[PATCH v2 2/6] aarch64: Port NEON reduction intrinsics to pragma-based framework
Dhruv Chawla
dhruvc@nvidia.com
Tue Oct 6 04:33:46 GMT 2026
On 15/09/26 15:04, Kyrylo Tkachov wrote:
>
>
>> On 15 Sep 2026, at 10:12, Dhruv Chawla <dhruvc@nvidia.com> wrote:
>>
>> From: Dhruv Chawla <dhruvc@nvidia.com>
>>
>> Port the following intrinsics to the pragma-based framework:
>> * vaddv
>> * vmaxv
>> * vminv
>> * vmaxnmv
>> * vminnmv
>>
>> The following are not ported in this patch as they cannot be directly
>> lowered from an IFN through an existing instruction pattern:
>> * vmaxv_f16
>> * vmaxv_f32
>> * vmaxvq_f16
>> * vmaxvq_f32
>> * vmaxvq_f64
>> * vminv_f16
>> * vminv_f32
>> * vminvq_f16
>> * vminvq_f32
>> * vminvq_f64
>>
>> This is due to the NaN semantics of these intrinsics which require
>> mapping to <optab>_nan variants.
>>
>
> I think these, as well as the vsqrt, vrnd, vmaxnm, vminnm from other patches in this series suffer from the non-LHS ICE. Please include no-LHS testcases for them and add the temp LHS fix to them as well.
Hi Kyrill,
Now that Karl's fix has gone in as r17-4816-gb24418a2212866, is this
series okay to commit? I've rebased and tested on trunk, no regressions.
--
Regards,
Dhruv
> Thanks,
> Kyrill
>
>> Bootstrapped and regtested on aarch64-linux-gnu.
>>
>> Signed-off-by: Dhruv Chawla <dhruvc@nvidia.com>
>>
>> gcc/ChangeLog:
>>
>> * config/aarch64/aarch64-acle-builtins.h (TYPES_s_float_bhs_integer,
>> TYPES_sd_float_all_integer, s_float_bhs_integer,
>> sd_float_all_integer): New type lists.
>> * config/aarch64/aarch64-neon-builtins-base.cc (vaddv, vaddvq, vmaxv,
>> vmaxvq, vminv, vminvq, vmaxnmv, vmaxnmvq, vminnmv, vminnmvq):
>> New function bases.
>> * config/aarch64/aarch64-neon-builtins-base.def (vaddv, vaddvq, vmaxv,
>> vmaxvq, vminv, vminvq, vmaxnmv, vmaxnmvq, vminnmv, vminnmvq):
>> New function groups.
>> * config/aarch64/arm_neon.h (vaddv_s8, vaddv_s16, vaddv_s32, vaddv_u8,
>> vaddv_u16, vaddv_u32, vaddvq_s8, vaddvq_s16, vaddvq_s32, vaddvq_s64,
>> vaddvq_u8, vaddvq_u16, vaddvq_u32, vaddvq_u64, vaddv_f32, vaddvq_f32,
>> vaddvq_f64, vmaxv_s8, vmaxv_s16, vmaxv_s32, vmaxv_u8, vmaxv_u16,
>> vmaxv_u32, vmaxvq_s8, vmaxvq_s16, vmaxvq_s32, vmaxvq_u8, vmaxvq_u16,
>> vmaxvq_u32, vmaxnmv_f32, vmaxnmvq_f32, vmaxnmvq_f64, vminv_s8,
>> vminv_s16, vminv_s32, vminv_u8, vminv_u16, vminv_u32, vminvq_s8,
>> vminvq_s16, vminvq_s32, vminvq_u8, vminvq_u16, vminvq_u32, vminnmv_f32,
>> vminnmvq_f32, vminnmvq_f64, vmaxnmv_f16, vmaxnmvq_f16, vminnmv_f16,
>> vminnmvq_f16): Delete functions.
>>
>> gcc/testsuite/ChangeLog:
>>
>> * gcc.target/aarch64/neon/vaddv.c: New test.
>> * gcc.target/aarch64/neon/vmaxnmv.c: New test.
>> * gcc.target/aarch64/neon/vmaxv.c: New test.
>> * gcc.target/aarch64/neon/vminnmv.c: New test.
>> * gcc.target/aarch64/neon/vminv.c: New test.
>> ---
>> gcc/config/aarch64/aarch64-acle-builtins.h | 14 +
>> .../aarch64/aarch64-neon-builtins-base.cc | 7 +
>> .../aarch64/aarch64-neon-builtins-base.def | 22 ++
>> gcc/config/aarch64/arm_neon.h | 363 ------------------
>> gcc/testsuite/gcc.target/aarch64/neon/vaddv.c | 138 +++++++
>> .../gcc.target/aarch64/neon/vmaxnmv.c | 39 ++
>> gcc/testsuite/gcc.target/aarch64/neon/vmaxv.c | 135 +++++++
>> .../gcc.target/aarch64/neon/vminnmv.c | 39 ++
>> gcc/testsuite/gcc.target/aarch64/neon/vminv.c | 135 +++++++
>> 9 files changed, 529 insertions(+), 363 deletions(-)
>> create mode 100644 gcc/testsuite/gcc.target/aarch64/neon/vaddv.c
>> create mode 100644 gcc/testsuite/gcc.target/aarch64/neon/vmaxnmv.c
>> create mode 100644 gcc/testsuite/gcc.target/aarch64/neon/vmaxv.c
>> create mode 100644 gcc/testsuite/gcc.target/aarch64/neon/vminnmv.c
>> create mode 100644 gcc/testsuite/gcc.target/aarch64/neon/vminv.c
>>
>> diff --git a/gcc/config/aarch64/aarch64-acle-builtins.h b/gcc/config/aarch64/aarch64-acle-builtins.h
>> index a9832c1689b..77641a672f5 100644
>> --- a/gcc/config/aarch64/aarch64-acle-builtins.h
>> +++ b/gcc/config/aarch64/aarch64-acle-builtins.h
>> @@ -1393,6 +1393,18 @@ function_expander::result_mode () const
>> #define TYPES_s_float_sd_integer(S, D, T) \
>> TYPES_s_float (S, D, T), TYPES_sd_integer (S, D, T)
>>
>> +/* _f32
>> + _s8 _s16 _s32
>> + _u8 _u16 _u32. */
>> +#define TYPES_s_float_bhs_integer(S, D, T) \
>> + TYPES_s_float (S, D, T), TYPES_bhs_integer (S, D, T)
>> +
>> +/* _f32 _f64
>> + _s8 _s16 _s32 _s64
>> + _u8 _u16 _u32 _u64. */
>> +#define TYPES_sd_float_all_integer(S, D, T) \
>> + TYPES_sd_float (S, D, T), TYPES_all_integer (S, D, T)
>> +
>> /* _s32. */
>> #define TYPES_s_signed(S, D, T) \
>> S (s32)
>> @@ -2108,6 +2120,8 @@ DEF_SVE_TYPES_ARRAY (hsd_integer);
>> DEF_SVE_TYPES_ARRAY (hsd_data);
>> DEF_SVE_TYPES_ARRAY (s_float);
>> DEF_SVE_TYPES_ARRAY (s_float_hsd_integer);
>> +DEF_SVE_TYPES_ARRAY (s_float_bhs_integer);
>> +DEF_SVE_TYPES_ARRAY (sd_float_all_integer);
>> DEF_SVE_TYPES_ARRAY (s_float_mf8);
>> DEF_SVE_TYPES_ARRAY (s_float_sd_integer);
>> DEF_SVE_TYPES_ARRAY (s_signed);
>> diff --git a/gcc/config/aarch64/aarch64-neon-builtins-base.cc b/gcc/config/aarch64/aarch64-neon-builtins-base.cc
>> index f9b70af84eb..74dff4a8917 100644
>> --- a/gcc/config/aarch64/aarch64-neon-builtins-base.cc
>> +++ b/gcc/config/aarch64/aarch64-neon-builtins-base.cc
>> @@ -768,6 +768,13 @@ NEON_FUNCTION (vaddq, gimple_expr, (PLUS_EXPR, PLUS_EXPR, BIT_XOR_EXPR))
>> NEON_FUNCTION (vqaddb, vqaddh, vqadds, vqaddd, vqadd, vqaddq, gimple_ifn, (IFN_SAT_ADD))
>> NEON_FUNCTION (vqsubb, vqsubh, vqsubs, vqsubd, vqsub, vqsubq, gimple_ifn, (IFN_SAT_SUB))
>>
>> +// Reductions
>> +NEON_FUNCTION (vaddv, vaddvq, gimple_ifn, (IFN_REDUC_PLUS))
>> +NEON_FUNCTION (vmaxv, vmaxvq, gimple_ifn, (IFN_REDUC_MAX))
>> +NEON_FUNCTION (vminv, vminvq, gimple_ifn, (IFN_REDUC_MIN))
>> +NEON_FUNCTION (vmaxnmv, vmaxnmvq, gimple_ifn, (IFN_REDUC_FMAX))
>> +NEON_FUNCTION (vminnmv, vminnmvq, gimple_ifn, (IFN_REDUC_FMIN))
>> +
>> // Bitwise operations
>> NEON_FUNCTION (vand, vandq, gimple_expr, (BIT_AND_EXPR))
>> NEON_FUNCTION (vbic, vbicq, gimple_not_rhs, (BIT_AND_EXPR))
>> diff --git a/gcc/config/aarch64/aarch64-neon-builtins-base.def b/gcc/config/aarch64/aarch64-neon-builtins-base.def
>> index 0077ff3b839..3b7cbd430ce 100644
>> --- a/gcc/config/aarch64/aarch64-neon-builtins-base.def
>> +++ b/gcc/config/aarch64/aarch64-neon-builtins-base.def
>> @@ -92,6 +92,28 @@ DEF_NEON_FUNCTION (vqsub, all_integer, ("D0,D0,D0"))
>> DEF_NEON_FUNCTION (vqsubq, all_integer, ("Q0,Q0,Q0"))
>> #undef REQUIRED_EXTENSIONS
>>
>> +// Reductions
>> +#define REQUIRED_EXTENSIONS nonstreaming_only (AARCH64_FL_SIMD)
>> +DEF_NEON_FUNCTION (vaddv, s_float_bhs_integer, ("s0,D0"))
>> +DEF_NEON_FUNCTION (vaddvq, sd_float_all_integer, ("s0,Q0"))
>> +DEF_NEON_FUNCTION (vmaxv, bhs_integer, ("s0,D0"))
>> +DEF_NEON_FUNCTION (vmaxvq, bhs_integer, ("s0,Q0"))
>> +DEF_NEON_FUNCTION (vminv, bhs_integer, ("s0,D0"))
>> +DEF_NEON_FUNCTION (vminvq, bhs_integer, ("s0,Q0"))
>> +DEF_NEON_FUNCTION (vmaxnmv, s_float, ("s0,D0"))
>> +DEF_NEON_FUNCTION (vmaxnmvq, sd_float, ("s0,Q0"))
>> +DEF_NEON_FUNCTION (vminnmv, s_float, ("s0,D0"))
>> +DEF_NEON_FUNCTION (vminnmvq, sd_float, ("s0,Q0"))
>> +#undef REQUIRED_EXTENSIONS
>> +
>> +// Reductions (FP16)
>> +#define REQUIRED_EXTENSIONS nonstreaming_only (AARCH64_FL_SIMD | AARCH64_FL_F16)
>> +DEF_NEON_FUNCTION (vmaxnmv, h_float, ("s0,D0"))
>> +DEF_NEON_FUNCTION (vmaxnmvq, h_float, ("s0,Q0"))
>> +DEF_NEON_FUNCTION (vminnmv, h_float, ("s0,D0"))
>> +DEF_NEON_FUNCTION (vminnmvq, h_float, ("s0,Q0"))
>> +#undef REQUIRED_EXTENSIONS
>> +
>> // Bitwise operations
>> #define REQUIRED_EXTENSIONS nonstreaming_only (AARCH64_FL_SIMD)
>> DEF_NEON_FUNCTION (vand, all_integer, ("D0,D0,D0"))
>> diff --git a/gcc/config/aarch64/arm_neon.h b/gcc/config/aarch64/arm_neon.h
>> index fb67a735eaa..71a27b0c8b0 100644
>> --- a/gcc/config/aarch64/arm_neon.h
>> +++ b/gcc/config/aarch64/arm_neon.h
>> @@ -5055,127 +5055,6 @@ vabsd_s64 (int64_t __a)
>> return __a < 0 ? - (uint64_t) __a : __a;
>> }
>>
>> -/* vaddv */
>> -
>> -__extension__ extern __inline int8_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vaddv_s8 (int8x8_t __a)
>> -{
>> - return __builtin_aarch64_reduc_plus_scal_v8qi (__a);
>> -}
>> -
>> -__extension__ extern __inline int16_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vaddv_s16 (int16x4_t __a)
>> -{
>> - return __builtin_aarch64_reduc_plus_scal_v4hi (__a);
>> -}
>> -
>> -__extension__ extern __inline int32_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vaddv_s32 (int32x2_t __a)
>> -{
>> - return __builtin_aarch64_reduc_plus_scal_v2si (__a);
>> -}
>> -
>> -__extension__ extern __inline uint8_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vaddv_u8 (uint8x8_t __a)
>> -{
>> - return __builtin_aarch64_reduc_plus_scal_v8qi_uu (__a);
>> -}
>> -
>> -__extension__ extern __inline uint16_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vaddv_u16 (uint16x4_t __a)
>> -{
>> - return __builtin_aarch64_reduc_plus_scal_v4hi_uu (__a);
>> -}
>> -
>> -__extension__ extern __inline uint32_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vaddv_u32 (uint32x2_t __a)
>> -{
>> - return __builtin_aarch64_reduc_plus_scal_v2si_uu (__a);
>> -}
>> -
>> -__extension__ extern __inline int8_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vaddvq_s8 (int8x16_t __a)
>> -{
>> - return __builtin_aarch64_reduc_plus_scal_v16qi (__a);
>> -}
>> -
>> -__extension__ extern __inline int16_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vaddvq_s16 (int16x8_t __a)
>> -{
>> - return __builtin_aarch64_reduc_plus_scal_v8hi (__a);
>> -}
>> -
>> -__extension__ extern __inline int32_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vaddvq_s32 (int32x4_t __a)
>> -{
>> - return __builtin_aarch64_reduc_plus_scal_v4si (__a);
>> -}
>> -
>> -__extension__ extern __inline int64_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vaddvq_s64 (int64x2_t __a)
>> -{
>> - return __builtin_aarch64_reduc_plus_scal_v2di (__a);
>> -}
>> -
>> -__extension__ extern __inline uint8_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vaddvq_u8 (uint8x16_t __a)
>> -{
>> - return __builtin_aarch64_reduc_plus_scal_v16qi_uu (__a);
>> -}
>> -
>> -__extension__ extern __inline uint16_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vaddvq_u16 (uint16x8_t __a)
>> -{
>> - return __builtin_aarch64_reduc_plus_scal_v8hi_uu (__a);
>> -}
>> -
>> -__extension__ extern __inline uint32_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vaddvq_u32 (uint32x4_t __a)
>> -{
>> - return __builtin_aarch64_reduc_plus_scal_v4si_uu (__a);
>> -}
>> -
>> -__extension__ extern __inline uint64_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vaddvq_u64 (uint64x2_t __a)
>> -{
>> - return __builtin_aarch64_reduc_plus_scal_v2di_uu (__a);
>> -}
>> -
>> -__extension__ extern __inline float32_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vaddv_f32 (float32x2_t __a)
>> -{
>> - return __builtin_aarch64_reduc_plus_scal_v2sf (__a);
>> -}
>> -
>> -__extension__ extern __inline float32_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vaddvq_f32 (float32x4_t __a)
>> -{
>> - return __builtin_aarch64_reduc_plus_scal_v4sf (__a);
>> -}
>> -
>> -__extension__ extern __inline float64_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vaddvq_f64 (float64x2_t __a)
>> -{
>> - return __builtin_aarch64_reduc_plus_scal_v2df (__a);
>> -}
>> -
>> /* ARMv8.1-A intrinsics. */
>> #pragma GCC push_options
>> #pragma GCC target ("+nothing+rdma")
>> @@ -12365,48 +12244,6 @@ vmaxv_f32 (float32x2_t __a)
>> return __builtin_aarch64_reduc_smax_nan_scal_v2sf (__a);
>> }
>>
>> -__extension__ extern __inline int8_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vmaxv_s8 (int8x8_t __a)
>> -{
>> - return __builtin_aarch64_reduc_smax_scal_v8qi (__a);
>> -}
>> -
>> -__extension__ extern __inline int16_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vmaxv_s16 (int16x4_t __a)
>> -{
>> - return __builtin_aarch64_reduc_smax_scal_v4hi (__a);
>> -}
>> -
>> -__extension__ extern __inline int32_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vmaxv_s32 (int32x2_t __a)
>> -{
>> - return __builtin_aarch64_reduc_smax_scal_v2si (__a);
>> -}
>> -
>> -__extension__ extern __inline uint8_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vmaxv_u8 (uint8x8_t __a)
>> -{
>> - return __builtin_aarch64_reduc_umax_scal_v8qi_uu (__a);
>> -}
>> -
>> -__extension__ extern __inline uint16_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vmaxv_u16 (uint16x4_t __a)
>> -{
>> - return __builtin_aarch64_reduc_umax_scal_v4hi_uu (__a);
>> -}
>> -
>> -__extension__ extern __inline uint32_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vmaxv_u32 (uint32x2_t __a)
>> -{
>> - return __builtin_aarch64_reduc_umax_scal_v2si_uu (__a);
>> -}
>> -
>> __extension__ extern __inline float32_t
>> __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> vmaxvq_f32 (float32x4_t __a)
>> @@ -12421,71 +12258,6 @@ vmaxvq_f64 (float64x2_t __a)
>> return __builtin_aarch64_reduc_smax_nan_scal_v2df (__a);
>> }
>>
>> -__extension__ extern __inline int8_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vmaxvq_s8 (int8x16_t __a)
>> -{
>> - return __builtin_aarch64_reduc_smax_scal_v16qi (__a);
>> -}
>> -
>> -__extension__ extern __inline int16_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vmaxvq_s16 (int16x8_t __a)
>> -{
>> - return __builtin_aarch64_reduc_smax_scal_v8hi (__a);
>> -}
>> -
>> -__extension__ extern __inline int32_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vmaxvq_s32 (int32x4_t __a)
>> -{
>> - return __builtin_aarch64_reduc_smax_scal_v4si (__a);
>> -}
>> -
>> -__extension__ extern __inline uint8_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vmaxvq_u8 (uint8x16_t __a)
>> -{
>> - return __builtin_aarch64_reduc_umax_scal_v16qi_uu (__a);
>> -}
>> -
>> -__extension__ extern __inline uint16_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vmaxvq_u16 (uint16x8_t __a)
>> -{
>> - return __builtin_aarch64_reduc_umax_scal_v8hi_uu (__a);
>> -}
>> -
>> -__extension__ extern __inline uint32_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vmaxvq_u32 (uint32x4_t __a)
>> -{
>> - return __builtin_aarch64_reduc_umax_scal_v4si_uu (__a);
>> -}
>> -
>> -/* vmaxnmv */
>> -
>> -__extension__ extern __inline float32_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vmaxnmv_f32 (float32x2_t __a)
>> -{
>> - return __builtin_aarch64_reduc_smax_scal_v2sf (__a);
>> -}
>> -
>> -__extension__ extern __inline float32_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vmaxnmvq_f32 (float32x4_t __a)
>> -{
>> - return __builtin_aarch64_reduc_smax_scal_v4sf (__a);
>> -}
>> -
>> -__extension__ extern __inline float64_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vmaxnmvq_f64 (float64x2_t __a)
>> -{
>> - return __builtin_aarch64_reduc_smax_scal_v2df (__a);
>> -}
>> -
>> /* vmin */
>>
>> __extension__ extern __inline float32x2_t
>> @@ -12649,48 +12421,6 @@ vminv_f32 (float32x2_t __a)
>> return __builtin_aarch64_reduc_smin_nan_scal_v2sf (__a);
>> }
>>
>> -__extension__ extern __inline int8_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vminv_s8 (int8x8_t __a)
>> -{
>> - return __builtin_aarch64_reduc_smin_scal_v8qi (__a);
>> -}
>> -
>> -__extension__ extern __inline int16_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vminv_s16 (int16x4_t __a)
>> -{
>> - return __builtin_aarch64_reduc_smin_scal_v4hi (__a);
>> -}
>> -
>> -__extension__ extern __inline int32_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vminv_s32 (int32x2_t __a)
>> -{
>> - return __builtin_aarch64_reduc_smin_scal_v2si (__a);
>> -}
>> -
>> -__extension__ extern __inline uint8_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vminv_u8 (uint8x8_t __a)
>> -{
>> - return __builtin_aarch64_reduc_umin_scal_v8qi_uu (__a);
>> -}
>> -
>> -__extension__ extern __inline uint16_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vminv_u16 (uint16x4_t __a)
>> -{
>> - return __builtin_aarch64_reduc_umin_scal_v4hi_uu (__a);
>> -}
>> -
>> -__extension__ extern __inline uint32_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vminv_u32 (uint32x2_t __a)
>> -{
>> - return __builtin_aarch64_reduc_umin_scal_v2si_uu (__a);
>> -}
>> -
>> __extension__ extern __inline float32_t
>> __attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> vminvq_f32 (float32x4_t __a)
>> @@ -12705,71 +12435,6 @@ vminvq_f64 (float64x2_t __a)
>> return __builtin_aarch64_reduc_smin_nan_scal_v2df (__a);
>> }
>>
>> -__extension__ extern __inline int8_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vminvq_s8 (int8x16_t __a)
>> -{
>> - return __builtin_aarch64_reduc_smin_scal_v16qi (__a);
>> -}
>> -
>> -__extension__ extern __inline int16_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vminvq_s16 (int16x8_t __a)
>> -{
>> - return __builtin_aarch64_reduc_smin_scal_v8hi (__a);
>> -}
>> -
>> -__extension__ extern __inline int32_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vminvq_s32 (int32x4_t __a)
>> -{
>> - return __builtin_aarch64_reduc_smin_scal_v4si (__a);
>> -}
>> -
>> -__extension__ extern __inline uint8_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vminvq_u8 (uint8x16_t __a)
>> -{
>> - return __builtin_aarch64_reduc_umin_scal_v16qi_uu (__a);
>> -}
>> -
>> -__extension__ extern __inline uint16_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vminvq_u16 (uint16x8_t __a)
>> -{
>> - return __builtin_aarch64_reduc_umin_scal_v8hi_uu (__a);
>> -}
>> -
>> -__extension__ extern __inline uint32_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vminvq_u32 (uint32x4_t __a)
>> -{
>> - return __builtin_aarch64_reduc_umin_scal_v4si_uu (__a);
>> -}
>> -
>> -/* vminnmv */
>> -
>> -__extension__ extern __inline float32_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vminnmv_f32 (float32x2_t __a)
>> -{
>> - return __builtin_aarch64_reduc_smin_scal_v2sf (__a);
>> -}
>> -
>> -__extension__ extern __inline float32_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vminnmvq_f32 (float32x4_t __a)
>> -{
>> - return __builtin_aarch64_reduc_smin_scal_v4sf (__a);
>> -}
>> -
>> -__extension__ extern __inline float64_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vminnmvq_f64 (float64x2_t __a)
>> -{
>> - return __builtin_aarch64_reduc_smin_scal_v2df (__a);
>> -}
>> -
>> /* vmla */
>>
>> __extension__ extern __inline float32x2_t
>> @@ -20513,34 +20178,6 @@ vminvq_f16 (float16x8_t __a)
>> return __builtin_aarch64_reduc_smin_nan_scal_v8hf (__a);
>> }
>>
>> -__extension__ extern __inline float16_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vmaxnmv_f16 (float16x4_t __a)
>> -{
>> - return __builtin_aarch64_reduc_smax_scal_v4hf (__a);
>> -}
>> -
>> -__extension__ extern __inline float16_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vmaxnmvq_f16 (float16x8_t __a)
>> -{
>> - return __builtin_aarch64_reduc_smax_scal_v8hf (__a);
>> -}
>> -
>> -__extension__ extern __inline float16_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vminnmv_f16 (float16x4_t __a)
>> -{
>> - return __builtin_aarch64_reduc_smin_scal_v4hf (__a);
>> -}
>> -
>> -__extension__ extern __inline float16_t
>> -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__))
>> -vminnmvq_f16 (float16x8_t __a)
>> -{
>> - return __builtin_aarch64_reduc_smin_scal_v8hf (__a);
>> -}
>> -
>> #pragma GCC pop_options
>>
>> /* AdvSIMD Dot Product intrinsics. */
>> diff --git a/gcc/testsuite/gcc.target/aarch64/neon/vaddv.c b/gcc/testsuite/gcc.target/aarch64/neon/vaddv.c
>> new file mode 100644
>> index 00000000000..43b91b102cf
>> --- /dev/null
>> +++ b/gcc/testsuite/gcc.target/aarch64/neon/vaddv.c
>> @@ -0,0 +1,138 @@
>> +/* { dg-do compile } */
>> +/* { dg-final { check-function-bodies "**" "" } } */
>> +
>> +#include "arm_neon_test.h"
>> +
>> +/*
>> +** test_vaddv_u8:
>> +** addv b([0-9]+), v0\.8b
>> +** umov w0, v\1\.b\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vaddv_u8, uint8_t, uint8x8_t)
>> +
>> +/*
>> +** test_vaddv_s8:
>> +** addv b([0-9]+), v0\.8b
>> +** umov w0, v\1\.b\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vaddv_s8, int8_t, int8x8_t)
>> +
>> +/*
>> +** test_vaddv_u16:
>> +** addv h([0-9]+), v0\.4h
>> +** umov w0, v\1\.h\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vaddv_u16, uint16_t, uint16x4_t)
>> +
>> +/*
>> +** test_vaddv_s16:
>> +** addv h([0-9]+), v0\.4h
>> +** umov w0, v\1\.h\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vaddv_s16, int16_t, int16x4_t)
>> +
>> +/*
>> +** test_vaddv_u32:
>> +** addp v([0-9]+)\.2s, v0\.2s, v0\.2s
>> +** fmov w0, s\1
>> +** ret
>> +*/
>> +TEST_UNARY (vaddv_u32, uint32_t, uint32x2_t)
>> +
>> +/*
>> +** test_vaddv_s32:
>> +** addp v([0-9]+)\.2s, v0\.2s, v0\.2s
>> +** fmov w0, s\1
>> +** ret
>> +*/
>> +TEST_UNARY (vaddv_s32, int32_t, int32x2_t)
>> +
>> +/*
>> +** test_vaddvq_u8:
>> +** addv b([0-9]+), v0\.16b
>> +** umov w0, v\1\.b\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vaddvq_u8, uint8_t, uint8x16_t)
>> +
>> +/*
>> +** test_vaddvq_s8:
>> +** addv b([0-9]+), v0\.16b
>> +** umov w0, v\1\.b\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vaddvq_s8, int8_t, int8x16_t)
>> +
>> +/*
>> +** test_vaddvq_u16:
>> +** addv h([0-9]+), v0\.8h
>> +** umov w0, v\1\.h\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vaddvq_u16, uint16_t, uint16x8_t)
>> +
>> +/*
>> +** test_vaddvq_s16:
>> +** addv h([0-9]+), v0\.8h
>> +** umov w0, v\1\.h\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vaddvq_s16, int16_t, int16x8_t)
>> +
>> +/*
>> +** test_vaddvq_u32:
>> +** addv (s[0-9]+), v0\.4s
>> +** fmov w0, \1
>> +** ret
>> +*/
>> +TEST_UNARY (vaddvq_u32, uint32_t, uint32x4_t)
>> +
>> +/*
>> +** test_vaddvq_s32:
>> +** addv (s[0-9]+), v0\.4s
>> +** fmov w0, \1
>> +** ret
>> +*/
>> +TEST_UNARY (vaddvq_s32, int32_t, int32x4_t)
>> +
>> +/*
>> +** test_vaddvq_u64:
>> +** addp (d[0-9]+), v0\.2d
>> +** fmov x0, \1
>> +** ret
>> +*/
>> +TEST_UNARY (vaddvq_u64, uint64_t, uint64x2_t)
>> +
>> +/*
>> +** test_vaddvq_s64:
>> +** addp (d[0-9]+), v0\.2d
>> +** fmov x0, \1
>> +** ret
>> +*/
>> +TEST_UNARY (vaddvq_s64, int64_t, int64x2_t)
>> +
>> +/*
>> +** test_vaddv_f32:
>> +** faddp s0, v0\.2s
>> +** ret
>> +*/
>> +TEST_UNARY (vaddv_f32, float32_t, float32x2_t)
>> +
>> +/*
>> +** test_vaddvq_f32:
>> +** faddp (v[0-9]+\.4s), v0\.4s, v0\.4s
>> +** faddp v0\.4s, \1, \1
>> +** ret
>> +*/
>> +TEST_UNARY (vaddvq_f32, float32_t, float32x4_t)
>> +
>> +/*
>> +** test_vaddvq_f64:
>> +** faddp d0, v0\.2d
>> +** ret
>> +*/
>> +TEST_UNARY (vaddvq_f64, float64_t, float64x2_t)
>> diff --git a/gcc/testsuite/gcc.target/aarch64/neon/vmaxnmv.c b/gcc/testsuite/gcc.target/aarch64/neon/vmaxnmv.c
>> new file mode 100644
>> index 00000000000..47f54144fdb
>> --- /dev/null
>> +++ b/gcc/testsuite/gcc.target/aarch64/neon/vmaxnmv.c
>> @@ -0,0 +1,39 @@
>> +/* { dg-do compile } */
>> +/* { dg-final { check-function-bodies "**" "" } } */
>> +
>> +#include "arm_neon_test.h"
>> +
>> +/*
>> +** test_vmaxnmv_f16:
>> +** fmaxnmv h0, v0\.4h
>> +** ret
>> +*/
>> +TEST_UNARY (vmaxnmv_f16, float16_t, float16x4_t)
>> +
>> +/*
>> +** test_vmaxnmv_f32:
>> +** fmaxnmp s0, v0\.2s
>> +** ret
>> +*/
>> +TEST_UNARY (vmaxnmv_f32, float32_t, float32x2_t)
>> +
>> +/*
>> +** test_vmaxnmvq_f16:
>> +** fmaxnmv h0, v0\.8h
>> +** ret
>> +*/
>> +TEST_UNARY (vmaxnmvq_f16, float16_t, float16x8_t)
>> +
>> +/*
>> +** test_vmaxnmvq_f32:
>> +** fmaxnmv s0, v0\.4s
>> +** ret
>> +*/
>> +TEST_UNARY (vmaxnmvq_f32, float32_t, float32x4_t)
>> +
>> +/*
>> +** test_vmaxnmvq_f64:
>> +** fmaxnmp d0, v0\.2d
>> +** ret
>> +*/
>> +TEST_UNARY (vmaxnmvq_f64, float64_t, float64x2_t)
>> diff --git a/gcc/testsuite/gcc.target/aarch64/neon/vmaxv.c b/gcc/testsuite/gcc.target/aarch64/neon/vmaxv.c
>> new file mode 100644
>> index 00000000000..a28da5fbb70
>> --- /dev/null
>> +++ b/gcc/testsuite/gcc.target/aarch64/neon/vmaxv.c
>> @@ -0,0 +1,135 @@
>> +/* { dg-do compile } */
>> +/* { dg-final { check-function-bodies "**" "" } } */
>> +
>> +#include "arm_neon_test.h"
>> +
>> +/*
>> +** test_vmaxv_u8:
>> +** umaxv b([0-9]+), v0\.8b
>> +** umov x0, v\1\.d\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vmaxv_u8, uint8_t, uint8x8_t)
>> +
>> +/*
>> +** test_vmaxv_s8:
>> +** smaxv b([0-9]+), v0\.8b
>> +** umov x0, v\1\.d\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vmaxv_s8, int8_t, int8x8_t)
>> +
>> +/*
>> +** test_vmaxv_u16:
>> +** umaxv h([0-9]+), v0\.4h
>> +** umov x0, v\1\.d\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vmaxv_u16, uint16_t, uint16x4_t)
>> +
>> +/*
>> +** test_vmaxv_s16:
>> +** smaxv h([0-9]+), v0\.4h
>> +** umov x0, v\1\.d\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vmaxv_s16, int16_t, int16x4_t)
>> +
>> +/*
>> +** test_vmaxv_u32:
>> +** umaxp v([0-9]+)\.2s, v0\.2s, v0\.2s
>> +** umov x0, v\1\.d\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vmaxv_u32, uint32_t, uint32x2_t)
>> +
>> +/*
>> +** test_vmaxv_s32:
>> +** smaxp v([0-9]+)\.2s, v0\.2s, v0\.2s
>> +** umov x0, v\1\.d\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vmaxv_s32, int32_t, int32x2_t)
>> +
>> +/*
>> +** test_vmaxvq_u8:
>> +** umaxv b([0-9]+), v0\.16b
>> +** umov w0, v\1\.b\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vmaxvq_u8, uint8_t, uint8x16_t)
>> +
>> +/*
>> +** test_vmaxvq_s8:
>> +** smaxv b([0-9]+), v0\.16b
>> +** umov w0, v\1\.b\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vmaxvq_s8, int8_t, int8x16_t)
>> +
>> +/*
>> +** test_vmaxvq_u16:
>> +** umaxv h([0-9]+), v0\.8h
>> +** umov w0, v\1\.h\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vmaxvq_u16, uint16_t, uint16x8_t)
>> +
>> +/*
>> +** test_vmaxvq_s16:
>> +** smaxv h([0-9]+), v0\.8h
>> +** umov w0, v\1\.h\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vmaxvq_s16, int16_t, int16x8_t)
>> +
>> +/*
>> +** test_vmaxvq_u32:
>> +** umaxv (s[0-9]+), v0\.4s
>> +** fmov w0, \1
>> +** ret
>> +*/
>> +TEST_UNARY (vmaxvq_u32, uint32_t, uint32x4_t)
>> +
>> +/*
>> +** test_vmaxvq_s32:
>> +** smaxv (s[0-9]+), v0\.4s
>> +** fmov w0, \1
>> +** ret
>> +*/
>> +TEST_UNARY (vmaxvq_s32, int32_t, int32x4_t)
>> +
>> +/*
>> +** test_vmaxv_f16:
>> +** fmaxv h0, v0\.4h
>> +** ret
>> +*/
>> +TEST_UNARY (vmaxv_f16, float16_t, float16x4_t)
>> +
>> +/*
>> +** test_vmaxv_f32:
>> +** fmaxp s0, v0\.2s
>> +** ret
>> +*/
>> +TEST_UNARY (vmaxv_f32, float32_t, float32x2_t)
>> +
>> +/*
>> +** test_vmaxvq_f16:
>> +** fmaxv h0, v0\.8h
>> +** ret
>> +*/
>> +TEST_UNARY (vmaxvq_f16, float16_t, float16x8_t)
>> +
>> +/*
>> +** test_vmaxvq_f32:
>> +** fmaxv s0, v0\.4s
>> +** ret
>> +*/
>> +TEST_UNARY (vmaxvq_f32, float32_t, float32x4_t)
>> +
>> +/*
>> +** test_vmaxvq_f64:
>> +** fmaxp d0, v0\.2d
>> +** ret
>> +*/
>> +TEST_UNARY (vmaxvq_f64, float64_t, float64x2_t)
>> diff --git a/gcc/testsuite/gcc.target/aarch64/neon/vminnmv.c b/gcc/testsuite/gcc.target/aarch64/neon/vminnmv.c
>> new file mode 100644
>> index 00000000000..04765f7d3bc
>> --- /dev/null
>> +++ b/gcc/testsuite/gcc.target/aarch64/neon/vminnmv.c
>> @@ -0,0 +1,39 @@
>> +/* { dg-do compile } */
>> +/* { dg-final { check-function-bodies "**" "" } } */
>> +
>> +#include "arm_neon_test.h"
>> +
>> +/*
>> +** test_vminnmv_f16:
>> +** fminnmv h0, v0\.4h
>> +** ret
>> +*/
>> +TEST_UNARY (vminnmv_f16, float16_t, float16x4_t)
>> +
>> +/*
>> +** test_vminnmv_f32:
>> +** fminnmp s0, v0\.2s
>> +** ret
>> +*/
>> +TEST_UNARY (vminnmv_f32, float32_t, float32x2_t)
>> +
>> +/*
>> +** test_vminnmvq_f16:
>> +** fminnmv h0, v0\.8h
>> +** ret
>> +*/
>> +TEST_UNARY (vminnmvq_f16, float16_t, float16x8_t)
>> +
>> +/*
>> +** test_vminnmvq_f32:
>> +** fminnmv s0, v0\.4s
>> +** ret
>> +*/
>> +TEST_UNARY (vminnmvq_f32, float32_t, float32x4_t)
>> +
>> +/*
>> +** test_vminnmvq_f64:
>> +** fminnmp d0, v0\.2d
>> +** ret
>> +*/
>> +TEST_UNARY (vminnmvq_f64, float64_t, float64x2_t)
>> diff --git a/gcc/testsuite/gcc.target/aarch64/neon/vminv.c b/gcc/testsuite/gcc.target/aarch64/neon/vminv.c
>> new file mode 100644
>> index 00000000000..9b7657357fd
>> --- /dev/null
>> +++ b/gcc/testsuite/gcc.target/aarch64/neon/vminv.c
>> @@ -0,0 +1,135 @@
>> +/* { dg-do compile } */
>> +/* { dg-final { check-function-bodies "**" "" } } */
>> +
>> +#include "arm_neon_test.h"
>> +
>> +/*
>> +** test_vminv_u8:
>> +** uminv b([0-9]+), v0\.8b
>> +** umov x0, v\1\.d\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vminv_u8, uint8_t, uint8x8_t)
>> +
>> +/*
>> +** test_vminv_s8:
>> +** sminv b([0-9]+), v0\.8b
>> +** umov x0, v\1\.d\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vminv_s8, int8_t, int8x8_t)
>> +
>> +/*
>> +** test_vminv_u16:
>> +** uminv h([0-9]+), v0\.4h
>> +** umov x0, v\1\.d\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vminv_u16, uint16_t, uint16x4_t)
>> +
>> +/*
>> +** test_vminv_s16:
>> +** sminv h([0-9]+), v0\.4h
>> +** umov x0, v\1\.d\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vminv_s16, int16_t, int16x4_t)
>> +
>> +/*
>> +** test_vminv_u32:
>> +** uminp v([0-9]+)\.2s, v0\.2s, v0\.2s
>> +** umov x0, v\1\.d\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vminv_u32, uint32_t, uint32x2_t)
>> +
>> +/*
>> +** test_vminv_s32:
>> +** sminp v([0-9]+)\.2s, v0\.2s, v0\.2s
>> +** umov x0, v\1\.d\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vminv_s32, int32_t, int32x2_t)
>> +
>> +/*
>> +** test_vminvq_u8:
>> +** uminv b([0-9]+), v0\.16b
>> +** umov w0, v\1\.b\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vminvq_u8, uint8_t, uint8x16_t)
>> +
>> +/*
>> +** test_vminvq_s8:
>> +** sminv b([0-9]+), v0\.16b
>> +** umov w0, v\1\.b\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vminvq_s8, int8_t, int8x16_t)
>> +
>> +/*
>> +** test_vminvq_u16:
>> +** uminv h([0-9]+), v0\.8h
>> +** umov w0, v\1\.h\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vminvq_u16, uint16_t, uint16x8_t)
>> +
>> +/*
>> +** test_vminvq_s16:
>> +** sminv h([0-9]+), v0\.8h
>> +** umov w0, v\1\.h\[0\]
>> +** ret
>> +*/
>> +TEST_UNARY (vminvq_s16, int16_t, int16x8_t)
>> +
>> +/*
>> +** test_vminvq_u32:
>> +** uminv (s[0-9]+), v0\.4s
>> +** fmov w0, \1
>> +** ret
>> +*/
>> +TEST_UNARY (vminvq_u32, uint32_t, uint32x4_t)
>> +
>> +/*
>> +** test_vminvq_s32:
>> +** sminv (s[0-9]+), v0\.4s
>> +** fmov w0, \1
>> +** ret
>> +*/
>> +TEST_UNARY (vminvq_s32, int32_t, int32x4_t)
>> +
>> +/*
>> +** test_vminv_f16:
>> +** fminv h0, v0\.4h
>> +** ret
>> +*/
>> +TEST_UNARY (vminv_f16, float16_t, float16x4_t)
>> +
>> +/*
>> +** test_vminv_f32:
>> +** fminp s0, v0\.2s
>> +** ret
>> +*/
>> +TEST_UNARY (vminv_f32, float32_t, float32x2_t)
>> +
>> +/*
>> +** test_vminvq_f16:
>> +** fminv h0, v0\.8h
>> +** ret
>> +*/
>> +TEST_UNARY (vminvq_f16, float16_t, float16x8_t)
>> +
>> +/*
>> +** test_vminvq_f32:
>> +** fminv s0, v0\.4s
>> +** ret
>> +*/
>> +TEST_UNARY (vminvq_f32, float32_t, float32x4_t)
>> +
>> +/*
>> +** test_vminvq_f64:
>> +** fminp d0, v0\.2d
>> +** ret
>> +*/
>> +TEST_UNARY (vminvq_f64, float64_t, float64x2_t)
>> --
>> 2.43.0
>>
>
More information about the Gcc-patches
mailing list