https://gcc.gnu.org/g:7d8c44d397e1d6b754c6f08d7b6c95a312884b5b
commit r17-6227-g7d8c44d397e1d6b754c6f08d7b6c95a312884b5b Author: Dhruv Chawla <[email protected]> Date: Thu Nov 5 15:43:55 2026 +0530 aarch64: Port NEON shift and insert intrinsics to the pragma-based framework Port the following families of NEON intrinsics to the pragma-based framework: * vsli_n * vsri_n * vslid_n * vsrid_n Bootstrapped and regtested on aarch64-linux-gnu. Signed-off-by: Dhruv Chawla <[email protected]> gcc/ChangeLog: * config/aarch64/aarch64-builtins.cc (aarch64_types_ternop_p_imm_qualifiers, aarch64_types_unsigned_shiftacc_qualifiers, TYPES_SHIFTINSERTP, TYPES_USHIFTACC): Delete. * config/aarch64/aarch64-neon-builtins-base.cc (code_for_si_n): New function. (vsli_n, vsliq_n, vslid_n, vsri_n, vsriq_n, vsrid_n): New function bases. * config/aarch64/aarch64-neon-builtins-base.def: Add the corresponding function group. * config/aarch64/aarch64-simd-builtins.def: Delete the ssli_n, usli_n, ssri_n and usri_n builtins, which no longer have any caller. * config/aarch64/arm_neon.h (vsli_n_p16, vsli_n_p64, vsli_n_p8, vsli_n_s16, vsli_n_s32, vsli_n_s64, vsli_n_s8, vsli_n_u16, vsli_n_u32, vsli_n_u64, vsli_n_u8, vslid_n_s64, vslid_n_u64, vsliq_n_p16, vsliq_n_p64, vsliq_n_p8, vsliq_n_s16, vsliq_n_s32, vsliq_n_s64, vsliq_n_s8, vsliq_n_u16, vsliq_n_u32, vsliq_n_u64, vsliq_n_u8, vsri_n_p16, vsri_n_p64, vsri_n_p8, vsri_n_s16, vsri_n_s32, vsri_n_s64, vsri_n_s8, vsri_n_u16, vsri_n_u32, vsri_n_u64, vsri_n_u8, vsrid_n_s64, vsrid_n_u64, vsriq_n_p16, vsriq_n_p64, vsriq_n_p8, vsriq_n_s16, vsriq_n_s32, vsriq_n_s64, vsriq_n_s8, vsriq_n_u16, vsriq_n_u32, vsriq_n_u64, vsriq_n_u8): Delete. gcc/testsuite/ChangeLog: * gcc.target/aarch64/neon/vsli_n.c: New test. Diff: --- gcc/config/aarch64/aarch64-builtins.cc | 10 - gcc/config/aarch64/aarch64-neon-builtins-base.cc | 11 + gcc/config/aarch64/aarch64-neon-builtins-base.def | 14 + gcc/config/aarch64/aarch64-simd-builtins.def | 7 - gcc/config/aarch64/arm_neon.h | 340 ---------------------- gcc/testsuite/gcc.target/aarch64/neon/vsli_n.c | 66 +++++ 6 files changed, 91 insertions(+), 357 deletions(-) diff --git a/gcc/config/aarch64/aarch64-builtins.cc b/gcc/config/aarch64/aarch64-builtins.cc index 3acc0b470f01..8fc58ddc4d9a 100644 --- a/gcc/config/aarch64/aarch64-builtins.cc +++ b/gcc/config/aarch64/aarch64-builtins.cc @@ -367,16 +367,6 @@ aarch64_types_ternop_s_imm_qualifiers[SIMD_MAX_BUILTIN_ARGS] #define TYPES_SHIFTACC (aarch64_types_ternop_s_imm_qualifiers) #define TYPES_SHIFT2IMM (aarch64_types_ternop_s_imm_qualifiers) -static enum aarch64_type_qualifiers -aarch64_types_ternop_p_imm_qualifiers[SIMD_MAX_BUILTIN_ARGS] - = { qualifier_poly, qualifier_poly, qualifier_poly, qualifier_immediate}; -#define TYPES_SHIFTINSERTP (aarch64_types_ternop_p_imm_qualifiers) - -static enum aarch64_type_qualifiers -aarch64_types_unsigned_shiftacc_qualifiers[SIMD_MAX_BUILTIN_ARGS] - = { qualifier_unsigned, qualifier_unsigned, qualifier_unsigned, - qualifier_immediate }; -#define TYPES_USHIFTACC (aarch64_types_unsigned_shiftacc_qualifiers) static enum aarch64_type_qualifiers aarch64_types_load1_qualifiers[SIMD_MAX_BUILTIN_ARGS] diff --git a/gcc/config/aarch64/aarch64-neon-builtins-base.cc b/gcc/config/aarch64/aarch64-neon-builtins-base.cc index 87fa2807114c..4393fd184121 100644 --- a/gcc/config/aarch64/aarch64-neon-builtins-base.cc +++ b/gcc/config/aarch64/aarch64-neon-builtins-base.cc @@ -337,6 +337,13 @@ code_for_qshl_n (int unspec, machine_mode mode) return code_for_aarch64_qshl_n (unspec, unspec, mode); } +/* Likewise for `aarch64_<sur>s<lr>i_n<mode>`. */ +static insn_code +code_for_si_n (int unspec, machine_mode mode) +{ + return code_for_aarch64_si_n (unspec, unspec, mode); +} + /* For intrinsics whose instruction is selected by a mode through the code_for_* helpers that insn-opinit.h generates for `@`-prefixed pattterns. */ @@ -1660,6 +1667,10 @@ NEON_FUNCTION (vqshl_n, vqshlq_n, vqshlb_n, vqshlh_n, vqshls_n, vqshld_n, NEON_FUNCTION (vqshlu_n, vqshluq_n, vqshlub_n, vqshluh_n, vqshlus_n, vqshlud_n, gimple_unspec_based_function, (code_for_qshl_n, UNSPEC_SQSHLU, UNSPEC_SQSHLU)) +// Shift and insert +NEON_FUNCTION (vsli_n, vsliq_n, vslid_n, gimple_unspec_based_function, (code_for_si_n, UNSPEC_SSLI, UNSPEC_USLI)) +NEON_FUNCTION (vsri_n, vsriq_n, vsrid_n, gimple_unspec_based_function, (code_for_si_n, UNSPEC_SSRI, UNSPEC_USRI)) + // Bitwise operations NEON_FUNCTION (vand, vandq, gimple_expr, (BIT_AND_EXPR)) diff --git a/gcc/config/aarch64/aarch64-neon-builtins-base.def b/gcc/config/aarch64/aarch64-neon-builtins-base.def index f6edd2102e2d..735faf039adf 100644 --- a/gcc/config/aarch64/aarch64-neon-builtins-base.def +++ b/gcc/config/aarch64/aarch64-neon-builtins-base.def @@ -449,6 +449,20 @@ DEF_NEON_FUNCTION (vqshlus_n, s_signed, ("su0,s0,ss32", shift<1>)) DEF_NEON_FUNCTION (vqshlud_n, d_signed, ("su0,s0,ss32", shift<1>)) #undef REQUIRED_EXTENSIONS +// Shift and insert +#define REQUIRED_EXTENSIONS nonstreaming_only (AARCH64_FL_SIMD) +DEF_NEON_FUNCTION (vsli_n, all_integer, ("D0,D0,D0,ss32", shift<2>)) +DEF_NEON_FUNCTION (vsli_n, bhd_poly, ("D0,D0,D0,ss32", shift<2>)) +DEF_NEON_FUNCTION (vsliq_n, all_integer, ("Q0,Q0,Q0,ss32", shift<2>)) +DEF_NEON_FUNCTION (vsliq_n, bhd_poly, ("Q0,Q0,Q0,ss32", shift<2>)) +DEF_NEON_FUNCTION (vsri_n, all_integer, ("D0,D0,D0,ss32", shift_right<2>)) +DEF_NEON_FUNCTION (vsri_n, bhd_poly, ("D0,D0,D0,ss32", shift_right<2>)) +DEF_NEON_FUNCTION (vsriq_n, all_integer, ("Q0,Q0,Q0,ss32", shift_right<2>)) +DEF_NEON_FUNCTION (vsriq_n, bhd_poly, ("Q0,Q0,Q0,ss32", shift_right<2>)) +DEF_NEON_FUNCTION (vslid_n, d_integer, ("s0,s0,s0,ss32", shift<2>)) +DEF_NEON_FUNCTION (vsrid_n, d_integer, ("s0,s0,s0,ss32", shift_right<2>)) +#undef REQUIRED_EXTENSIONS + // Bitwise operations (FEAT_SHA3) #define REQUIRED_EXTENSIONS nonstreaming_only (AARCH64_FL_SHA3) DEF_NEON_FUNCTION (vbcaxq, all_integer, ("Q0,Q0,Q0,Q0")) diff --git a/gcc/config/aarch64/aarch64-simd-builtins.def b/gcc/config/aarch64/aarch64-simd-builtins.def index 909ea518b166..fa400a946f6a 100644 --- a/gcc/config/aarch64/aarch64-simd-builtins.def +++ b/gcc/config/aarch64/aarch64-simd-builtins.def @@ -432,13 +432,6 @@ BUILTIN_VQN (USHIFT2IMM, uqushrn2_n, 0, DEFAULT) BUILTIN_VQN (SHIFT2IMM, sqrshrn2_n, 0, DEFAULT) BUILTIN_VQN (USHIFT2IMM, uqrshrn2_n, 0, DEFAULT) - /* Implemented by aarch64_<sur>s<lr>i_n<mode>. */ - BUILTIN_VSDQ_I_DI (SHIFTINSERT, ssri_n, 0, DEFAULT) - BUILTIN_VALLP (SHIFTINSERTP, ssri_n, 0, DEFAULT) - BUILTIN_VSDQ_I_DI (USHIFTACC, usri_n, 0, DEFAULT) - BUILTIN_VSDQ_I_DI (SHIFTINSERT, ssli_n, 0, DEFAULT) - BUILTIN_VALLP (SHIFTINSERTP, ssli_n, 0, DEFAULT) - BUILTIN_VSDQ_I_DI (USHIFTACC, usli_n, 0, DEFAULT) /* Implemented by aarch64_xtn2<mode>. */ BUILTIN_VQN (BINOP, xtn2, 0, DEFAULT) diff --git a/gcc/config/aarch64/arm_neon.h b/gcc/config/aarch64/arm_neon.h index 68e4c2ef7569..e9238dfdcd3f 100644 --- a/gcc/config/aarch64/arm_neon.h +++ b/gcc/config/aarch64/arm_neon.h @@ -2899,76 +2899,6 @@ vshrn_high_n_u64 (uint32x2_t __a, uint64x2_t __b, const int __c) return __builtin_aarch64_ushrn2_nv2di_uuus (__a, __b, __c); } -__extension__ extern __inline poly8x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsli_n_p8 (poly8x8_t __a, poly8x8_t __b, const int __c) -{ - return __builtin_aarch64_ssli_nv8qi_ppps (__a, __b, __c); -} - -__extension__ extern __inline poly16x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsli_n_p16 (poly16x4_t __a, poly16x4_t __b, const int __c) -{ - return __builtin_aarch64_ssli_nv4hi_ppps (__a, __b, __c); -} - -__extension__ extern __inline poly8x16_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsliq_n_p8 (poly8x16_t __a, poly8x16_t __b, const int __c) -{ - return __builtin_aarch64_ssli_nv16qi_ppps (__a, __b, __c); -} - -__extension__ extern __inline poly16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsliq_n_p16 (poly16x8_t __a, poly16x8_t __b, const int __c) -{ - return __builtin_aarch64_ssli_nv8hi_ppps (__a, __b, __c); -} - -__extension__ extern __inline poly8x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsri_n_p8 (poly8x8_t __a, poly8x8_t __b, const int __c) -{ - return __builtin_aarch64_ssri_nv8qi_ppps (__a, __b, __c); -} - -__extension__ extern __inline poly16x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsri_n_p16 (poly16x4_t __a, poly16x4_t __b, const int __c) -{ - return __builtin_aarch64_ssri_nv4hi_ppps (__a, __b, __c); -} - -__extension__ extern __inline poly64x1_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsri_n_p64 (poly64x1_t __a, poly64x1_t __b, const int __c) -{ - return (poly64x1_t) __builtin_aarch64_ssri_ndi_ppps (__a[0], __b[0], __c); -} - -__extension__ extern __inline poly8x16_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsriq_n_p8 (poly8x16_t __a, poly8x16_t __b, const int __c) -{ - return __builtin_aarch64_ssri_nv16qi_ppps (__a, __b, __c); -} - -__extension__ extern __inline poly16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsriq_n_p16 (poly16x8_t __a, poly16x8_t __b, const int __c) -{ - return __builtin_aarch64_ssri_nv8hi_ppps (__a, __b, __c); -} - -__extension__ extern __inline poly64x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsriq_n_p64 (poly64x2_t __a, poly64x2_t __b, const int __c) -{ - return __builtin_aarch64_ssri_nv2di_ppps (__a, __b, __c); -} - __extension__ extern __inline uint8x8_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) vtst_p8 (poly8x8_t __a, poly8x8_t __b) @@ -11886,148 +11816,6 @@ vshll_n_u32 (uint32x2_t __a, const int __b) return __builtin_aarch64_ushll_nv2si_uus (__a, __b); } -/* vsli */ - -__extension__ extern __inline int8x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsli_n_s8 (int8x8_t __a, int8x8_t __b, const int __c) -{ - return (int8x8_t) __builtin_aarch64_ssli_nv8qi (__a, __b, __c); -} - -__extension__ extern __inline int16x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsli_n_s16 (int16x4_t __a, int16x4_t __b, const int __c) -{ - return (int16x4_t) __builtin_aarch64_ssli_nv4hi (__a, __b, __c); -} - -__extension__ extern __inline int32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsli_n_s32 (int32x2_t __a, int32x2_t __b, const int __c) -{ - return (int32x2_t) __builtin_aarch64_ssli_nv2si (__a, __b, __c); -} - -__extension__ extern __inline int64x1_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsli_n_s64 (int64x1_t __a, int64x1_t __b, const int __c) -{ - return (int64x1_t) {__builtin_aarch64_ssli_ndi (__a[0], __b[0], __c)}; -} - -__extension__ extern __inline uint8x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsli_n_u8 (uint8x8_t __a, uint8x8_t __b, const int __c) -{ - return __builtin_aarch64_usli_nv8qi_uuus (__a, __b, __c); -} - -__extension__ extern __inline uint16x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsli_n_u16 (uint16x4_t __a, uint16x4_t __b, const int __c) -{ - return __builtin_aarch64_usli_nv4hi_uuus (__a, __b, __c); -} - -__extension__ extern __inline uint32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsli_n_u32 (uint32x2_t __a, uint32x2_t __b, const int __c) -{ - return __builtin_aarch64_usli_nv2si_uuus (__a, __b, __c); -} - -__extension__ extern __inline uint64x1_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsli_n_u64 (uint64x1_t __a, uint64x1_t __b, const int __c) -{ - return (uint64x1_t) {__builtin_aarch64_usli_ndi_uuus (__a[0], __b[0], __c)}; -} - -__extension__ extern __inline poly64x1_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsli_n_p64 (poly64x1_t __a, poly64x1_t __b, const int __c) -{ - return (poly64x1_t) {__builtin_aarch64_ssli_ndi_ppps (__a[0], __b[0], __c)}; -} - -__extension__ extern __inline int8x16_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsliq_n_s8 (int8x16_t __a, int8x16_t __b, const int __c) -{ - return (int8x16_t) __builtin_aarch64_ssli_nv16qi (__a, __b, __c); -} - -__extension__ extern __inline int16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsliq_n_s16 (int16x8_t __a, int16x8_t __b, const int __c) -{ - return (int16x8_t) __builtin_aarch64_ssli_nv8hi (__a, __b, __c); -} - -__extension__ extern __inline int32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsliq_n_s32 (int32x4_t __a, int32x4_t __b, const int __c) -{ - return (int32x4_t) __builtin_aarch64_ssli_nv4si (__a, __b, __c); -} - -__extension__ extern __inline int64x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsliq_n_s64 (int64x2_t __a, int64x2_t __b, const int __c) -{ - return (int64x2_t) __builtin_aarch64_ssli_nv2di (__a, __b, __c); -} - -__extension__ extern __inline uint8x16_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsliq_n_u8 (uint8x16_t __a, uint8x16_t __b, const int __c) -{ - return __builtin_aarch64_usli_nv16qi_uuus (__a, __b, __c); -} - -__extension__ extern __inline uint16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsliq_n_u16 (uint16x8_t __a, uint16x8_t __b, const int __c) -{ - return __builtin_aarch64_usli_nv8hi_uuus (__a, __b, __c); -} - -__extension__ extern __inline uint32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsliq_n_u32 (uint32x4_t __a, uint32x4_t __b, const int __c) -{ - return __builtin_aarch64_usli_nv4si_uuus (__a, __b, __c); -} - -__extension__ extern __inline uint64x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsliq_n_u64 (uint64x2_t __a, uint64x2_t __b, const int __c) -{ - return __builtin_aarch64_usli_nv2di_uuus (__a, __b, __c); -} - -__extension__ extern __inline poly64x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsliq_n_p64 (poly64x2_t __a, poly64x2_t __b, const int __c) -{ - return __builtin_aarch64_ssli_nv2di_ppps (__a, __b, __c); -} - -__extension__ extern __inline int64_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vslid_n_s64 (int64_t __a, int64_t __b, const int __c) -{ - return __builtin_aarch64_ssli_ndi (__a, __b, __c); -} - -__extension__ extern __inline uint64_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vslid_n_u64 (uint64_t __a, uint64_t __b, const int __c) -{ - return __builtin_aarch64_usli_ndi_uuus (__a, __b, __c); -} - /* vsqadd */ __extension__ extern __inline uint8x8_t @@ -12114,134 +11902,6 @@ vsqaddd_u64 (uint64_t __a, int64_t __b) return __builtin_aarch64_usqadddi_uus (__a, __b); } -/* vsri */ - -__extension__ extern __inline int8x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsri_n_s8 (int8x8_t __a, int8x8_t __b, const int __c) -{ - return (int8x8_t) __builtin_aarch64_ssri_nv8qi (__a, __b, __c); -} - -__extension__ extern __inline int16x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsri_n_s16 (int16x4_t __a, int16x4_t __b, const int __c) -{ - return (int16x4_t) __builtin_aarch64_ssri_nv4hi (__a, __b, __c); -} - -__extension__ extern __inline int32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsri_n_s32 (int32x2_t __a, int32x2_t __b, const int __c) -{ - return (int32x2_t) __builtin_aarch64_ssri_nv2si (__a, __b, __c); -} - -__extension__ extern __inline int64x1_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsri_n_s64 (int64x1_t __a, int64x1_t __b, const int __c) -{ - return (int64x1_t) {__builtin_aarch64_ssri_ndi (__a[0], __b[0], __c)}; -} - -__extension__ extern __inline uint8x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsri_n_u8 (uint8x8_t __a, uint8x8_t __b, const int __c) -{ - return __builtin_aarch64_usri_nv8qi_uuus (__a, __b, __c); -} - -__extension__ extern __inline uint16x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsri_n_u16 (uint16x4_t __a, uint16x4_t __b, const int __c) -{ - return __builtin_aarch64_usri_nv4hi_uuus (__a, __b, __c); -} - -__extension__ extern __inline uint32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsri_n_u32 (uint32x2_t __a, uint32x2_t __b, const int __c) -{ - return __builtin_aarch64_usri_nv2si_uuus (__a, __b, __c); -} - -__extension__ extern __inline uint64x1_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsri_n_u64 (uint64x1_t __a, uint64x1_t __b, const int __c) -{ - return (uint64x1_t) {__builtin_aarch64_usri_ndi_uuus (__a[0], __b[0], __c)}; -} - -__extension__ extern __inline int8x16_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsriq_n_s8 (int8x16_t __a, int8x16_t __b, const int __c) -{ - return (int8x16_t) __builtin_aarch64_ssri_nv16qi (__a, __b, __c); -} - -__extension__ extern __inline int16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsriq_n_s16 (int16x8_t __a, int16x8_t __b, const int __c) -{ - return (int16x8_t) __builtin_aarch64_ssri_nv8hi (__a, __b, __c); -} - -__extension__ extern __inline int32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsriq_n_s32 (int32x4_t __a, int32x4_t __b, const int __c) -{ - return (int32x4_t) __builtin_aarch64_ssri_nv4si (__a, __b, __c); -} - -__extension__ extern __inline int64x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsriq_n_s64 (int64x2_t __a, int64x2_t __b, const int __c) -{ - return (int64x2_t) __builtin_aarch64_ssri_nv2di (__a, __b, __c); -} - -__extension__ extern __inline uint8x16_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsriq_n_u8 (uint8x16_t __a, uint8x16_t __b, const int __c) -{ - return __builtin_aarch64_usri_nv16qi_uuus (__a, __b, __c); -} - -__extension__ extern __inline uint16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsriq_n_u16 (uint16x8_t __a, uint16x8_t __b, const int __c) -{ - return __builtin_aarch64_usri_nv8hi_uuus (__a, __b, __c); -} - -__extension__ extern __inline uint32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsriq_n_u32 (uint32x4_t __a, uint32x4_t __b, const int __c) -{ - return __builtin_aarch64_usri_nv4si_uuus (__a, __b, __c); -} - -__extension__ extern __inline uint64x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsriq_n_u64 (uint64x2_t __a, uint64x2_t __b, const int __c) -{ - return __builtin_aarch64_usri_nv2di_uuus (__a, __b, __c); -} - -__extension__ extern __inline int64_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsrid_n_s64 (int64_t __a, int64_t __b, const int __c) -{ - return __builtin_aarch64_ssri_ndi (__a, __b, __c); -} - -__extension__ extern __inline uint64_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vsrid_n_u64 (uint64_t __a, uint64_t __b, const int __c) -{ - return __builtin_aarch64_usri_ndi_uuus (__a, __b, __c); -} - /* vst1 */ __extension__ extern __inline void diff --git a/gcc/testsuite/gcc.target/aarch64/neon/vsli_n.c b/gcc/testsuite/gcc.target/aarch64/neon/vsli_n.c new file mode 100644 index 000000000000..af833470f2e4 --- /dev/null +++ b/gcc/testsuite/gcc.target/aarch64/neon/vsli_n.c @@ -0,0 +1,66 @@ +/* { dg-do compile } */ +/* { dg-final { check-function-bodies "**" "" } } */ + +#include "arm_neon_test.h" + +/* +** test_vsli_n_s8_3: +** sli v0\.8b, v1\.8b, 3 +** ret +*/ +TEST_UNIFORM_BINARY_N (vsli_n_s8, int8x8_t, 3) + +/* +** test_vsliq_n_p64_63: +** sli v0\.2d, v1\.2d, 63 +** ret +*/ +TEST_UNIFORM_BINARY_N (vsliq_n_p64, poly64x2_t, 63) + +/* +** test_vsli_n_u64_7: +** sli d0, d1, 7 +** ret +*/ +TEST_UNIFORM_BINARY_N (vsli_n_u64, uint64x1_t, 7) + +/* +** test_vsri_n_p8_3: +** sri v0\.8b, v1\.8b, 3 +** ret +*/ +TEST_UNIFORM_BINARY_N (vsri_n_p8, poly8x8_t, 3) + +/* +** test_vsriq_n_u64_64: +** sri v0\.2d, v1\.2d, 64 +** ret +*/ +TEST_UNIFORM_BINARY_N (vsriq_n_u64, uint64x2_t, 64) + +/* +** test_vsri_n_s64_7: +** sri d0, d1, 7 +** ret +*/ +TEST_UNIFORM_BINARY_N (vsri_n_s64, int64x1_t, 7) + +/* +** test_vslid_n_s64_7: +** fmov d[0-9]+, x0 +** fmov d[0-9]+, x1 +** sli d[0-9]+, d[0-9]+, 7 +** fmov x0, d[0-9]+ +** ret +*/ +TEST_UNIFORM_BINARY_N (vslid_n_s64, int64_t, 7) + +/* +** test_vsrid_n_u64_7: +** fmov d[0-9]+, x0 +** fmov d[0-9]+, x1 +** sri d[0-9]+, d[0-9]+, 7 +** fmov x0, d[0-9]+ +** ret +*/ +TEST_UNIFORM_BINARY_N (vsrid_n_u64, uint64_t, 7)
