Implement vqshluq using the new MVE builtins framework. 2022-12-12 Christophe Lyon <christophe.l...@arm.com>
gcc/ * config/arm/arm-mve-builtins-base.cc (vqshluq): New. * config/arm/arm-mve-builtins-base.def (vqshluq): New. * config/arm/arm-mve-builtins-base.h (vqshluq): New. * config/arm/arm_mve.h (vqshluq): Remove. (vqshluq_m): Remove. (vqshluq_n_s8): Remove. (vqshluq_n_s16): Remove. (vqshluq_n_s32): Remove. (vqshluq_m_n_s8): Remove. (vqshluq_m_n_s16): Remove. (vqshluq_m_n_s32): Remove. (__arm_vqshluq_n_s8): Remove. (__arm_vqshluq_n_s16): Remove. (__arm_vqshluq_n_s32): Remove. (__arm_vqshluq_m_n_s8): Remove. (__arm_vqshluq_m_n_s16): Remove. (__arm_vqshluq_m_n_s32): Remove. (__arm_vqshluq): Remove. (__arm_vqshluq_m): Remove. --- gcc/config/arm/arm-mve-builtins-base.cc | 1 + gcc/config/arm/arm-mve-builtins-base.def | 1 + gcc/config/arm/arm-mve-builtins-base.h | 1 + gcc/config/arm/arm_mve.h | 111 ----------------------- 4 files changed, 3 insertions(+), 111 deletions(-) diff --git a/gcc/config/arm/arm-mve-builtins-base.cc b/gcc/config/arm/arm-mve-builtins-base.cc index a2b227bb2aa..739ab604843 100644 --- a/gcc/config/arm/arm-mve-builtins-base.cc +++ b/gcc/config/arm/arm-mve-builtins-base.cc @@ -332,6 +332,7 @@ FUNCTION_WITHOUT_N_NO_U_F (vqnegq, VQNEGQ) FUNCTION_WITH_M_N_NO_F (vqrshlq, VQRSHLQ) FUNCTION_WITH_M_N_NO_U_F (vqrdmulhq, VQRDMULHQ) FUNCTION_WITH_M_N_R (vqshlq, VQSHLQ) +FUNCTION_ONLY_N_NO_U_F (vqshluq, VQSHLUQ) FUNCTION_ONLY_N_NO_F (vqrshrnbq, VQRSHRNBQ) FUNCTION_ONLY_N_NO_F (vqrshrntq, VQRSHRNTQ) FUNCTION_ONLY_N_NO_U_F (vqrshrunbq, VQRSHRUNBQ) diff --git a/gcc/config/arm/arm-mve-builtins-base.def b/gcc/config/arm/arm-mve-builtins-base.def index c4ef74169dd..3f7bb414e40 100644 --- a/gcc/config/arm/arm-mve-builtins-base.def +++ b/gcc/config/arm/arm-mve-builtins-base.def @@ -106,6 +106,7 @@ DEF_MVE_FUNCTION (vqrshrunbq, binary_rshift_narrow_unsigned, signed_16_32, m_or_ DEF_MVE_FUNCTION (vqrshruntq, binary_rshift_narrow_unsigned, signed_16_32, m_or_none) DEF_MVE_FUNCTION (vqshlq, binary_lshift, all_integer, m_or_none) DEF_MVE_FUNCTION (vqshlq, binary_lshift_r, all_integer, m_or_none) +DEF_MVE_FUNCTION (vqshluq, binary_lshift_unsigned, all_signed, m_or_none) DEF_MVE_FUNCTION (vqshrnbq, binary_rshift_narrow, integer_16_32, m_or_none) DEF_MVE_FUNCTION (vqshrntq, binary_rshift_narrow, integer_16_32, m_or_none) DEF_MVE_FUNCTION (vqshrunbq, binary_rshift_narrow_unsigned, signed_16_32, m_or_none) diff --git a/gcc/config/arm/arm-mve-builtins-base.h b/gcc/config/arm/arm-mve-builtins-base.h index 41b2e19c2d7..797f8ba2f5e 100644 --- a/gcc/config/arm/arm-mve-builtins-base.h +++ b/gcc/config/arm/arm-mve-builtins-base.h @@ -118,6 +118,7 @@ extern const function_base *const vqrshrntq; extern const function_base *const vqrshrunbq; extern const function_base *const vqrshruntq; extern const function_base *const vqshlq; +extern const function_base *const vqshluq; extern const function_base *const vqshrnbq; extern const function_base *const vqshrntq; extern const function_base *const vqshrunbq; diff --git a/gcc/config/arm/arm_mve.h b/gcc/config/arm/arm_mve.h index c995093e12f..673a3df1bfd 100644 --- a/gcc/config/arm/arm_mve.h +++ b/gcc/config/arm/arm_mve.h @@ -50,7 +50,6 @@ #define vcaddq_rot270(__a, __b) __arm_vcaddq_rot270(__a, __b) #define vbicq(__a, __b) __arm_vbicq(__a, __b) #define vbrsrq(__a, __b) __arm_vbrsrq(__a, __b) -#define vqshluq(__a, __imm) __arm_vqshluq(__a, __imm) #define vhcaddq_rot90(__a, __b) __arm_vhcaddq_rot90(__a, __b) #define vhcaddq_rot270(__a, __b) __arm_vhcaddq_rot270(__a, __b) #define vmulltq_poly(__a, __b) __arm_vmulltq_poly(__a, __b) @@ -62,7 +61,6 @@ #define vsriq(__a, __b, __imm) __arm_vsriq(__a, __b, __imm) #define vsliq(__a, __b, __imm) __arm_vsliq(__a, __b, __imm) #define vsriq_m(__a, __b, __imm, __p) __arm_vsriq_m(__a, __b, __imm, __p) -#define vqshluq_m(__inactive, __a, __imm, __p) __arm_vqshluq_m(__inactive, __a, __imm, __p) #define vbicq_m(__inactive, __a, __b, __p) __arm_vbicq_m(__inactive, __a, __b, __p) #define vbrsrq_m(__inactive, __a, __b, __p) __arm_vbrsrq_m(__inactive, __a, __b, __p) #define vcaddq_rot270_m(__inactive, __a, __b, __p) __arm_vcaddq_rot270_m(__inactive, __a, __b, __p) @@ -284,7 +282,6 @@ #define vcaddq_rot270_u8(__a, __b) __arm_vcaddq_rot270_u8(__a, __b) #define vbicq_u8(__a, __b) __arm_vbicq_u8(__a, __b) #define vbrsrq_n_u8(__a, __b) __arm_vbrsrq_n_u8(__a, __b) -#define vqshluq_n_s8(__a, __imm) __arm_vqshluq_n_s8(__a, __imm) #define vornq_s8(__a, __b) __arm_vornq_s8(__a, __b) #define vmulltq_int_s8(__a, __b) __arm_vmulltq_int_s8(__a, __b) #define vmullbq_int_s8(__a, __b) __arm_vmullbq_int_s8(__a, __b) @@ -301,7 +298,6 @@ #define vcaddq_rot270_u16(__a, __b) __arm_vcaddq_rot270_u16(__a, __b) #define vbicq_u16(__a, __b) __arm_vbicq_u16(__a, __b) #define vbrsrq_n_u16(__a, __b) __arm_vbrsrq_n_u16(__a, __b) -#define vqshluq_n_s16(__a, __imm) __arm_vqshluq_n_s16(__a, __imm) #define vornq_s16(__a, __b) __arm_vornq_s16(__a, __b) #define vmulltq_int_s16(__a, __b) __arm_vmulltq_int_s16(__a, __b) #define vmullbq_int_s16(__a, __b) __arm_vmullbq_int_s16(__a, __b) @@ -318,7 +314,6 @@ #define vcaddq_rot270_u32(__a, __b) __arm_vcaddq_rot270_u32(__a, __b) #define vbicq_u32(__a, __b) __arm_vbicq_u32(__a, __b) #define vbrsrq_n_u32(__a, __b) __arm_vbrsrq_n_u32(__a, __b) -#define vqshluq_n_s32(__a, __imm) __arm_vqshluq_n_s32(__a, __imm) #define vornq_s32(__a, __b) __arm_vornq_s32(__a, __b) #define vmulltq_int_s32(__a, __b) __arm_vmulltq_int_s32(__a, __b) #define vmullbq_int_s32(__a, __b) __arm_vmullbq_int_s32(__a, __b) @@ -446,16 +441,13 @@ #define vcvtq_m_u32_f32(__inactive, __a, __p) __arm_vcvtq_m_u32_f32(__inactive, __a, __p) #define vsriq_m_n_s8(__a, __b, __imm, __p) __arm_vsriq_m_n_s8(__a, __b, __imm, __p) #define vcvtq_m_n_f16_u16(__inactive, __a, __imm6, __p) __arm_vcvtq_m_n_f16_u16(__inactive, __a, __imm6, __p) -#define vqshluq_m_n_s8(__inactive, __a, __imm, __p) __arm_vqshluq_m_n_s8(__inactive, __a, __imm, __p) #define vsriq_m_n_u8(__a, __b, __imm, __p) __arm_vsriq_m_n_u8(__a, __b, __imm, __p) #define vcvtq_m_n_f16_s16(__inactive, __a, __imm6, __p) __arm_vcvtq_m_n_f16_s16(__inactive, __a, __imm6, __p) #define vsriq_m_n_s16(__a, __b, __imm, __p) __arm_vsriq_m_n_s16(__a, __b, __imm, __p) #define vcvtq_m_n_f32_u32(__inactive, __a, __imm6, __p) __arm_vcvtq_m_n_f32_u32(__inactive, __a, __imm6, __p) -#define vqshluq_m_n_s16(__inactive, __a, __imm, __p) __arm_vqshluq_m_n_s16(__inactive, __a, __imm, __p) #define vsriq_m_n_u16(__a, __b, __imm, __p) __arm_vsriq_m_n_u16(__a, __b, __imm, __p) #define vcvtq_m_n_f32_s32(__inactive, __a, __imm6, __p) __arm_vcvtq_m_n_f32_s32(__inactive, __a, __imm6, __p) #define vsriq_m_n_s32(__a, __b, __imm, __p) __arm_vsriq_m_n_s32(__a, __b, __imm, __p) -#define vqshluq_m_n_s32(__inactive, __a, __imm, __p) __arm_vqshluq_m_n_s32(__inactive, __a, __imm, __p) #define vsriq_m_n_u32(__a, __b, __imm, __p) __arm_vsriq_m_n_u32(__a, __b, __imm, __p) #define vbicq_m_s8(__inactive, __a, __b, __p) __arm_vbicq_m_s8(__inactive, __a, __b, __p) #define vbicq_m_s32(__inactive, __a, __b, __p) __arm_vbicq_m_s32(__inactive, __a, __b, __p) @@ -1287,13 +1279,6 @@ __arm_vbrsrq_n_u8 (uint8x16_t __a, int32_t __b) return __builtin_mve_vbrsrq_n_uv16qi (__a, __b); } -__extension__ extern __inline uint8x16_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -__arm_vqshluq_n_s8 (int8x16_t __a, const int __imm) -{ - return __builtin_mve_vqshluq_n_sv16qi (__a, __imm); -} - __extension__ extern __inline int8x16_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) __arm_vornq_s8 (int8x16_t __a, int8x16_t __b) @@ -1408,13 +1393,6 @@ __arm_vbrsrq_n_u16 (uint16x8_t __a, int32_t __b) return __builtin_mve_vbrsrq_n_uv8hi (__a, __b); } -__extension__ extern __inline uint16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -__arm_vqshluq_n_s16 (int16x8_t __a, const int __imm) -{ - return __builtin_mve_vqshluq_n_sv8hi (__a, __imm); -} - __extension__ extern __inline int16x8_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) __arm_vornq_s16 (int16x8_t __a, int16x8_t __b) @@ -1529,13 +1507,6 @@ __arm_vbrsrq_n_u32 (uint32x4_t __a, int32_t __b) return __builtin_mve_vbrsrq_n_uv4si (__a, __b); } -__extension__ extern __inline uint32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -__arm_vqshluq_n_s32 (int32x4_t __a, const int __imm) -{ - return __builtin_mve_vqshluq_n_sv4si (__a, __imm); -} - __extension__ extern __inline int32x4_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) __arm_vornq_s32 (int32x4_t __a, int32x4_t __b) @@ -1982,13 +1953,6 @@ __arm_vsriq_m_n_s8 (int8x16_t __a, int8x16_t __b, const int __imm, mve_pred16_t return __builtin_mve_vsriq_m_n_sv16qi (__a, __b, __imm, __p); } -__extension__ extern __inline uint8x16_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -__arm_vqshluq_m_n_s8 (uint8x16_t __inactive, int8x16_t __a, const int __imm, mve_pred16_t __p) -{ - return __builtin_mve_vqshluq_m_n_sv16qi (__inactive, __a, __imm, __p); -} - __extension__ extern __inline uint8x16_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) __arm_vsriq_m_n_u8 (uint8x16_t __a, uint8x16_t __b, const int __imm, mve_pred16_t __p) @@ -2003,13 +1967,6 @@ __arm_vsriq_m_n_s16 (int16x8_t __a, int16x8_t __b, const int __imm, mve_pred16_t return __builtin_mve_vsriq_m_n_sv8hi (__a, __b, __imm, __p); } -__extension__ extern __inline uint16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -__arm_vqshluq_m_n_s16 (uint16x8_t __inactive, int16x8_t __a, const int __imm, mve_pred16_t __p) -{ - return __builtin_mve_vqshluq_m_n_sv8hi (__inactive, __a, __imm, __p); -} - __extension__ extern __inline uint16x8_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) __arm_vsriq_m_n_u16 (uint16x8_t __a, uint16x8_t __b, const int __imm, mve_pred16_t __p) @@ -2024,13 +1981,6 @@ __arm_vsriq_m_n_s32 (int32x4_t __a, int32x4_t __b, const int __imm, mve_pred16_t return __builtin_mve_vsriq_m_n_sv4si (__a, __b, __imm, __p); } -__extension__ extern __inline uint32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -__arm_vqshluq_m_n_s32 (uint32x4_t __inactive, int32x4_t __a, const int __imm, mve_pred16_t __p) -{ - return __builtin_mve_vqshluq_m_n_sv4si (__inactive, __a, __imm, __p); -} - __extension__ extern __inline uint32x4_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) __arm_vsriq_m_n_u32 (uint32x4_t __a, uint32x4_t __b, const int __imm, mve_pred16_t __p) @@ -7377,13 +7327,6 @@ __arm_vbrsrq (uint8x16_t __a, int32_t __b) return __arm_vbrsrq_n_u8 (__a, __b); } -__extension__ extern __inline uint8x16_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -__arm_vqshluq (int8x16_t __a, const int __imm) -{ - return __arm_vqshluq_n_s8 (__a, __imm); -} - __extension__ extern __inline int8x16_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) __arm_vornq (int8x16_t __a, int8x16_t __b) @@ -7496,13 +7439,6 @@ __arm_vbrsrq (uint16x8_t __a, int32_t __b) return __arm_vbrsrq_n_u16 (__a, __b); } -__extension__ extern __inline uint16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -__arm_vqshluq (int16x8_t __a, const int __imm) -{ - return __arm_vqshluq_n_s16 (__a, __imm); -} - __extension__ extern __inline int16x8_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) __arm_vornq (int16x8_t __a, int16x8_t __b) @@ -7615,13 +7551,6 @@ __arm_vbrsrq (uint32x4_t __a, int32_t __b) return __arm_vbrsrq_n_u32 (__a, __b); } -__extension__ extern __inline uint32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -__arm_vqshluq (int32x4_t __a, const int __imm) -{ - return __arm_vqshluq_n_s32 (__a, __imm); -} - __extension__ extern __inline int32x4_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) __arm_vornq (int32x4_t __a, int32x4_t __b) @@ -8028,13 +7957,6 @@ __arm_vsriq_m (int8x16_t __a, int8x16_t __b, const int __imm, mve_pred16_t __p) return __arm_vsriq_m_n_s8 (__a, __b, __imm, __p); } -__extension__ extern __inline uint8x16_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -__arm_vqshluq_m (uint8x16_t __inactive, int8x16_t __a, const int __imm, mve_pred16_t __p) -{ - return __arm_vqshluq_m_n_s8 (__inactive, __a, __imm, __p); -} - __extension__ extern __inline uint8x16_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) __arm_vsriq_m (uint8x16_t __a, uint8x16_t __b, const int __imm, mve_pred16_t __p) @@ -8049,13 +7971,6 @@ __arm_vsriq_m (int16x8_t __a, int16x8_t __b, const int __imm, mve_pred16_t __p) return __arm_vsriq_m_n_s16 (__a, __b, __imm, __p); } -__extension__ extern __inline uint16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -__arm_vqshluq_m (uint16x8_t __inactive, int16x8_t __a, const int __imm, mve_pred16_t __p) -{ - return __arm_vqshluq_m_n_s16 (__inactive, __a, __imm, __p); -} - __extension__ extern __inline uint16x8_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) __arm_vsriq_m (uint16x8_t __a, uint16x8_t __b, const int __imm, mve_pred16_t __p) @@ -8070,13 +7985,6 @@ __arm_vsriq_m (int32x4_t __a, int32x4_t __b, const int __imm, mve_pred16_t __p) return __arm_vsriq_m_n_s32 (__a, __b, __imm, __p); } -__extension__ extern __inline uint32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -__arm_vqshluq_m (uint32x4_t __inactive, int32x4_t __a, const int __imm, mve_pred16_t __p) -{ - return __arm_vqshluq_m_n_s32 (__inactive, __a, __imm, __p); -} - __extension__ extern __inline uint32x4_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) __arm_vsriq_m (uint32x4_t __a, uint32x4_t __b, const int __imm, mve_pred16_t __p) @@ -12553,12 +12461,6 @@ extern void *__ARM_undef; int (*)[__ARM_mve_type_float16x8_t]: __arm_vbrsrq_n_f16 (__ARM_mve_coerce(__p0, float16x8_t), p1), \ int (*)[__ARM_mve_type_float32x4_t]: __arm_vbrsrq_n_f32 (__ARM_mve_coerce(__p0, float32x4_t), p1));}) -#define __arm_vqshluq(p0,p1) ({ __typeof(p0) __p0 = (p0); \ - _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \ - int (*)[__ARM_mve_type_int8x16_t]: __arm_vqshluq_n_s8 (__ARM_mve_coerce(__p0, int8x16_t), p1), \ - int (*)[__ARM_mve_type_int16x8_t]: __arm_vqshluq_n_s16 (__ARM_mve_coerce(__p0, int16x8_t), p1), \ - int (*)[__ARM_mve_type_int32x4_t]: __arm_vqshluq_n_s32 (__ARM_mve_coerce(__p0, int32x4_t), p1));}) - #define __arm_vmulltq_poly(p0,p1) ({ __typeof(p0) __p0 = (p0); \ __typeof(p1) __p1 = (p1); \ _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p1)])0, \ @@ -13383,12 +13285,6 @@ extern void *__ARM_undef; int (*)[__ARM_mve_type_uint16x8_t]: __arm_vmvnq_u16 (__ARM_mve_coerce(__p0, uint16x8_t)), \ int (*)[__ARM_mve_type_uint32x4_t]: __arm_vmvnq_u32 (__ARM_mve_coerce(__p0, uint32x4_t)));}) -#define __arm_vqshluq(p0,p1) ({ __typeof(p0) __p0 = (p0); \ - _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \ - int (*)[__ARM_mve_type_int8x16_t]: __arm_vqshluq_n_s8 (__ARM_mve_coerce(__p0, int8x16_t), p1), \ - int (*)[__ARM_mve_type_int16x8_t]: __arm_vqshluq_n_s16 (__ARM_mve_coerce(__p0, int16x8_t), p1), \ - int (*)[__ARM_mve_type_int32x4_t]: __arm_vqshluq_n_s32 (__ARM_mve_coerce(__p0, int32x4_t), p1));}) - #define __arm_vornq(p0,p1) ({ __typeof(p0) __p0 = (p0); \ __typeof(p1) __p1 = (p1); \ _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p1)])0, \ @@ -14161,13 +14057,6 @@ extern void *__ARM_undef; int (*)[__ARM_mve_type_uint16x8_t][__ARM_mve_type_int_n]: __arm_vmvnq_m_n_u16 (__ARM_mve_coerce(__p0, uint16x8_t), __ARM_mve_coerce1(__p1, int) , p2), \ int (*)[__ARM_mve_type_uint32x4_t][__ARM_mve_type_int_n]: __arm_vmvnq_m_n_u32 (__ARM_mve_coerce(__p0, uint32x4_t), __ARM_mve_coerce1(__p1, int) , p2));}) -#define __arm_vqshluq_m(p0,p1,p2,p3) ({ __typeof(p0) __p0 = (p0); \ - __typeof(p1) __p1 = (p1); \ - _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p1)])0, \ - int (*)[__ARM_mve_type_uint8x16_t][__ARM_mve_type_int8x16_t]: __arm_vqshluq_m_n_s8 (__ARM_mve_coerce(__p0, uint8x16_t), __ARM_mve_coerce(__p1, int8x16_t), p2, p3), \ - int (*)[__ARM_mve_type_uint16x8_t][__ARM_mve_type_int16x8_t]: __arm_vqshluq_m_n_s16 (__ARM_mve_coerce(__p0, uint16x8_t), __ARM_mve_coerce(__p1, int16x8_t), p2, p3), \ - int (*)[__ARM_mve_type_uint32x4_t][__ARM_mve_type_int32x4_t]: __arm_vqshluq_m_n_s32 (__ARM_mve_coerce(__p0, uint32x4_t), __ARM_mve_coerce(__p1, int32x4_t), p2, p3));}) - #define __arm_vsriq_m(p0,p1,p2,p3) ({ __typeof(p0) __p0 = (p0); \ __typeof(p1) __p1 = (p1); \ _Generic( (int (*)[__ARM_mve_typeid(__p0)][__ARM_mve_typeid(__p1)])0, \ -- 2.34.1