From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: by sourceware.org (Postfix, from userid 7835) id 99B79396ECAC; Tue, 17 Aug 2021 10:45:08 +0000 (GMT) DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 99B79396ECAC MIME-Version: 1.0 Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset="utf-8" From: Jonathan Wright To: gcc-cvs@gcc.gnu.org Subject: [gcc r12-2951] aarch64: Remove macros for vld4[q]_lane Neon intrinsics X-Act-Checkin: gcc X-Git-Author: Jonathan Wright X-Git-Refname: refs/heads/master X-Git-Oldrev: 08f83812e5c5fdd9a7a4a1b9e46bb33725185c5a X-Git-Newrev: d1819df86fbe42125cccb2fc2959a0bf51e524d6 Message-Id: <20210817104508.99B79396ECAC@sourceware.org> Date: Tue, 17 Aug 2021 10:45:08 +0000 (GMT) X-BeenThere: gcc-cvs@gcc.gnu.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Gcc-cvs mailing list List-Unsubscribe: , List-Archive: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 17 Aug 2021 10:45:08 -0000 https://gcc.gnu.org/g:d1819df86fbe42125cccb2fc2959a0bf51e524d6 commit r12-2951-gd1819df86fbe42125cccb2fc2959a0bf51e524d6 Author: Jonathan Wright Date: Mon Aug 16 14:37:18 2021 +0100 aarch64: Remove macros for vld4[q]_lane Neon intrinsics Remove macros for vld4[q]_lane Neon intrinsics. This is a preparatory step before adding new modes for structures of Advanced SIMD vectors. gcc/ChangeLog: 2021-08-16 Jonathan Wright * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete. (__LD4Q_LANE_FUNC): Likewise. (vld4_lane_u8): Define without macro. (vld4_lane_u16): Likewise. (vld4_lane_u32): Likewise. (vld4_lane_u64): Likewise. (vld4_lane_s8): Likewise. (vld4_lane_s16): Likewise. (vld4_lane_s32): Likewise. (vld4_lane_s64): Likewise. (vld4_lane_f16): Likewise. (vld4_lane_f32): Likewise. (vld4_lane_f64): Likewise. (vld4_lane_p8): Likewise. (vld4_lane_p16): Likewise. (vld4_lane_p64): Likewise. (vld4q_lane_u8): Likewise. (vld4q_lane_u16): Likewise. (vld4q_lane_u32): Likewise. (vld4q_lane_u64): Likewise. (vld4q_lane_s8): Likewise. (vld4q_lane_s16): Likewise. (vld4q_lane_s32): Likewise. (vld4q_lane_s64): Likewise. (vld4q_lane_f16): Likewise. (vld4q_lane_f32): Likewise. (vld4q_lane_f64): Likewise. (vld4q_lane_p8): Likewise. (vld4q_lane_p16): Likewise. (vld4q_lane_p64): Likewise. (vld4_lane_bf16): Likewise. (vld4q_lane_bf16): Likewise. Diff: --- gcc/config/aarch64/arm_neon.h | 728 ++++++++++++++++++++++++++++++++++++------ 1 file changed, 624 insertions(+), 104 deletions(-) diff --git a/gcc/config/aarch64/arm_neon.h b/gcc/config/aarch64/arm_neon.h index 29b62988a91..d8b29706a20 100644 --- a/gcc/config/aarch64/arm_neon.h +++ b/gcc/config/aarch64/arm_neon.h @@ -20856,110 +20856,595 @@ vld3q_lane_p64 (const poly64_t * __ptr, poly64x2x3_t __b, const int __c) /* vld4_lane */ -#define __LD4_LANE_FUNC(intype, vectype, largetype, ptrtype, mode, \ - qmode, ptrmode, funcsuffix, signedtype) \ -__extension__ extern __inline intype \ -__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) \ -vld4_lane_##funcsuffix (const ptrtype * __ptr, intype __b, const int __c) \ -{ \ - __builtin_aarch64_simd_xi __o; \ - largetype __temp; \ - __temp.val[0] = \ - vcombine_##funcsuffix (__b.val[0], vcreate_##funcsuffix (0)); \ - __temp.val[1] = \ - vcombine_##funcsuffix (__b.val[1], vcreate_##funcsuffix (0)); \ - __temp.val[2] = \ - vcombine_##funcsuffix (__b.val[2], vcreate_##funcsuffix (0)); \ - __temp.val[3] = \ - vcombine_##funcsuffix (__b.val[3], vcreate_##funcsuffix (0)); \ - __o = __builtin_aarch64_set_qregxi##qmode (__o, \ - (signedtype) __temp.val[0], \ - 0); \ - __o = __builtin_aarch64_set_qregxi##qmode (__o, \ - (signedtype) __temp.val[1], \ - 1); \ - __o = __builtin_aarch64_set_qregxi##qmode (__o, \ - (signedtype) __temp.val[2], \ - 2); \ - __o = __builtin_aarch64_set_qregxi##qmode (__o, \ - (signedtype) __temp.val[3], \ - 3); \ - __o = __builtin_aarch64_ld4_lane##mode ( \ - (__builtin_aarch64_simd_##ptrmode *) __ptr, __o, __c); \ - __b.val[0] = (vectype) __builtin_aarch64_get_dregxidi (__o, 0); \ - __b.val[1] = (vectype) __builtin_aarch64_get_dregxidi (__o, 1); \ - __b.val[2] = (vectype) __builtin_aarch64_get_dregxidi (__o, 2); \ - __b.val[3] = (vectype) __builtin_aarch64_get_dregxidi (__o, 3); \ - return __b; \ +__extension__ extern __inline uint8x8x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4_lane_u8 (const uint8_t * __ptr, uint8x8x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + uint8x16x4_t __temp; + __temp.val[0] = vcombine_u8 (__b.val[0], vcreate_u8 (0)); + __temp.val[1] = vcombine_u8 (__b.val[1], vcreate_u8 (0)); + __temp.val[2] = vcombine_u8 (__b.val[2], vcreate_u8 (0)); + __temp.val[3] = vcombine_u8 (__b.val[3], vcreate_u8 (0)); + __o = __builtin_aarch64_set_qregxiv16qi (__o, (int8x16_t) __temp.val[0], 0); + __o = __builtin_aarch64_set_qregxiv16qi (__o, (int8x16_t) __temp.val[1], 1); + __o = __builtin_aarch64_set_qregxiv16qi (__o, (int8x16_t) __temp.val[2], 2); + __o = __builtin_aarch64_set_qregxiv16qi (__o, (int8x16_t) __temp.val[3], 3); + __o = __builtin_aarch64_ld4_lanev8qi ( + (__builtin_aarch64_simd_qi *) __ptr, __o, __c); + __b.val[0] = (uint8x8_t) __builtin_aarch64_get_dregxidi (__o, 0); + __b.val[1] = (uint8x8_t) __builtin_aarch64_get_dregxidi (__o, 1); + __b.val[2] = (uint8x8_t) __builtin_aarch64_get_dregxidi (__o, 2); + __b.val[3] = (uint8x8_t) __builtin_aarch64_get_dregxidi (__o, 3); + return __b; } -/* vld4q_lane */ +__extension__ extern __inline uint16x4x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4_lane_u16 (const uint16_t * __ptr, uint16x4x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + uint16x8x4_t __temp; + __temp.val[0] = vcombine_u16 (__b.val[0], vcreate_u16 (0)); + __temp.val[1] = vcombine_u16 (__b.val[1], vcreate_u16 (0)); + __temp.val[2] = vcombine_u16 (__b.val[2], vcreate_u16 (0)); + __temp.val[3] = vcombine_u16 (__b.val[3], vcreate_u16 (0)); + __o = __builtin_aarch64_set_qregxiv8hi (__o, (int16x8_t) __temp.val[0], 0); + __o = __builtin_aarch64_set_qregxiv8hi (__o, (int16x8_t) __temp.val[1], 1); + __o = __builtin_aarch64_set_qregxiv8hi (__o, (int16x8_t) __temp.val[2], 2); + __o = __builtin_aarch64_set_qregxiv8hi (__o, (int16x8_t) __temp.val[3], 3); + __o = __builtin_aarch64_ld4_lanev4hi ( + (__builtin_aarch64_simd_hi *) __ptr, __o, __c); + __b.val[0] = (uint16x4_t) __builtin_aarch64_get_dregxidi (__o, 0); + __b.val[1] = (uint16x4_t) __builtin_aarch64_get_dregxidi (__o, 1); + __b.val[2] = (uint16x4_t) __builtin_aarch64_get_dregxidi (__o, 2); + __b.val[3] = (uint16x4_t) __builtin_aarch64_get_dregxidi (__o, 3); + return __b; +} + +__extension__ extern __inline uint32x2x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4_lane_u32 (const uint32_t * __ptr, uint32x2x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + uint32x4x4_t __temp; + __temp.val[0] = vcombine_u32 (__b.val[0], vcreate_u32 (0)); + __temp.val[1] = vcombine_u32 (__b.val[1], vcreate_u32 (0)); + __temp.val[2] = vcombine_u32 (__b.val[2], vcreate_u32 (0)); + __temp.val[3] = vcombine_u32 (__b.val[3], vcreate_u32 (0)); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __temp.val[0], 0); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __temp.val[1], 1); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __temp.val[2], 2); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __temp.val[3], 3); + __o = __builtin_aarch64_ld4_lanev2si ( + (__builtin_aarch64_simd_si *) __ptr, __o, __c); + __b.val[0] = (uint32x2_t) __builtin_aarch64_get_dregxidi (__o, 0); + __b.val[1] = (uint32x2_t) __builtin_aarch64_get_dregxidi (__o, 1); + __b.val[2] = (uint32x2_t) __builtin_aarch64_get_dregxidi (__o, 2); + __b.val[3] = (uint32x2_t) __builtin_aarch64_get_dregxidi (__o, 3); + return __b; +} + +__extension__ extern __inline uint64x1x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4_lane_u64 (const uint64_t * __ptr, uint64x1x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + uint64x2x4_t __temp; + __temp.val[0] = vcombine_u64 (__b.val[0], vcreate_u64 (0)); + __temp.val[1] = vcombine_u64 (__b.val[1], vcreate_u64 (0)); + __temp.val[2] = vcombine_u64 (__b.val[2], vcreate_u64 (0)); + __temp.val[3] = vcombine_u64 (__b.val[3], vcreate_u64 (0)); + __o = __builtin_aarch64_set_qregxiv2di (__o, (int64x2_t) __temp.val[0], 0); + __o = __builtin_aarch64_set_qregxiv2di (__o, (int64x2_t) __temp.val[1], 1); + __o = __builtin_aarch64_set_qregxiv2di (__o, (int64x2_t) __temp.val[2], 2); + __o = __builtin_aarch64_set_qregxiv2di (__o, (int64x2_t) __temp.val[3], 3); + __o = __builtin_aarch64_ld4_lanedi ( + (__builtin_aarch64_simd_di *) __ptr, __o, __c); + __b.val[0] = (uint64x1_t) __builtin_aarch64_get_dregxidi (__o, 0); + __b.val[1] = (uint64x1_t) __builtin_aarch64_get_dregxidi (__o, 1); + __b.val[2] = (uint64x1_t) __builtin_aarch64_get_dregxidi (__o, 2); + __b.val[3] = (uint64x1_t) __builtin_aarch64_get_dregxidi (__o, 3); + return __b; +} + +__extension__ extern __inline int8x8x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4_lane_s8 (const int8_t * __ptr, int8x8x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + int8x16x4_t __temp; + __temp.val[0] = vcombine_s8 (__b.val[0], vcreate_s8 (0)); + __temp.val[1] = vcombine_s8 (__b.val[1], vcreate_s8 (0)); + __temp.val[2] = vcombine_s8 (__b.val[2], vcreate_s8 (0)); + __temp.val[3] = vcombine_s8 (__b.val[3], vcreate_s8 (0)); + __o = __builtin_aarch64_set_qregxiv16qi (__o, (int8x16_t) __temp.val[0], 0); + __o = __builtin_aarch64_set_qregxiv16qi (__o, (int8x16_t) __temp.val[1], 1); + __o = __builtin_aarch64_set_qregxiv16qi (__o, (int8x16_t) __temp.val[2], 2); + __o = __builtin_aarch64_set_qregxiv16qi (__o, (int8x16_t) __temp.val[3], 3); + __o = __builtin_aarch64_ld4_lanev8qi ( + (__builtin_aarch64_simd_qi *) __ptr, __o, __c); + __b.val[0] = (int8x8_t) __builtin_aarch64_get_dregxidi (__o, 0); + __b.val[1] = (int8x8_t) __builtin_aarch64_get_dregxidi (__o, 1); + __b.val[2] = (int8x8_t) __builtin_aarch64_get_dregxidi (__o, 2); + __b.val[3] = (int8x8_t) __builtin_aarch64_get_dregxidi (__o, 3); + return __b; +} + +__extension__ extern __inline int16x4x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4_lane_s16 (const int16_t * __ptr, int16x4x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + int16x8x4_t __temp; + __temp.val[0] = vcombine_s16 (__b.val[0], vcreate_s16 (0)); + __temp.val[1] = vcombine_s16 (__b.val[1], vcreate_s16 (0)); + __temp.val[2] = vcombine_s16 (__b.val[2], vcreate_s16 (0)); + __temp.val[3] = vcombine_s16 (__b.val[3], vcreate_s16 (0)); + __o = __builtin_aarch64_set_qregxiv8hi (__o, (int16x8_t) __temp.val[0], 0); + __o = __builtin_aarch64_set_qregxiv8hi (__o, (int16x8_t) __temp.val[1], 1); + __o = __builtin_aarch64_set_qregxiv8hi (__o, (int16x8_t) __temp.val[2], 2); + __o = __builtin_aarch64_set_qregxiv8hi (__o, (int16x8_t) __temp.val[3], 3); + __o = __builtin_aarch64_ld4_lanev4hi ( + (__builtin_aarch64_simd_hi *) __ptr, __o, __c); + __b.val[0] = (int16x4_t) __builtin_aarch64_get_dregxidi (__o, 0); + __b.val[1] = (int16x4_t) __builtin_aarch64_get_dregxidi (__o, 1); + __b.val[2] = (int16x4_t) __builtin_aarch64_get_dregxidi (__o, 2); + __b.val[3] = (int16x4_t) __builtin_aarch64_get_dregxidi (__o, 3); + return __b; +} + +__extension__ extern __inline int32x2x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4_lane_s32 (const int32_t * __ptr, int32x2x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + int32x4x4_t __temp; + __temp.val[0] = vcombine_s32 (__b.val[0], vcreate_s32 (0)); + __temp.val[1] = vcombine_s32 (__b.val[1], vcreate_s32 (0)); + __temp.val[2] = vcombine_s32 (__b.val[2], vcreate_s32 (0)); + __temp.val[3] = vcombine_s32 (__b.val[3], vcreate_s32 (0)); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __temp.val[0], 0); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __temp.val[1], 1); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __temp.val[2], 2); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __temp.val[3], 3); + __o = __builtin_aarch64_ld4_lanev2si ( + (__builtin_aarch64_simd_si *) __ptr, __o, __c); + __b.val[0] = (int32x2_t) __builtin_aarch64_get_dregxidi (__o, 0); + __b.val[1] = (int32x2_t) __builtin_aarch64_get_dregxidi (__o, 1); + __b.val[2] = (int32x2_t) __builtin_aarch64_get_dregxidi (__o, 2); + __b.val[3] = (int32x2_t) __builtin_aarch64_get_dregxidi (__o, 3); + return __b; +} + +__extension__ extern __inline int64x1x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4_lane_s64 (const int64_t * __ptr, int64x1x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + int64x2x4_t __temp; + __temp.val[0] = vcombine_s64 (__b.val[0], vcreate_s64 (0)); + __temp.val[1] = vcombine_s64 (__b.val[1], vcreate_s64 (0)); + __temp.val[2] = vcombine_s64 (__b.val[2], vcreate_s64 (0)); + __temp.val[3] = vcombine_s64 (__b.val[3], vcreate_s64 (0)); + __o = __builtin_aarch64_set_qregxiv2di (__o, (int64x2_t) __temp.val[0], 0); + __o = __builtin_aarch64_set_qregxiv2di (__o, (int64x2_t) __temp.val[1], 1); + __o = __builtin_aarch64_set_qregxiv2di (__o, (int64x2_t) __temp.val[2], 2); + __o = __builtin_aarch64_set_qregxiv2di (__o, (int64x2_t) __temp.val[3], 3); + __o = __builtin_aarch64_ld4_lanedi ( + (__builtin_aarch64_simd_di *) __ptr, __o, __c); + __b.val[0] = (int64x1_t) __builtin_aarch64_get_dregxidi (__o, 0); + __b.val[1] = (int64x1_t) __builtin_aarch64_get_dregxidi (__o, 1); + __b.val[2] = (int64x1_t) __builtin_aarch64_get_dregxidi (__o, 2); + __b.val[3] = (int64x1_t) __builtin_aarch64_get_dregxidi (__o, 3); + return __b; +} + +__extension__ extern __inline float16x4x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4_lane_f16 (const float16_t * __ptr, float16x4x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + float16x8x4_t __temp; + __temp.val[0] = vcombine_f16 (__b.val[0], vcreate_f16 (0)); + __temp.val[1] = vcombine_f16 (__b.val[1], vcreate_f16 (0)); + __temp.val[2] = vcombine_f16 (__b.val[2], vcreate_f16 (0)); + __temp.val[3] = vcombine_f16 (__b.val[3], vcreate_f16 (0)); + __o = __builtin_aarch64_set_qregxiv8hf (__o, (float16x8_t) __temp.val[0], 0); + __o = __builtin_aarch64_set_qregxiv8hf (__o, (float16x8_t) __temp.val[1], 1); + __o = __builtin_aarch64_set_qregxiv8hf (__o, (float16x8_t) __temp.val[2], 2); + __o = __builtin_aarch64_set_qregxiv8hf (__o, (float16x8_t) __temp.val[3], 3); + __o = __builtin_aarch64_ld4_lanev4hf ( + (__builtin_aarch64_simd_hf *) __ptr, __o, __c); + __b.val[0] = (float16x4_t) __builtin_aarch64_get_dregxidi (__o, 0); + __b.val[1] = (float16x4_t) __builtin_aarch64_get_dregxidi (__o, 1); + __b.val[2] = (float16x4_t) __builtin_aarch64_get_dregxidi (__o, 2); + __b.val[3] = (float16x4_t) __builtin_aarch64_get_dregxidi (__o, 3); + return __b; +} + +__extension__ extern __inline float32x2x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4_lane_f32 (const float32_t * __ptr, float32x2x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + float32x4x4_t __temp; + __temp.val[0] = vcombine_f32 (__b.val[0], vcreate_f32 (0)); + __temp.val[1] = vcombine_f32 (__b.val[1], vcreate_f32 (0)); + __temp.val[2] = vcombine_f32 (__b.val[2], vcreate_f32 (0)); + __temp.val[3] = vcombine_f32 (__b.val[3], vcreate_f32 (0)); + __o = __builtin_aarch64_set_qregxiv4sf (__o, (float32x4_t) __temp.val[0], 0); + __o = __builtin_aarch64_set_qregxiv4sf (__o, (float32x4_t) __temp.val[1], 1); + __o = __builtin_aarch64_set_qregxiv4sf (__o, (float32x4_t) __temp.val[2], 2); + __o = __builtin_aarch64_set_qregxiv4sf (__o, (float32x4_t) __temp.val[3], 3); + __o = __builtin_aarch64_ld4_lanev2si ( + (__builtin_aarch64_simd_sf *) __ptr, __o, __c); + __b.val[0] = (float32x2_t) __builtin_aarch64_get_dregxidi (__o, 0); + __b.val[1] = (float32x2_t) __builtin_aarch64_get_dregxidi (__o, 1); + __b.val[2] = (float32x2_t) __builtin_aarch64_get_dregxidi (__o, 2); + __b.val[3] = (float32x2_t) __builtin_aarch64_get_dregxidi (__o, 3); + return __b; +} + +__extension__ extern __inline float64x1x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4_lane_f64 (const float64_t * __ptr, float64x1x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + float64x2x4_t __temp; + __temp.val[0] = vcombine_f64 (__b.val[0], vcreate_f64 (0)); + __temp.val[1] = vcombine_f64 (__b.val[1], vcreate_f64 (0)); + __temp.val[2] = vcombine_f64 (__b.val[2], vcreate_f64 (0)); + __temp.val[3] = vcombine_f64 (__b.val[3], vcreate_f64 (0)); + __o = __builtin_aarch64_set_qregxiv2df (__o, (float64x2_t) __temp.val[0], 0); + __o = __builtin_aarch64_set_qregxiv2df (__o, (float64x2_t) __temp.val[1], 1); + __o = __builtin_aarch64_set_qregxiv2df (__o, (float64x2_t) __temp.val[2], 2); + __o = __builtin_aarch64_set_qregxiv2df (__o, (float64x2_t) __temp.val[3], 3); + __o = __builtin_aarch64_ld4_lanedf ( + (__builtin_aarch64_simd_df *) __ptr, __o, __c); + __b.val[0] = (float64x1_t) __builtin_aarch64_get_dregxidi (__o, 0); + __b.val[1] = (float64x1_t) __builtin_aarch64_get_dregxidi (__o, 1); + __b.val[2] = (float64x1_t) __builtin_aarch64_get_dregxidi (__o, 2); + __b.val[3] = (float64x1_t) __builtin_aarch64_get_dregxidi (__o, 3); + return __b; +} + +__extension__ extern __inline poly8x8x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4_lane_p8 (const poly8_t * __ptr, poly8x8x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + poly8x16x4_t __temp; + __temp.val[0] = vcombine_p8 (__b.val[0], vcreate_p8 (0)); + __temp.val[1] = vcombine_p8 (__b.val[1], vcreate_p8 (0)); + __temp.val[2] = vcombine_p8 (__b.val[2], vcreate_p8 (0)); + __temp.val[3] = vcombine_p8 (__b.val[3], vcreate_p8 (0)); + __o = __builtin_aarch64_set_qregxiv16qi (__o, (int8x16_t) __temp.val[0], 0); + __o = __builtin_aarch64_set_qregxiv16qi (__o, (int8x16_t) __temp.val[1], 1); + __o = __builtin_aarch64_set_qregxiv16qi (__o, (int8x16_t) __temp.val[2], 2); + __o = __builtin_aarch64_set_qregxiv16qi (__o, (int8x16_t) __temp.val[3], 3); + __o = __builtin_aarch64_ld4_lanev8qi ( + (__builtin_aarch64_simd_qi *) __ptr, __o, __c); + __b.val[0] = (poly8x8_t) __builtin_aarch64_get_dregxidi (__o, 0); + __b.val[1] = (poly8x8_t) __builtin_aarch64_get_dregxidi (__o, 1); + __b.val[2] = (poly8x8_t) __builtin_aarch64_get_dregxidi (__o, 2); + __b.val[3] = (poly8x8_t) __builtin_aarch64_get_dregxidi (__o, 3); + return __b; +} -__LD4_LANE_FUNC (float16x4x4_t, float16x4_t, float16x8x4_t, float16_t, v4hf, - v8hf, hf, f16, float16x8_t) -__LD4_LANE_FUNC (float32x2x4_t, float32x2_t, float32x4x4_t, float32_t, v2sf, v4sf, - sf, f32, float32x4_t) -__LD4_LANE_FUNC (float64x1x4_t, float64x1_t, float64x2x4_t, float64_t, df, v2df, - df, f64, float64x2_t) -__LD4_LANE_FUNC (poly8x8x4_t, poly8x8_t, poly8x16x4_t, poly8_t, v8qi, v16qi, qi, p8, - int8x16_t) -__LD4_LANE_FUNC (poly16x4x4_t, poly16x4_t, poly16x8x4_t, poly16_t, v4hi, v8hi, hi, - p16, int16x8_t) -__LD4_LANE_FUNC (poly64x1x4_t, poly64x1_t, poly64x2x4_t, poly64_t, di, - v2di_ssps, di, p64, poly64x2_t) -__LD4_LANE_FUNC (int8x8x4_t, int8x8_t, int8x16x4_t, int8_t, v8qi, v16qi, qi, s8, - int8x16_t) -__LD4_LANE_FUNC (int16x4x4_t, int16x4_t, int16x8x4_t, int16_t, v4hi, v8hi, hi, s16, - int16x8_t) -__LD4_LANE_FUNC (int32x2x4_t, int32x2_t, int32x4x4_t, int32_t, v2si, v4si, si, s32, - int32x4_t) -__LD4_LANE_FUNC (int64x1x4_t, int64x1_t, int64x2x4_t, int64_t, di, v2di, di, s64, - int64x2_t) -__LD4_LANE_FUNC (uint8x8x4_t, uint8x8_t, uint8x16x4_t, uint8_t, v8qi, v16qi, qi, u8, - int8x16_t) -__LD4_LANE_FUNC (uint16x4x4_t, uint16x4_t, uint16x8x4_t, uint16_t, v4hi, v8hi, hi, - u16, int16x8_t) -__LD4_LANE_FUNC (uint32x2x4_t, uint32x2_t, uint32x4x4_t, uint32_t, v2si, v4si, si, - u32, int32x4_t) -__LD4_LANE_FUNC (uint64x1x4_t, uint64x1_t, uint64x2x4_t, uint64_t, di, v2di, di, - u64, int64x2_t) +__extension__ extern __inline poly16x4x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4_lane_p16 (const poly16_t * __ptr, poly16x4x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + poly16x8x4_t __temp; + __temp.val[0] = vcombine_p16 (__b.val[0], vcreate_p16 (0)); + __temp.val[1] = vcombine_p16 (__b.val[1], vcreate_p16 (0)); + __temp.val[2] = vcombine_p16 (__b.val[2], vcreate_p16 (0)); + __temp.val[3] = vcombine_p16 (__b.val[3], vcreate_p16 (0)); + __o = __builtin_aarch64_set_qregxiv8hi (__o, (int16x8_t) __temp.val[0], 0); + __o = __builtin_aarch64_set_qregxiv8hi (__o, (int16x8_t) __temp.val[1], 1); + __o = __builtin_aarch64_set_qregxiv8hi (__o, (int16x8_t) __temp.val[2], 2); + __o = __builtin_aarch64_set_qregxiv8hi (__o, (int16x8_t) __temp.val[3], 3); + __o = __builtin_aarch64_ld4_lanev4hi ( + (__builtin_aarch64_simd_hi *) __ptr, __o, __c); + __b.val[0] = (poly16x4_t) __builtin_aarch64_get_dregxidi (__o, 0); + __b.val[1] = (poly16x4_t) __builtin_aarch64_get_dregxidi (__o, 1); + __b.val[2] = (poly16x4_t) __builtin_aarch64_get_dregxidi (__o, 2); + __b.val[3] = (poly16x4_t) __builtin_aarch64_get_dregxidi (__o, 3); + return __b; +} + +__extension__ extern __inline poly64x1x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4_lane_p64 (const poly64_t * __ptr, poly64x1x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + poly64x2x4_t __temp; + __temp.val[0] = vcombine_p64 (__b.val[0], vcreate_p64 (0)); + __temp.val[1] = vcombine_p64 (__b.val[1], vcreate_p64 (0)); + __temp.val[2] = vcombine_p64 (__b.val[2], vcreate_p64 (0)); + __temp.val[3] = vcombine_p64 (__b.val[3], vcreate_p64 (0)); + __o = __builtin_aarch64_set_qregxiv2di (__o, (int64x2_t) __temp.val[0], 0); + __o = __builtin_aarch64_set_qregxiv2di (__o, (int64x2_t) __temp.val[1], 1); + __o = __builtin_aarch64_set_qregxiv2di (__o, (int64x2_t) __temp.val[2], 2); + __o = __builtin_aarch64_set_qregxiv2di (__o, (int64x2_t) __temp.val[3], 3); + __o = __builtin_aarch64_ld4_lanedi ( + (__builtin_aarch64_simd_di *) __ptr, __o, __c); + __b.val[0] = (poly64x1_t) __builtin_aarch64_get_dregxidi (__o, 0); + __b.val[1] = (poly64x1_t) __builtin_aarch64_get_dregxidi (__o, 1); + __b.val[2] = (poly64x1_t) __builtin_aarch64_get_dregxidi (__o, 2); + __b.val[3] = (poly64x1_t) __builtin_aarch64_get_dregxidi (__o, 3); + return __b; +} /* vld4q_lane */ -#define __LD4Q_LANE_FUNC(intype, vtype, ptrtype, mode, ptrmode, funcsuffix) \ -__extension__ extern __inline intype \ -__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) \ -vld4q_lane_##funcsuffix (const ptrtype * __ptr, intype __b, const int __c) \ -{ \ - __builtin_aarch64_simd_xi __o; \ - intype ret; \ - __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[0], 0); \ - __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[1], 1); \ - __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[2], 2); \ - __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[3], 3); \ - __o = __builtin_aarch64_ld4_lane##mode ( \ - (__builtin_aarch64_simd_##ptrmode *) __ptr, __o, __c); \ - ret.val[0] = (vtype) __builtin_aarch64_get_qregxiv4si (__o, 0); \ - ret.val[1] = (vtype) __builtin_aarch64_get_qregxiv4si (__o, 1); \ - ret.val[2] = (vtype) __builtin_aarch64_get_qregxiv4si (__o, 2); \ - ret.val[3] = (vtype) __builtin_aarch64_get_qregxiv4si (__o, 3); \ - return ret; \ -} - -__LD4Q_LANE_FUNC (float16x8x4_t, float16x8_t, float16_t, v8hf, hf, f16) -__LD4Q_LANE_FUNC (float32x4x4_t, float32x4_t, float32_t, v4sf, sf, f32) -__LD4Q_LANE_FUNC (float64x2x4_t, float64x2_t, float64_t, v2df, df, f64) -__LD4Q_LANE_FUNC (poly8x16x4_t, poly8x16_t, poly8_t, v16qi, qi, p8) -__LD4Q_LANE_FUNC (poly16x8x4_t, poly16x8_t, poly16_t, v8hi, hi, p16) -__LD4Q_LANE_FUNC (poly64x2x4_t, poly64x2_t, poly64_t, v2di, di, p64) -__LD4Q_LANE_FUNC (int8x16x4_t, int8x16_t, int8_t, v16qi, qi, s8) -__LD4Q_LANE_FUNC (int16x8x4_t, int16x8_t, int16_t, v8hi, hi, s16) -__LD4Q_LANE_FUNC (int32x4x4_t, int32x4_t, int32_t, v4si, si, s32) -__LD4Q_LANE_FUNC (int64x2x4_t, int64x2_t, int64_t, v2di, di, s64) -__LD4Q_LANE_FUNC (uint8x16x4_t, uint8x16_t, uint8_t, v16qi, qi, u8) -__LD4Q_LANE_FUNC (uint16x8x4_t, uint16x8_t, uint16_t, v8hi, hi, u16) -__LD4Q_LANE_FUNC (uint32x4x4_t, uint32x4_t, uint32_t, v4si, si, u32) -__LD4Q_LANE_FUNC (uint64x2x4_t, uint64x2_t, uint64_t, v2di, di, u64) +__extension__ extern __inline uint8x16x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4q_lane_u8 (const uint8_t * __ptr, uint8x16x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + uint8x16x4_t ret; + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[0], 0); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[1], 1); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[2], 2); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[3], 3); + __o = __builtin_aarch64_ld4_lanev16qi ( + (__builtin_aarch64_simd_qi *) __ptr, __o, __c); + ret.val[0] = (uint8x16_t) __builtin_aarch64_get_qregxiv4si (__o, 0); + ret.val[1] = (uint8x16_t) __builtin_aarch64_get_qregxiv4si (__o, 1); + ret.val[2] = (uint8x16_t) __builtin_aarch64_get_qregxiv4si (__o, 2); + ret.val[3] = (uint8x16_t) __builtin_aarch64_get_qregxiv4si (__o, 3); + return ret; +} + +__extension__ extern __inline uint16x8x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4q_lane_u16 (const uint16_t * __ptr, uint16x8x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + uint16x8x4_t ret; + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[0], 0); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[1], 1); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[2], 2); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[3], 3); + __o = __builtin_aarch64_ld4_lanev8hi ( + (__builtin_aarch64_simd_hi *) __ptr, __o, __c); + ret.val[0] = (uint16x8_t) __builtin_aarch64_get_qregxiv4si (__o, 0); + ret.val[1] = (uint16x8_t) __builtin_aarch64_get_qregxiv4si (__o, 1); + ret.val[2] = (uint16x8_t) __builtin_aarch64_get_qregxiv4si (__o, 2); + ret.val[3] = (uint16x8_t) __builtin_aarch64_get_qregxiv4si (__o, 3); + return ret; +} + +__extension__ extern __inline uint32x4x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4q_lane_u32 (const uint32_t * __ptr, uint32x4x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + uint32x4x4_t ret; + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[0], 0); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[1], 1); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[2], 2); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[3], 3); + __o = __builtin_aarch64_ld4_lanev4si ( + (__builtin_aarch64_simd_si *) __ptr, __o, __c); + ret.val[0] = (uint32x4_t) __builtin_aarch64_get_qregxiv4si (__o, 0); + ret.val[1] = (uint32x4_t) __builtin_aarch64_get_qregxiv4si (__o, 1); + ret.val[2] = (uint32x4_t) __builtin_aarch64_get_qregxiv4si (__o, 2); + ret.val[3] = (uint32x4_t) __builtin_aarch64_get_qregxiv4si (__o, 3); + return ret; +} + +__extension__ extern __inline uint64x2x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4q_lane_u64 (const uint64_t * __ptr, uint64x2x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + uint64x2x4_t ret; + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[0], 0); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[1], 1); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[2], 2); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[3], 3); + __o = __builtin_aarch64_ld4_lanev2di ( + (__builtin_aarch64_simd_di *) __ptr, __o, __c); + ret.val[0] = (uint64x2_t) __builtin_aarch64_get_qregxiv4si (__o, 0); + ret.val[1] = (uint64x2_t) __builtin_aarch64_get_qregxiv4si (__o, 1); + ret.val[2] = (uint64x2_t) __builtin_aarch64_get_qregxiv4si (__o, 2); + ret.val[3] = (uint64x2_t) __builtin_aarch64_get_qregxiv4si (__o, 3); + return ret; +} + +__extension__ extern __inline int8x16x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4q_lane_s8 (const int8_t * __ptr, int8x16x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + int8x16x4_t ret; + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[0], 0); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[1], 1); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[2], 2); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[3], 3); + __o = __builtin_aarch64_ld4_lanev16qi ( + (__builtin_aarch64_simd_qi *) __ptr, __o, __c); + ret.val[0] = (int8x16_t) __builtin_aarch64_get_qregxiv4si (__o, 0); + ret.val[1] = (int8x16_t) __builtin_aarch64_get_qregxiv4si (__o, 1); + ret.val[2] = (int8x16_t) __builtin_aarch64_get_qregxiv4si (__o, 2); + ret.val[3] = (int8x16_t) __builtin_aarch64_get_qregxiv4si (__o, 3); + return ret; +} + +__extension__ extern __inline int16x8x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4q_lane_s16 (const int16_t * __ptr, int16x8x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + int16x8x4_t ret; + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[0], 0); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[1], 1); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[2], 2); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[3], 3); + __o = __builtin_aarch64_ld4_lanev8hi ( + (__builtin_aarch64_simd_hi *) __ptr, __o, __c); + ret.val[0] = (int16x8_t) __builtin_aarch64_get_qregxiv4si (__o, 0); + ret.val[1] = (int16x8_t) __builtin_aarch64_get_qregxiv4si (__o, 1); + ret.val[2] = (int16x8_t) __builtin_aarch64_get_qregxiv4si (__o, 2); + ret.val[3] = (int16x8_t) __builtin_aarch64_get_qregxiv4si (__o, 3); + return ret; +} + +__extension__ extern __inline int32x4x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4q_lane_s32 (const int32_t * __ptr, int32x4x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + int32x4x4_t ret; + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[0], 0); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[1], 1); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[2], 2); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[3], 3); + __o = __builtin_aarch64_ld4_lanev4si ( + (__builtin_aarch64_simd_si *) __ptr, __o, __c); + ret.val[0] = (int32x4_t) __builtin_aarch64_get_qregxiv4si (__o, 0); + ret.val[1] = (int32x4_t) __builtin_aarch64_get_qregxiv4si (__o, 1); + ret.val[2] = (int32x4_t) __builtin_aarch64_get_qregxiv4si (__o, 2); + ret.val[3] = (int32x4_t) __builtin_aarch64_get_qregxiv4si (__o, 3); + return ret; +} + +__extension__ extern __inline int64x2x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4q_lane_s64 (const int64_t * __ptr, int64x2x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + int64x2x4_t ret; + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[0], 0); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[1], 1); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[2], 2); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[3], 3); + __o = __builtin_aarch64_ld4_lanev2di ( + (__builtin_aarch64_simd_di *) __ptr, __o, __c); + ret.val[0] = (int64x2_t) __builtin_aarch64_get_qregxiv4si (__o, 0); + ret.val[1] = (int64x2_t) __builtin_aarch64_get_qregxiv4si (__o, 1); + ret.val[2] = (int64x2_t) __builtin_aarch64_get_qregxiv4si (__o, 2); + ret.val[3] = (int64x2_t) __builtin_aarch64_get_qregxiv4si (__o, 3); + return ret; +} + +__extension__ extern __inline float16x8x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4q_lane_f16 (const float16_t * __ptr, float16x8x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + float16x8x4_t ret; + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[0], 0); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[1], 1); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[2], 2); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[3], 3); + __o = __builtin_aarch64_ld4_lanev8hf ( + (__builtin_aarch64_simd_hf *) __ptr, __o, __c); + ret.val[0] = (float16x8_t) __builtin_aarch64_get_qregxiv4si (__o, 0); + ret.val[1] = (float16x8_t) __builtin_aarch64_get_qregxiv4si (__o, 1); + ret.val[2] = (float16x8_t) __builtin_aarch64_get_qregxiv4si (__o, 2); + ret.val[3] = (float16x8_t) __builtin_aarch64_get_qregxiv4si (__o, 3); + return ret; +} + +__extension__ extern __inline float32x4x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4q_lane_f32 (const float32_t * __ptr, float32x4x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + float32x4x4_t ret; + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[0], 0); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[1], 1); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[2], 2); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[3], 3); + __o = __builtin_aarch64_ld4_lanev4sf ( + (__builtin_aarch64_simd_sf *) __ptr, __o, __c); + ret.val[0] = (float32x4_t) __builtin_aarch64_get_qregxiv4si (__o, 0); + ret.val[1] = (float32x4_t) __builtin_aarch64_get_qregxiv4si (__o, 1); + ret.val[2] = (float32x4_t) __builtin_aarch64_get_qregxiv4si (__o, 2); + ret.val[3] = (float32x4_t) __builtin_aarch64_get_qregxiv4si (__o, 3); + return ret; +} + +__extension__ extern __inline float64x2x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4q_lane_f64 (const float64_t * __ptr, float64x2x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + float64x2x4_t ret; + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[0], 0); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[1], 1); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[2], 2); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[3], 3); + __o = __builtin_aarch64_ld4_lanev2df ( + (__builtin_aarch64_simd_di *) __ptr, __o, __c); + ret.val[0] = (float64x2_t) __builtin_aarch64_get_qregxiv4si (__o, 0); + ret.val[1] = (float64x2_t) __builtin_aarch64_get_qregxiv4si (__o, 1); + ret.val[2] = (float64x2_t) __builtin_aarch64_get_qregxiv4si (__o, 2); + ret.val[3] = (float64x2_t) __builtin_aarch64_get_qregxiv4si (__o, 3); + return ret; +} + +__extension__ extern __inline poly8x16x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4q_lane_p8 (const poly8_t * __ptr, poly8x16x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + poly8x16x4_t ret; + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[0], 0); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[1], 1); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[2], 2); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[3], 3); + __o = __builtin_aarch64_ld4_lanev16qi ( + (__builtin_aarch64_simd_qi *) __ptr, __o, __c); + ret.val[0] = (poly8x16_t) __builtin_aarch64_get_qregxiv4si (__o, 0); + ret.val[1] = (poly8x16_t) __builtin_aarch64_get_qregxiv4si (__o, 1); + ret.val[2] = (poly8x16_t) __builtin_aarch64_get_qregxiv4si (__o, 2); + ret.val[3] = (poly8x16_t) __builtin_aarch64_get_qregxiv4si (__o, 3); + return ret; +} + +__extension__ extern __inline poly16x8x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4q_lane_p16 (const poly16_t * __ptr, poly16x8x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + poly16x8x4_t ret; + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[0], 0); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[1], 1); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[2], 2); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[3], 3); + __o = __builtin_aarch64_ld4_lanev8hi ( + (__builtin_aarch64_simd_hi *) __ptr, __o, __c); + ret.val[0] = (poly16x8_t) __builtin_aarch64_get_qregxiv4si (__o, 0); + ret.val[1] = (poly16x8_t) __builtin_aarch64_get_qregxiv4si (__o, 1); + ret.val[2] = (poly16x8_t) __builtin_aarch64_get_qregxiv4si (__o, 2); + ret.val[3] = (poly16x8_t) __builtin_aarch64_get_qregxiv4si (__o, 3); + return ret; +} + +__extension__ extern __inline poly64x2x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4q_lane_p64 (const poly64_t * __ptr, poly64x2x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + poly64x2x4_t ret; + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[0], 0); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[1], 1); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[2], 2); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[3], 3); + __o = __builtin_aarch64_ld4_lanev2di ( + (__builtin_aarch64_simd_di *) __ptr, __o, __c); + ret.val[0] = (poly64x2_t) __builtin_aarch64_get_qregxiv4si (__o, 0); + ret.val[1] = (poly64x2_t) __builtin_aarch64_get_qregxiv4si (__o, 1); + ret.val[2] = (poly64x2_t) __builtin_aarch64_get_qregxiv4si (__o, 2); + ret.val[3] = (poly64x2_t) __builtin_aarch64_get_qregxiv4si (__o, 3); + return ret; +} /* vmax */ @@ -35441,9 +35926,47 @@ vld3q_lane_bf16 (const bfloat16_t * __ptr, bfloat16x8x3_t __b, const int __c) return ret; } -__LD4_LANE_FUNC (bfloat16x4x4_t, bfloat16x4_t, bfloat16x8x4_t, bfloat16_t, v4bf, - v8bf, bf, bf16, bfloat16x8_t) -__LD4Q_LANE_FUNC (bfloat16x8x4_t, bfloat16x8_t, bfloat16_t, v8bf, bf, bf16) +__extension__ extern __inline bfloat16x4x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4_lane_bf16 (const bfloat16_t * __ptr, bfloat16x4x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + bfloat16x8x4_t __temp; + __temp.val[0] = vcombine_bf16 (__b.val[0], vcreate_bf16 (0)); + __temp.val[1] = vcombine_bf16 (__b.val[1], vcreate_bf16 (0)); + __temp.val[2] = vcombine_bf16 (__b.val[2], vcreate_bf16 (0)); + __temp.val[3] = vcombine_bf16 (__b.val[3], vcreate_bf16 (0)); + __o = __builtin_aarch64_set_qregxiv8bf (__o, (bfloat16x8_t) __temp.val[0], 0); + __o = __builtin_aarch64_set_qregxiv8bf (__o, (bfloat16x8_t) __temp.val[1], 1); + __o = __builtin_aarch64_set_qregxiv8bf (__o, (bfloat16x8_t) __temp.val[2], 2); + __o = __builtin_aarch64_set_qregxiv8bf (__o, (bfloat16x8_t) __temp.val[3], 3); + __o = __builtin_aarch64_ld4_lanev4bf ( + (__builtin_aarch64_simd_bf *) __ptr, __o, __c); + __b.val[0] = (bfloat16x4_t) __builtin_aarch64_get_dregxidi (__o, 0); + __b.val[1] = (bfloat16x4_t) __builtin_aarch64_get_dregxidi (__o, 1); + __b.val[2] = (bfloat16x4_t) __builtin_aarch64_get_dregxidi (__o, 2); + __b.val[3] = (bfloat16x4_t) __builtin_aarch64_get_dregxidi (__o, 3); + return __b; +} + +__extension__ extern __inline bfloat16x8x4_t +__attribute__ ((__always_inline__, __gnu_inline__,__artificial__)) +vld4q_lane_bf16 (const bfloat16_t * __ptr, bfloat16x8x4_t __b, const int __c) +{ + __builtin_aarch64_simd_xi __o; + bfloat16x8x4_t ret; + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[0], 0); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[1], 1); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[2], 2); + __o = __builtin_aarch64_set_qregxiv4si (__o, (int32x4_t) __b.val[3], 3); + __o = __builtin_aarch64_ld4_lanev8bf ( + (__builtin_aarch64_simd_bf *) __ptr, __o, __c); + ret.val[0] = (bfloat16x8_t) __builtin_aarch64_get_qregxiv4si (__o, 0); + ret.val[1] = (bfloat16x8_t) __builtin_aarch64_get_qregxiv4si (__o, 1); + ret.val[2] = (bfloat16x8_t) __builtin_aarch64_get_qregxiv4si (__o, 2); + ret.val[3] = (bfloat16x8_t) __builtin_aarch64_get_qregxiv4si (__o, 3); + return ret; +} __extension__ extern __inline void __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) @@ -35739,7 +36262,4 @@ vaddq_p128 (poly128_t __a, poly128_t __b) #undef __aarch64_vdupq_laneq_u32 #undef __aarch64_vdupq_laneq_u64 -#undef __LD4_LANE_FUNC -#undef __LD4Q_LANE_FUNC - #endif