/* SPDX-License-Identifier: MIT * * Permission is hereby granted, free of charge, to any person * obtaining a copy of this software and associated documentation * files (the "Software"), to deal in the Software without * restriction, including without limitation the rights to use, copy, * modify, merge, publish, distribute, sublicense, and/or sell copies * of the Software, and to permit persons to whom the Software is * furnished to do so, subject to the following conditions: * * The above copyright notice and this permission notice shall be * included in all copies or substantial portions of the Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. * * Copyright: * 2020 Evan Nemerson * 2020 Sean Maher (Copyright owned by Google, LLC) * 2023 Yi-Yen Chung (Copyright owned by Andes Technology) */ #if !defined(SIMDE_ARM_NEON_UZP1_H) #define SIMDE_ARM_NEON_UZP1_H #include "types.h" HEDLEY_DIAGNOSTIC_PUSH SIMDE_DISABLE_UNWANTED_DIAGNOSTICS SIMDE_BEGIN_DECLS_ SIMDE_FUNCTION_ATTRIBUTES simde_float16x4_t simde_vuzp1_f16(simde_float16x4_t a, simde_float16x4_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) && defined(SIMDE_ARM_NEON_FP16) return vuzp1_f16(a, b); #else simde_float16x4_private r_, a_ = simde_float16x4_to_private(a), b_ = simde_float16x4_to_private(b); const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } return simde_float16x4_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1_f16 #define vuzp1_f16(a, b) simde_vuzp1_f16((a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_float32x2_t simde_vuzp1_f32(simde_float32x2_t a, simde_float32x2_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) return vuzp1_f32(a, b); #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) float32x2x2_t t = vuzp_f32(a, b); return t.val[0]; #else simde_float32x2_private r_, a_ = simde_float32x2_to_private(a), b_ = simde_float32x2_to_private(b); #if defined(SIMDE_SHUFFLE_VECTOR_) r_.values = SIMDE_SHUFFLE_VECTOR_(32, 8, a_.values, b_.values, 0, 2); #else const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } #endif return simde_float32x2_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1_f32 #define vuzp1_f32(a, b) simde_vuzp1_f32((a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_int8x8_t simde_vuzp1_s8(simde_int8x8_t a, simde_int8x8_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) return vuzp1_s8(a, b); #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) int8x8x2_t t = vuzp_s8(a, b); return t.val[0]; #else simde_int8x8_private r_, a_ = simde_int8x8_to_private(a), b_ = simde_int8x8_to_private(b); #if defined(SIMDE_SHUFFLE_VECTOR_) r_.values = SIMDE_SHUFFLE_VECTOR_(8, 8, a_.values, b_.values, 0, 2, 4, 6, 8, 10, 12, 14); #else const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } #endif return simde_int8x8_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1_s8 #define vuzp1_s8(a, b) simde_vuzp1_s8((a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_int16x4_t simde_vuzp1_s16(simde_int16x4_t a, simde_int16x4_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) return vuzp1_s16(a, b); #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) int16x4x2_t t = vuzp_s16(a, b); return t.val[0]; #else simde_int16x4_private r_, a_ = simde_int16x4_to_private(a), b_ = simde_int16x4_to_private(b); #if defined(SIMDE_SHUFFLE_VECTOR_) r_.values = SIMDE_SHUFFLE_VECTOR_(16, 8, a_.values, b_.values, 0, 2, 4, 6); #else const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } #endif return simde_int16x4_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1_s16 #define vuzp1_s16(a, b) simde_vuzp1_s16((a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_int32x2_t simde_vuzp1_s32(simde_int32x2_t a, simde_int32x2_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) return vuzp1_s32(a, b); #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) int32x2x2_t t = vuzp_s32(a, b); return t.val[0]; #else simde_int32x2_private r_, a_ = simde_int32x2_to_private(a), b_ = simde_int32x2_to_private(b); #if defined(SIMDE_SHUFFLE_VECTOR_) r_.values = SIMDE_SHUFFLE_VECTOR_(32, 8, a_.values, b_.values, 0, 2); #else const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } #endif return simde_int32x2_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1_s32 #define vuzp1_s32(a, b) simde_vuzp1_s32((a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_uint8x8_t simde_vuzp1_u8(simde_uint8x8_t a, simde_uint8x8_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) return vuzp1_u8(a, b); #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) uint8x8x2_t t = vuzp_u8(a, b); return t.val[0]; #else simde_uint8x8_private r_, a_ = simde_uint8x8_to_private(a), b_ = simde_uint8x8_to_private(b); #if defined(SIMDE_SHUFFLE_VECTOR_) r_.values = SIMDE_SHUFFLE_VECTOR_(8, 8, a_.values, b_.values, 0, 2, 4, 6, 8, 10, 12, 14); #else const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } #endif return simde_uint8x8_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1_u8 #define vuzp1_u8(a, b) simde_vuzp1_u8((a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_uint16x4_t simde_vuzp1_u16(simde_uint16x4_t a, simde_uint16x4_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) return vuzp1_u16(a, b); #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) uint16x4x2_t t = vuzp_u16(a, b); return t.val[0]; #else simde_uint16x4_private r_, a_ = simde_uint16x4_to_private(a), b_ = simde_uint16x4_to_private(b); #if defined(SIMDE_SHUFFLE_VECTOR_) r_.values = SIMDE_SHUFFLE_VECTOR_(16, 8, a_.values, b_.values, 0, 2, 4, 6); #else const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } #endif return simde_uint16x4_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1_u16 #define vuzp1_u16(a, b) simde_vuzp1_u16((a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_uint32x2_t simde_vuzp1_u32(simde_uint32x2_t a, simde_uint32x2_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) return vuzp1_u32(a, b); #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) uint32x2x2_t t = vuzp_u32(a, b); return t.val[0]; #else simde_uint32x2_private r_, a_ = simde_uint32x2_to_private(a), b_ = simde_uint32x2_to_private(b); #if defined(SIMDE_SHUFFLE_VECTOR_) r_.values = SIMDE_SHUFFLE_VECTOR_(32, 8, a_.values, b_.values, 0, 2); #else const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } #endif return simde_uint32x2_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1_u32 #define vuzp1_u32(a, b) simde_vuzp1_u32((a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_float16x8_t simde_vuzp1q_f16(simde_float16x8_t a, simde_float16x8_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) && defined(SIMDE_ARM_NEON_FP16) return vuzp1q_f16(a, b); #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) && defined(SIMDE_ARM_NEON_FP16) float16x8x2_t t = vuzpq_f16(a, b); return t.val[0]; #else simde_float16x8_private r_, a_ = simde_float16x8_to_private(a), b_ = simde_float16x8_to_private(b); const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } return simde_float16x8_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1q_f16 #define vuzp1q_f16(a, b) simde_vuzp1q_f16((a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_float32x4_t simde_vuzp1q_f32(simde_float32x4_t a, simde_float32x4_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) return vuzp1q_f32(a, b); #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) float32x4x2_t t = vuzpq_f32(a, b); return t.val[0]; #else simde_float32x4_private r_, a_ = simde_float32x4_to_private(a), b_ = simde_float32x4_to_private(b); #if defined(SIMDE_WASM_SIMD128_NATIVE) r_.v128 = wasm_i32x4_shuffle(a_.v128, b_.v128, 0, 2, 4, 6); #elif defined(SIMDE_X86_SSE_NATIVE) r_.m128 = _mm_shuffle_ps(a_.m128, b_.m128, 0x88); #elif defined(SIMDE_SHUFFLE_VECTOR_) r_.values = SIMDE_SHUFFLE_VECTOR_(32, 16, a_.values, b_.values, 0, 2, 4, 6); #else const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } #endif return simde_float32x4_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1q_f32 #define vuzp1q_f32(a, b) simde_vuzp1q_f32((a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_float64x2_t simde_vuzp1q_f64(simde_float64x2_t a, simde_float64x2_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) return vuzp1q_f64(a, b); #else simde_float64x2_private r_, a_ = simde_float64x2_to_private(a), b_ = simde_float64x2_to_private(b); #if defined(SIMDE_WASM_SIMD128_NATIVE) r_.v128 = wasm_i64x2_shuffle(a_.v128, b_.v128, 0, 2); #elif defined(SIMDE_X86_SSE2_NATIVE) r_.m128d = _mm_castps_pd(_mm_movelh_ps(_mm_castpd_ps(a_.m128d), _mm_castpd_ps(b_.m128d))); #elif defined(SIMDE_SHUFFLE_VECTOR_) r_.values = SIMDE_SHUFFLE_VECTOR_(64, 16, a_.values, b_.values, 0, 2); #else const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } #endif return simde_float64x2_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1q_f64 #define vuzp1q_f64(a, b) simde_vuzp1q_f64((a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_int8x16_t simde_vuzp1q_s8(simde_int8x16_t a, simde_int8x16_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) return vuzp1q_s8(a, b); #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) int8x16x2_t t = vuzpq_s8(a, b); return t.val[0]; #else simde_int8x16_private r_, a_ = simde_int8x16_to_private(a), b_ = simde_int8x16_to_private(b); #if defined(SIMDE_WASM_SIMD128_NATIVE) r_.v128 = wasm_i8x16_shuffle(a_.v128, b_.v128, 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30); #elif defined(SIMDE_SHUFFLE_VECTOR_) r_.values = SIMDE_SHUFFLE_VECTOR_(8, 16, a_.values, b_.values, 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30); #else const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } #endif return simde_int8x16_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1q_s8 #define vuzp1q_s8(a, b) simde_vuzp1q_s8((a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_int16x8_t simde_vuzp1q_s16(simde_int16x8_t a, simde_int16x8_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) return vuzp1q_s16(a, b); #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) int16x8x2_t t = vuzpq_s16(a, b); return t.val[0]; #else simde_int16x8_private r_, a_ = simde_int16x8_to_private(a), b_ = simde_int16x8_to_private(b); #if defined(SIMDE_WASM_SIMD128_NATIVE) r_.v128 = wasm_i16x8_shuffle(a_.v128, b_.v128, 0, 2, 4, 6, 8, 10, 12, 14); #elif defined(SIMDE_SHUFFLE_VECTOR_) r_.values = SIMDE_SHUFFLE_VECTOR_(16, 16, a_.values, b_.values, 0, 2, 4, 6, 8, 10, 12, 14); #else const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } #endif return simde_int16x8_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1q_s16 #define vuzp1q_s16(a, b) simde_vuzp1q_s16((a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_int32x4_t simde_vuzp1q_s32(simde_int32x4_t a, simde_int32x4_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) return vuzp1q_s32(a, b); #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) int32x4x2_t t = vuzpq_s32(a, b); return t.val[0]; #else simde_int32x4_private r_, a_ = simde_int32x4_to_private(a), b_ = simde_int32x4_to_private(b); #if defined(SIMDE_WASM_SIMD128_NATIVE) r_.v128 = wasm_i32x4_shuffle(a_.v128, b_.v128, 0, 2, 4, 6); #elif defined(SIMDE_X86_SSE2_NATIVE) r_.m128i = _mm_castps_si128(_mm_shuffle_ps(_mm_castsi128_ps(a_.m128i), _mm_castsi128_ps(b_.m128i), 0x88)); #elif defined(SIMDE_SHUFFLE_VECTOR_) r_.values = SIMDE_SHUFFLE_VECTOR_(32, 16, a_.values, b_.values, 0, 2, 4, 6); #else const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } #endif return simde_int32x4_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1q_s32 #define vuzp1q_s32(a, b) simde_vuzp1q_s32((a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_int64x2_t simde_vuzp1q_s64(simde_int64x2_t a, simde_int64x2_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) return vuzp1q_s64(a, b); #else simde_int64x2_private r_, a_ = simde_int64x2_to_private(a), b_ = simde_int64x2_to_private(b); #if defined(SIMDE_WASM_SIMD128_NATIVE) r_.v128 = wasm_i64x2_shuffle(a_.v128, b_.v128, 0, 2); #elif defined(SIMDE_X86_SSE2_NATIVE) r_.m128i = _mm_castps_si128(_mm_movelh_ps(_mm_castsi128_ps(a_.m128i), _mm_castsi128_ps(b_.m128i))); #elif defined(SIMDE_SHUFFLE_VECTOR_) r_.values = SIMDE_SHUFFLE_VECTOR_(64, 16, a_.values, b_.values, 0, 2); #else const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } #endif return simde_int64x2_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1q_s64 #define vuzp1q_s64(a, b) simde_vuzp1q_s64((a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_uint8x16_t simde_vuzp1q_u8(simde_uint8x16_t a, simde_uint8x16_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) return vuzp1q_u8(a, b); #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) uint8x16x2_t t = vuzpq_u8(a, b); return t.val[0]; #else simde_uint8x16_private r_, a_ = simde_uint8x16_to_private(a), b_ = simde_uint8x16_to_private(b); #if defined(SIMDE_WASM_SIMD128_NATIVE) r_.v128 = wasm_i8x16_shuffle(a_.v128, b_.v128, 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30); #elif defined(SIMDE_SHUFFLE_VECTOR_) r_.values = SIMDE_SHUFFLE_VECTOR_(8, 16, a_.values, b_.values, 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30); #else const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } #endif return simde_uint8x16_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1q_u8 #define vuzp1q_u8(a, b) simde_vuzp1q_u8((a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_uint16x8_t simde_vuzp1q_u16(simde_uint16x8_t a, simde_uint16x8_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) return vuzp1q_u16(a, b); #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) uint16x8x2_t t = vuzpq_u16(a, b); return t.val[0]; #else simde_uint16x8_private r_, a_ = simde_uint16x8_to_private(a), b_ = simde_uint16x8_to_private(b); #if defined(SIMDE_WASM_SIMD128_NATIVE) r_.v128 = wasm_i16x8_shuffle(a_.v128, b_.v128, 0, 2, 4, 6, 8, 10, 12, 14); #elif defined(SIMDE_SHUFFLE_VECTOR_) r_.values = SIMDE_SHUFFLE_VECTOR_(16, 16, a_.values, b_.values, 0, 2, 4, 6, 8, 10, 12, 14); #else const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } #endif return simde_uint16x8_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1q_u16 #define vuzp1q_u16(a, b) simde_vuzp1q_u16((a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_uint32x4_t simde_vuzp1q_u32(simde_uint32x4_t a, simde_uint32x4_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) return vuzp1q_u32(a, b); #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) uint32x4x2_t t = vuzpq_u32(a, b); return t.val[0]; #else simde_uint32x4_private r_, a_ = simde_uint32x4_to_private(a), b_ = simde_uint32x4_to_private(b); #if defined(SIMDE_WASM_SIMD128_NATIVE) r_.v128 = wasm_i32x4_shuffle(a_.v128, b_.v128, 0, 2, 4, 6); #elif defined(SIMDE_X86_SSE2_NATIVE) r_.m128i = _mm_castps_si128(_mm_shuffle_ps(_mm_castsi128_ps(a_.m128i), _mm_castsi128_ps(b_.m128i), 0x88)); #elif defined(SIMDE_SHUFFLE_VECTOR_) r_.values = SIMDE_SHUFFLE_VECTOR_(32, 16, a_.values, b_.values, 0, 2, 4, 6); #else const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } #endif return simde_uint32x4_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1q_u32 #define vuzp1q_u32(a, b) simde_vuzp1q_u32((a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_uint64x2_t simde_vuzp1q_u64(simde_uint64x2_t a, simde_uint64x2_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) return vuzp1q_u64(a, b); #else simde_uint64x2_private r_, a_ = simde_uint64x2_to_private(a), b_ = simde_uint64x2_to_private(b); #if defined(SIMDE_WASM_SIMD128_NATIVE) r_.v128 = wasm_i64x2_shuffle(a_.v128, b_.v128, 0, 2); #elif defined(SIMDE_X86_SSE2_NATIVE) /* _mm_movelh_ps?!?! SSE is weird. */ r_.m128i = _mm_castps_si128(_mm_movelh_ps(_mm_castsi128_ps(a_.m128i), _mm_castsi128_ps(b_.m128i))); #elif defined(SIMDE_SHUFFLE_VECTOR_) r_.values = SIMDE_SHUFFLE_VECTOR_(64, 16, a_.values, b_.values, 0, 2); #else const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } #endif return simde_uint64x2_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1q_u64 #define vuzp1q_u64(a, b) simde_vuzp1q_u64((a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_poly8x8_t simde_vuzp1_p8(simde_poly8x8_t a, simde_poly8x8_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) return vuzp1_p8(a, b); #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) simde_poly8x8x2_t t = vuzp_p8(a, b); return t.val[0]; #else simde_poly8x8_private r_, a_ = simde_poly8x8_to_private(a), b_ = simde_poly8x8_to_private(b); const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } return simde_poly8x8_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1_p8 #define vuzp1_p8(a, b) simde_vuzp1_p8((a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_poly16x4_t simde_vuzp1_p16(simde_poly16x4_t a, simde_poly16x4_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) return vuzp1_p16(a, b); #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) simde_poly16x4x2_t t = vuzp_p16(a, b); return t.val[0]; #else simde_poly16x4_private r_, a_ = simde_poly16x4_to_private(a), b_ = simde_poly16x4_to_private(b); const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } return simde_poly16x4_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1_p16 #define vuzp1_p16(a, b) simde_vuzp1_p16((a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_poly8x16_t simde_vuzp1q_p8(simde_poly8x16_t a, simde_poly8x16_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) return vuzp1q_p8(a, b); #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) simde_poly8x16x2_t t = vuzpq_p8(a, b); return t.val[0]; #else simde_poly8x16_private r_, a_ = simde_poly8x16_to_private(a), b_ = simde_poly8x16_to_private(b); const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } return simde_poly8x16_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1q_p8 #define vuzp1q_p8(a, b) simde_vuzp1q_p8((a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_poly16x8_t simde_vuzp1q_p16(simde_poly16x8_t a, simde_poly16x8_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) return vuzp1q_p16(a, b); #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE) simde_poly16x8x2_t t = vuzpq_p16(a, b); return t.val[0]; #else simde_poly16x8_private r_, a_ = simde_poly16x8_to_private(a), b_ = simde_poly16x8_to_private(b); const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } return simde_poly16x8_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1q_p16 #define vuzp1q_p16(a, b) simde_vuzp1q_p16((a), (b)) #endif SIMDE_FUNCTION_ATTRIBUTES simde_poly64x2_t simde_vuzp1q_p64(simde_poly64x2_t a, simde_poly64x2_t b) { #if defined(SIMDE_ARM_NEON_A64V8_NATIVE) return vuzp1q_p64(a, b); #else simde_poly64x2_private r_, a_ = simde_poly64x2_to_private(a), b_ = simde_poly64x2_to_private(b); const size_t halfway_point = sizeof(r_.values) / sizeof(r_.values[0]) / 2; SIMDE_VECTORIZE for (size_t i = 0 ; i < halfway_point ; i++) { const size_t idx = i << 1; r_.values[ i ] = a_.values[idx]; r_.values[i + halfway_point] = b_.values[idx]; } return simde_poly64x2_from_private(r_); #endif } #if defined(SIMDE_ARM_NEON_A64V8_ENABLE_NATIVE_ALIASES) #undef vuzp1q_p64 #define vuzp1q_p64(a, b) simde_vuzp1q_p64((a), (b)) #endif SIMDE_END_DECLS_ HEDLEY_DIAGNOSTIC_POP #endif /* !defined(SIMDE_ARM_NEON_UZP1_H) */