Quelle  mmx.h   Sprache: C

/* 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:
 *   2017-2020 Evan Nemerson <evan@nemerson.com>
 */

#if !defined(SIMDE_X86_MMX_H)
#define SIMDE_X86_MMX_H

#include "../simde-common.h"

HEDLEY_DIAGNOSTIC_PUSH
SIMDE_DISABLE_UNWANTED_DIAGNOSTICS

#if defined(SIMDE_X86_MMX_NATIVE)
  #define SIMDE_X86_MMX_USE_NATIVE_TYPE
#elif defined(SIMDE_X86_SSE_NATIVE)
  #define SIMDE_X86_MMX_USE_NATIVE_TYPE
#endif

#if defined(SIMDE_X86_MMX_USE_NATIVE_TYPE)
  #include <mmintrin.h>
#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)
  #include <arm_neon.h>
#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
  #include <loongson-mmiintrin.h>
#endif

#include <stdint.h>
#include <limits.h>

SIMDE_BEGIN_DECLS_

typedef union {
  #if defined(SIMDE_VECTOR_SUBSCRIPT)
    SIMDE_ALIGN_TO_8 int8_t          i8 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;
    SIMDE_ALIGN_TO_8 int16_t        i16 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;
    SIMDE_ALIGN_TO_8 int32_t        i32 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;
    SIMDE_ALIGN_TO_8 int64_t        i64 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;
    SIMDE_ALIGN_TO_8 uint8_t         u8 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;
    SIMDE_ALIGN_TO_8 uint16_t       u16 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;
    SIMDE_ALIGN_TO_8 uint32_t       u32 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;
    SIMDE_ALIGN_TO_8 uint64_t       u64 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;
    SIMDE_ALIGN_TO_8 simde_float32  f32 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;
    SIMDE_ALIGN_TO_8 int_fast32_t  i32f SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;
    SIMDE_ALIGN_TO_8 uint_fast32_t u32f SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;
  #else
    SIMDE_ALIGN_TO_8 int8_t          i8[8];
    SIMDE_ALIGN_TO_8 int16_t        i16[4];
    SIMDE_ALIGN_TO_8 int32_t        i32[2];
    SIMDE_ALIGN_TO_8 int64_t        i64[1];
    SIMDE_ALIGN_TO_8 uint8_t         u8[8];
    SIMDE_ALIGN_TO_8 uint16_t       u16[4];
    SIMDE_ALIGN_TO_8 uint32_t       u32[2];
    SIMDE_ALIGN_TO_8 uint64_t       u64[1];
    SIMDE_ALIGN_TO_8 simde_float32  f32[2];
    SIMDE_ALIGN_TO_8 int_fast32_t  i32f[8 / sizeof(int_fast32_t)];
    SIMDE_ALIGN_TO_8 uint_fast32_t u32f[8 / sizeof(uint_fast32_t)];
  #endif

  #if defined(SIMDE_X86_MMX_USE_NATIVE_TYPE)
    __m64          n;
  #endif
  #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
    int8x8_t       neon_i8;
    int16x4_t      neon_i16;
    int32x2_t      neon_i32;
    int64x1_t      neon_i64;
    uint8x8_t      neon_u8;
    uint16x4_t     neon_u16;
    uint32x2_t     neon_u32;
    uint64x1_t     neon_u64;
    float32x2_t    neon_f32;
  #endif
  #if defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
    int8x8_t       mmi_i8;
    int16x4_t      mmi_i16;
    int32x2_t      mmi_i32;
    int64_t        mmi_i64;
    uint8x8_t      mmi_u8;
    uint16x4_t     mmi_u16;
    uint32x2_t     mmi_u32;
    uint64_t       mmi_u64;
  #endif
} simde__m64_private;

#if defined(SIMDE_X86_MMX_USE_NATIVE_TYPE)
  typedef __m64 simde__m64;
#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)
  typedef int32x2_t simde__m64;
#elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
  typedef int32x2_t simde__m64;
#elif defined(SIMDE_VECTOR_SUBSCRIPT)
  typedef int32_t simde__m64 SIMDE_ALIGN_TO_8 SIMDE_VECTOR(8) SIMDE_MAY_ALIAS;
#else
  typedef simde__m64_private simde__m64;
#endif

#if !defined(SIMDE_X86_MMX_USE_NATIVE_TYPE) && defined(SIMDE_ENABLE_NATIVE_ALIASES)
  #define SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES
  typedef simde__m64 __m64;
#endif

HEDLEY_STATIC_ASSERT(8 == sizeof(simde__m64), "__m64 size incorrect");
HEDLEY_STATIC_ASSERT(8 == sizeof(simde__m64_private), "__m64 size incorrect");
#if defined(SIMDE_CHECK_ALIGNMENT) && defined(SIMDE_ALIGN_OF)
HEDLEY_STATIC_ASSERT(SIMDE_ALIGN_OF(simde__m64) == 8, "simde__m64 is not 8-byte aligned");
HEDLEY_STATIC_ASSERT(SIMDE_ALIGN_OF(simde__m64_private) == 8, "simde__m64_private is not 8-byte aligned");
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde__m64_from_private(simde__m64_private v) {
  simde__m64 r;
  simde_memcpy(&r, &v, sizeof(r));
  return r;
}

SIMDE_FUNCTION_ATTRIBUTES
simde__m64_private
simde__m64_to_private(simde__m64 v) {
  simde__m64_private r;
  simde_memcpy(&r, &v, sizeof(r));
  return r;
}

#define SIMDE_X86_GENERATE_CONVERSION_FUNCTION(simde_type, source_type, isax, fragment) \
  SIMDE_FUNCTION_ATTRIBUTES \
  simde__##simde_type \
  simde__##simde_type##_from_##isax##_##fragment(source_type value) { \
    simde__##simde_type##_private r_; \
    r_.isax##_##fragment = value; \
    return simde__##simde_type##_from_private(r_); \
  } \
  \
  SIMDE_FUNCTION_ATTRIBUTES \
  source_type \
  simde__##simde_type##_to_##isax##_##fragment(simde__##simde_type value) { \
    simde__##simde_type##_private r_ = simde__##simde_type##_to_private(value); \
    return r_.isax##_##fragment; \
  }

#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
  SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, int8x8_t, neon, i8)
  SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, int16x4_t, neon, i16)
  SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, int32x2_t, neon, i32)
  SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, int64x1_t, neon, i64)
  SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, uint8x8_t, neon, u8)
  SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, uint16x4_t, neon, u16)
  SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, uint32x2_t, neon, u32)
  SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, uint64x1_t, neon, u64)
  SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, float32x2_t, neon, f32)
#endif /* defined(SIMDE_ARM_NEON_A32V7_NATIVE) */

#if defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
  SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, int8x8_t, mmi, i8)
  SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, int16x4_t, mmi, i16)
  SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, int32x2_t, mmi, i32)
  SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, int64_t, mmi, i64)
  SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, uint8x8_t, mmi, u8)
  SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, uint16x4_t, mmi, u16)
  SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, uint32x2_t, mmi, u32)
  SIMDE_X86_GENERATE_CONVERSION_FUNCTION(m64, uint64_t, mmi, u64)
#endif /* defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE) */

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_add_pi8 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_add_pi8(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_i8 = vadd_s8(a_.neon_i8, b_.neon_i8);
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i8 = paddb_s(a_.mmi_i8, b_.mmi_i8);
    #elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)
      r_.i8 = a_.i8 + b_.i8;
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i8) / sizeof(r_.i8[0])) ; i++) {
        r_.i8[i] = a_.i8[i] + b_.i8[i];
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_paddb(a, b) simde_mm_add_pi8(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_add_pi8(a, b) simde_mm_add_pi8(a, b)
#  define _m_paddb(a, b) simde_m_paddb(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_add_pi16 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_add_pi16(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

  #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
    r_.neon_i16 = vadd_s16(a_.neon_i16, b_.neon_i16);
  #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
    r_.mmi_i16 = paddh_s(a_.mmi_i16, b_.mmi_i16);
  #elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)
    r_.i16 = a_.i16 + b_.i16;
  #else
    SIMDE_VECTORIZE
    for (size_t i = 0 ; i < (sizeof(r_.i16) / sizeof(r_.i16[0])) ; i++) {
      r_.i16[i] = a_.i16[i] + b_.i16[i];
    }
  #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_paddw(a, b) simde_mm_add_pi16(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_add_pi16(a, b) simde_mm_add_pi16(a, b)
#  define _m_paddw(a, b) simde_mm_add_pi16(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_add_pi32 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_add_pi32(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_i32 = vadd_s32(a_.neon_i32, b_.neon_i32);
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i32 = paddw_s(a_.mmi_i32, b_.mmi_i32);
    #elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)
      r_.i32 = a_.i32 + b_.i32;
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i32) / sizeof(r_.i32[0])) ; i++) {
        r_.i32[i] = a_.i32[i] + b_.i32[i];
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_paddd(a, b) simde_mm_add_pi32(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_add_pi32(a, b) simde_mm_add_pi32(a, b)
#  define _m_paddd(a, b) simde_mm_add_pi32(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_adds_pi8 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_adds_pi8(a, b);
  #else
    simde__m64_private
      r_,
      a_ = simde__m64_to_private(a),
      b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_i8 = vqadd_s8(a_.neon_i8, b_.neon_i8);
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i8 = paddsb(a_.mmi_i8, b_.mmi_i8);
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i8) / sizeof(r_.i8[0])) ; i++) {
        if ((((b_.i8[i]) > 0) && ((a_.i8[i]) > (INT8_MAX - (b_.i8[i]))))) {
          r_.i8[i] = INT8_MAX;
        } else if ((((b_.i8[i]) < 0) && ((a_.i8[i]) < (INT8_MIN - (b_.i8[i]))))) {
          r_.i8[i] = INT8_MIN;
        } else {
          r_.i8[i] = (a_.i8[i]) + (b_.i8[i]);
        }
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_paddsb(a, b) simde_mm_adds_pi8(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_adds_pi8(a, b) simde_mm_adds_pi8(a, b)
#  define _m_paddsb(a, b) simde_mm_adds_pi8(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_adds_pu8 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_adds_pu8(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_u8 = vqadd_u8(a_.neon_u8, b_.neon_u8);
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_u8 = paddusb(a_.mmi_u8, b_.mmi_u8);
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.u8) / sizeof(r_.u8[0])) ; i++) {
        const uint_fast16_t x = HEDLEY_STATIC_CAST(uint_fast16_t, a_.u8[i]) + HEDLEY_STATIC_CAST(uint_fast16_t, b_.u8[i]);
        if (x > UINT8_MAX)
          r_.u8[i] = UINT8_MAX;
        else
          r_.u8[i] = HEDLEY_STATIC_CAST(uint8_t, x);
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_paddusb(a, b) simde_mm_adds_pu8(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_adds_pu8(a, b) simde_mm_adds_pu8(a, b)
#  define _m_paddusb(a, b) simde_mm_adds_pu8(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_adds_pi16 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_adds_pi16(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_i16 = vqadd_s16(a_.neon_i16, b_.neon_i16);
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i16 = paddsh(a_.mmi_i16, b_.mmi_i16);
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i16) / sizeof(r_.i16[0])) ; i++) {
        if ((((b_.i16[i]) > 0) && ((a_.i16[i]) > (INT16_MAX - (b_.i16[i]))))) {
          r_.i16[i] = INT16_MAX;
        } else if ((((b_.i16[i]) < 0) && ((a_.i16[i]) < (SHRT_MIN - (b_.i16[i]))))) {
          r_.i16[i] = SHRT_MIN;
        } else {
          r_.i16[i] = (a_.i16[i]) + (b_.i16[i]);
        }
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_paddsw(a, b) simde_mm_adds_pi16(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_adds_pi16(a, b) simde_mm_adds_pi16(a, b)
#  define _m_paddsw(a, b) simde_mm_adds_pi16(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_adds_pu16 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_adds_pu16(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_u16 = vqadd_u16(a_.neon_u16, b_.neon_u16);
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_u16 = paddush(a_.mmi_u16, b_.mmi_u16);
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i16) / sizeof(r_.i16[0])) ; i++) {
        const uint32_t x = a_.u16[i] + b_.u16[i];
        if (x > UINT16_MAX)
          r_.u16[i] = UINT16_MAX;
        else
          r_.u16[i] = HEDLEY_STATIC_CAST(uint16_t, x);
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_paddusw(a, b) simde_mm_adds_pu16(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_adds_pu16(a, b) simde_mm_adds_pu16(a, b)
#  define _m_paddusw(a, b) simde_mm_adds_pu16(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_and_si64 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_and_si64(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_i32 = vand_s32(a_.neon_i32, b_.neon_i32);
    #elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)
      r_.i64 = a_.i64 & b_.i64;
    #else
      r_.i64[0] = a_.i64[0] & b_.i64[0];
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_pand(a, b) simde_mm_and_si64(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_and_si64(a, b) simde_mm_and_si64(a, b)
#  define _m_pand(a, b) simde_mm_and_si64(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_andnot_si64 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_andnot_si64(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_i32 = vbic_s32(b_.neon_i32, a_.neon_i32);
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i32 = pandn_sw(a_.mmi_i32, b_.mmi_i32);
    #elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)
      r_.i32f = ~a_.i32f & b_.i32f;
    #else
      r_.u64[0] = (~(a_.u64[0])) & (b_.u64[0]);
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_pandn(a, b) simde_mm_andnot_si64(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_andnot_si64(a, b) simde_mm_andnot_si64(a, b)
#  define _m_pandn(a, b) simde_mm_andnot_si64(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_cmpeq_pi8 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_cmpeq_pi8(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_u8 = vceq_s8(a_.neon_i8, b_.neon_i8);
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i8 = pcmpeqb_s(a_.mmi_i8, b_.mmi_i8);
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i8) / sizeof(r_.i8[0])) ; i++) {
        r_.i8[i] = (a_.i8[i] == b_.i8[i]) ? ~INT8_C(0) : INT8_C(0);
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_pcmpeqb(a, b) simde_mm_cmpeq_pi8(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_cmpeq_pi8(a, b) simde_mm_cmpeq_pi8(a, b)
#  define _m_pcmpeqb(a, b) simde_mm_cmpeq_pi8(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_cmpeq_pi16 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_cmpeq_pi16(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_u16 = vceq_s16(a_.neon_i16, b_.neon_i16);
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i16 = pcmpeqh_s(a_.mmi_i16, b_.mmi_i16);
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i16) / sizeof(r_.i16[0])) ; i++) {
        r_.i16[i] = (a_.i16[i] == b_.i16[i]) ? ~INT16_C(0) : INT16_C(0);
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_pcmpeqw(a, b) simde_mm_cmpeq_pi16(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_cmpeq_pi16(a, b) simde_mm_cmpeq_pi16(a, b)
#  define _m_pcmpeqw(a, b) simde_mm_cmpeq_pi16(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_cmpeq_pi32 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_cmpeq_pi32(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_u32 = vceq_s32(a_.neon_i32, b_.neon_i32);
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i32 = pcmpeqw_s(a_.mmi_i32, b_.mmi_i32);
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i32) / sizeof(r_.i32[0])) ; i++) {
        r_.i32[i] = (a_.i32[i] == b_.i32[i]) ? ~INT32_C(0) : INT32_C(0);
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_pcmpeqd(a, b) simde_mm_cmpeq_pi32(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_cmpeq_pi32(a, b) simde_mm_cmpeq_pi32(a, b)
#  define _m_pcmpeqd(a, b) simde_mm_cmpeq_pi32(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_cmpgt_pi8 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_cmpgt_pi8(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_u8 = vcgt_s8(a_.neon_i8, b_.neon_i8);
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i8 = pcmpgtb_s(a_.mmi_i8, b_.mmi_i8);
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i8) / sizeof(r_.i8[0])) ; i++) {
        r_.i8[i] = (a_.i8[i] > b_.i8[i]) ? ~INT8_C(0) : INT8_C(0);
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_pcmpgtb(a, b) simde_mm_cmpgt_pi8(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_cmpgt_pi8(a, b) simde_mm_cmpgt_pi8(a, b)
#  define _m_pcmpgtb(a, b) simde_mm_cmpgt_pi8(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_cmpgt_pi16 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_cmpgt_pi16(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_u16 = vcgt_s16(a_.neon_i16, b_.neon_i16);
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i16 = pcmpgth_s(a_.mmi_i16, b_.mmi_i16);
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i16) / sizeof(r_.i16[0])) ; i++) {
        r_.i16[i] = (a_.i16[i] > b_.i16[i]) ? ~INT16_C(0) : INT16_C(0);
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_pcmpgtw(a, b) simde_mm_cmpgt_pi16(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_cmpgt_pi16(a, b) simde_mm_cmpgt_pi16(a, b)
#  define _m_pcmpgtw(a, b) simde_mm_cmpgt_pi16(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_cmpgt_pi32 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_cmpgt_pi32(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_u32 = vcgt_s32(a_.neon_i32, b_.neon_i32);
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i32 = pcmpgtw_s(a_.mmi_i32, b_.mmi_i32);
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i32) / sizeof(r_.i32[0])) ; i++) {
        r_.i32[i] = (a_.i32[i] > b_.i32[i]) ? ~INT32_C(0) : INT32_C(0);
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_pcmpgtd(a, b) simde_mm_cmpgt_pi32(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_cmpgt_pi32(a, b) simde_mm_cmpgt_pi32(a, b)
#  define _m_pcmpgtd(a, b) simde_mm_cmpgt_pi32(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
int64_t
simde_mm_cvtm64_si64 (simde__m64 a) {
  #if defined(SIMDE_X86_MMX_NATIVE) && defined(SIMDE_ARCH_AMD64) && !defined(__PGI)
    return _mm_cvtm64_si64(a);
  #else
    simde__m64_private a_ = simde__m64_to_private(a);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      HEDLEY_DIAGNOSTIC_PUSH
      #if HEDLEY_HAS_WARNING("-Wvector-conversion") && SIMDE_DETECT_CLANG_VERSION_NOT(10,0,0)
        #pragma clang diagnostic ignored "-Wvector-conversion"
      #endif
      return vget_lane_s64(a_.neon_i64, 0);
      HEDLEY_DIAGNOSTIC_POP
    #else
      return a_.i64[0];
    #endif
  #endif
}
#define simde_m_to_int64(a) simde_mm_cvtm64_si64(a)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES) || (defined(SIMDE_ENABLE_NATIVE_ALIASES) && !defined(SIMDE_ARCH_AMD64))
#  define _mm_cvtm64_si64(a) simde_mm_cvtm64_si64(a)
#  define _m_to_int64(a) simde_mm_cvtm64_si64(a)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_cvtsi32_si64 (int32_t a) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_cvtsi32_si64(a);
  #else
    simde__m64_private r_;

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      const int32_t av[2] = { a, 0 };
      r_.neon_i32 = vld1_s32(av);
    #else
      r_.i32[0] = a;
      r_.i32[1] = 0;
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_from_int(a) simde_mm_cvtsi32_si64(a)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_cvtsi32_si64(a) simde_mm_cvtsi32_si64(a)
#  define _m_from_int(a) simde_mm_cvtsi32_si64(a)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_cvtsi64_m64 (int64_t a) {
  #if defined(SIMDE_X86_MMX_NATIVE) && defined(SIMDE_ARCH_AMD64) && !defined(__PGI)
    return _mm_cvtsi64_m64(a);
  #else
    simde__m64_private r_;

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_i64 = vld1_s64(&a);
    #else
      r_.i64[0] = a;
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_from_int64(a) simde_mm_cvtsi64_m64(a)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES) || (defined(SIMDE_ENABLE_NATIVE_ALIASES) && !defined(SIMDE_ARCH_AMD64))
#  define _mm_cvtsi64_m64(a) simde_mm_cvtsi64_m64(a)
#  define _m_from_int64(a) simde_mm_cvtsi64_m64(a)
#endif

SIMDE_FUNCTION_ATTRIBUTES
int32_t
simde_mm_cvtsi64_si32 (simde__m64 a) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_cvtsi64_si32(a);
  #else
    simde__m64_private a_ = simde__m64_to_private(a);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      HEDLEY_DIAGNOSTIC_PUSH
      #if HEDLEY_HAS_WARNING("-Wvector-conversion") && SIMDE_DETECT_CLANG_VERSION_NOT(10,0,0)
        #pragma clang diagnostic ignored "-Wvector-conversion"
      #endif
      return vget_lane_s32(a_.neon_i32, 0);
      HEDLEY_DIAGNOSTIC_POP
    #else
      return a_.i32[0];
    #endif
  #endif
}
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_cvtsi64_si32(a) simde_mm_cvtsi64_si32(a)
#endif

SIMDE_FUNCTION_ATTRIBUTES
void
simde_mm_empty (void) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    _mm_empty();
  #else
    /* noop */
  #endif
}
#define simde_m_empty() simde_mm_empty()
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_empty() simde_mm_empty()
#  define _m_empty() simde_mm_empty()
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_madd_pi16 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_madd_pi16(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      int32x4_t i1 = vmull_s16(a_.neon_i16, b_.neon_i16);
      r_.neon_i32 = vpadd_s32(vget_low_s32(i1), vget_high_s32(i1));
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i32 = pmaddhw(a_.mmi_i16, b_.mmi_i16);
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i16) / sizeof(r_.i16[0])) ; i += 2) {
        r_.i32[i / 2] = (a_.i16[i] * b_.i16[i]) + (a_.i16[i + 1] * b_.i16[i + 1]);
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_pmaddwd(a, b) simde_mm_madd_pi16(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_madd_pi16(a, b) simde_mm_madd_pi16(a, b)
#  define _m_pmaddwd(a, b) simde_mm_madd_pi16(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_mulhi_pi16 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_mulhi_pi16(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      const int32x4_t t1 = vmull_s16(a_.neon_i16, b_.neon_i16);
      const uint32x4_t t2 = vshrq_n_u32(vreinterpretq_u32_s32(t1), 16);
      const uint16x4_t t3 = vmovn_u32(t2);
      r_.neon_u16 = t3;
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i16 = pmulhh(a_.mmi_i16, b_.mmi_i16);
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i16) / sizeof(r_.i16[0])) ; i++) {
        r_.i16[i] = HEDLEY_STATIC_CAST(int16_t, ((a_.i16[i] * b_.i16[i]) >> 16));
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_pmulhw(a, b) simde_mm_mulhi_pi16(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_mulhi_pi16(a, b) simde_mm_mulhi_pi16(a, b)
#  define _m_pmulhw(a, b) simde_mm_mulhi_pi16(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_mullo_pi16 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_mullo_pi16(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      const int32x4_t t1 = vmull_s16(a_.neon_i16, b_.neon_i16);
      const uint16x4_t t2 = vmovn_u32(vreinterpretq_u32_s32(t1));
      r_.neon_u16 = t2;
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i16 = pmullh(a_.mmi_i16, b_.mmi_i16);
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i16) / sizeof(r_.i16[0])) ; i++) {
        r_.i16[i] = HEDLEY_STATIC_CAST(int16_t, ((a_.i16[i] * b_.i16[i]) & 0xffff));
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_pmullw(a, b) simde_mm_mullo_pi16(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_mullo_pi16(a, b) simde_mm_mullo_pi16(a, b)
#  define _m_pmullw(a, b) simde_mm_mullo_pi16(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_or_si64 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_or_si64(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_i32 = vorr_s32(a_.neon_i32, b_.neon_i32);
    #elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)
      r_.i64 = a_.i64 | b_.i64;
    #else
      r_.i64[0] = a_.i64[0] | b_.i64[0];
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_por(a, b) simde_mm_or_si64(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_or_si64(a, b) simde_mm_or_si64(a, b)
#  define _m_por(a, b) simde_mm_or_si64(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_packs_pi16 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_packs_pi16(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_i8 = vqmovn_s16(vcombine_s16(a_.neon_i16, b_.neon_i16));
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i8 = packsshb(a_.mmi_i16, b_.mmi_i16);
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i16) / sizeof(r_.i16[0])) ; i++) {
        if (a_.i16[i] < INT8_MIN) {
          r_.i8[i] = INT8_MIN;
        } else if (a_.i16[i] > INT8_MAX) {
          r_.i8[i] = INT8_MAX;
        } else {
          r_.i8[i] = HEDLEY_STATIC_CAST(int8_t, a_.i16[i]);
        }
      }

      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i16) / sizeof(r_.i16[0])) ; i++) {
        if (b_.i16[i] < INT8_MIN) {
          r_.i8[i + 4] = INT8_MIN;
        } else if (b_.i16[i] > INT8_MAX) {
          r_.i8[i + 4] = INT8_MAX;
        } else {
          r_.i8[i + 4] = HEDLEY_STATIC_CAST(int8_t, b_.i16[i]);
        }
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_packsswb(a, b) simde_mm_packs_pi16(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_packs_pi16(a, b) simde_mm_packs_pi16(a, b)
#  define _m_packsswb(a, b) simde_mm_packs_pi16(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_packs_pi32 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_packs_pi32(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_i16 = vqmovn_s32(vcombine_s32(a_.neon_i32, b_.neon_i32));
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i16 = packsswh(a_.mmi_i32, b_.mmi_i32);
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (8 / sizeof(a_.i32[0])) ; i++) {
        if (a_.i32[i] < SHRT_MIN) {
          r_.i16[i] = SHRT_MIN;
        } else if (a_.i32[i] > INT16_MAX) {
          r_.i16[i] = INT16_MAX;
        } else {
          r_.i16[i] = HEDLEY_STATIC_CAST(int16_t, a_.i32[i]);
        }
      }

      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (8 / sizeof(b_.i32[0])) ; i++) {
        if (b_.i32[i] < SHRT_MIN) {
          r_.i16[i + 2] = SHRT_MIN;
        } else if (b_.i32[i] > INT16_MAX) {
          r_.i16[i + 2] = INT16_MAX;
        } else {
          r_.i16[i + 2] = HEDLEY_STATIC_CAST(int16_t, b_.i32[i]);
        }
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_packssdw(a, b) simde_mm_packs_pi32(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_packs_pi32(a, b) simde_mm_packs_pi32(a, b)
#  define _m_packssdw(a, b) simde_mm_packs_pi32(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_packs_pu16 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_packs_pu16(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A64V8_NATIVE)
      const int16x8_t t1 = vcombine_s16(a_.neon_i16, b_.neon_i16);

      /* Set elements which are < 0 to 0 */
      const int16x8_t t2 = vandq_s16(t1, vreinterpretq_s16_u16(vcgezq_s16(t1)));

      /* Vector with all s16 elements set to UINT8_MAX */
      const int16x8_t vmax = vmovq_n_s16(HEDLEY_STATIC_CAST(int16_t, UINT8_MAX));

      /* Elements which are within the acceptable range */
      const int16x8_t le_max = vandq_s16(t2, vreinterpretq_s16_u16(vcleq_s16(t2, vmax)));
      const int16x8_t gt_max = vandq_s16(vmax, vreinterpretq_s16_u16(vcgtq_s16(t2, vmax)));

      /* Final values as 16-bit integers */
      const int16x8_t values = vorrq_s16(le_max, gt_max);

      r_.neon_u8 = vmovn_u16(vreinterpretq_u16_s16(values));
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_u8 = packushb(a_.mmi_u16, b_.mmi_u16);
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i16) / sizeof(r_.i16[0])) ; i++) {
        if (a_.i16[i] > UINT8_MAX) {
          r_.u8[i] = UINT8_MAX;
        } else if (a_.i16[i] < 0) {
          r_.u8[i] = 0;
        } else {
          r_.u8[i] = HEDLEY_STATIC_CAST(uint8_t, a_.i16[i]);
        }
      }

      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i16) / sizeof(r_.i16[0])) ; i++) {
        if (b_.i16[i] > UINT8_MAX) {
          r_.u8[i + 4] = UINT8_MAX;
        } else if (b_.i16[i] < 0) {
          r_.u8[i + 4] = 0;
        } else {
          r_.u8[i + 4] = HEDLEY_STATIC_CAST(uint8_t, b_.i16[i]);
        }
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_packuswb(a, b) simde_mm_packs_pu16(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_packs_pu16(a, b) simde_mm_packs_pu16(a, b)
#  define _m_packuswb(a, b) simde_mm_packs_pu16(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_set_pi8 (int8_t e7, int8_t e6, int8_t e5, int8_t e4, int8_t e3, int8_t e2, int8_t e1, int8_t e0) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_set_pi8(e7, e6, e5, e4, e3, e2, e1, e0);
  #else
    simde__m64_private r_;

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      const int8_t v[sizeof(r_.i8) / sizeof(r_.i8[0])] = { e0, e1, e2, e3, e4, e5, e6, e7 };
      r_.neon_i8 = vld1_s8(v);
    #else
      r_.i8[0] = e0;
      r_.i8[1] = e1;
      r_.i8[2] = e2;
      r_.i8[3] = e3;
      r_.i8[4] = e4;
      r_.i8[5] = e5;
      r_.i8[6] = e6;
      r_.i8[7] = e7;
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_set_pi8(e7, e6, e5, e4, e3, e2, e1, e0) simde_mm_set_pi8(e7, e6, e5, e4, e3, e2, e1, e0)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_x_mm_set_pu8 (uint8_t e7, uint8_t e6, uint8_t e5, uint8_t e4, uint8_t e3, uint8_t e2, uint8_t e1, uint8_t e0) {
  simde__m64_private r_;

  #if defined(SIMDE_X86_MMX_NATIVE)
    r_.n = _mm_set_pi8(
        HEDLEY_STATIC_CAST(int8_t, e7),
        HEDLEY_STATIC_CAST(int8_t, e6),
        HEDLEY_STATIC_CAST(int8_t, e5),
        HEDLEY_STATIC_CAST(int8_t, e4),
        HEDLEY_STATIC_CAST(int8_t, e3),
        HEDLEY_STATIC_CAST(int8_t, e2),
        HEDLEY_STATIC_CAST(int8_t, e1),
        HEDLEY_STATIC_CAST(int8_t, e0));
  #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)
    const uint8_t v[sizeof(r_.u8) / sizeof(r_.u8[0])] = { e0, e1, e2, e3, e4, e5, e6, e7 };
    r_.neon_u8 = vld1_u8(v);
  #else
    r_.u8[0] = e0;
    r_.u8[1] = e1;
    r_.u8[2] = e2;
    r_.u8[3] = e3;
    r_.u8[4] = e4;
    r_.u8[5] = e5;
    r_.u8[6] = e6;
    r_.u8[7] = e7;
  #endif

  return simde__m64_from_private(r_);
}

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_set_pi16 (int16_t e3, int16_t e2, int16_t e1, int16_t e0) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_set_pi16(e3, e2, e1, e0);
  #else
    simde__m64_private r_;

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      const int16_t v[sizeof(r_.i16) / sizeof(r_.i16[0])] = { e0, e1, e2, e3 };
      r_.neon_i16 = vld1_s16(v);
    #else
      r_.i16[0] = e0;
      r_.i16[1] = e1;
      r_.i16[2] = e2;
      r_.i16[3] = e3;
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_set_pi16(e3, e2, e1, e0) simde_mm_set_pi16(e3, e2, e1, e0)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_x_mm_set_pu16 (uint16_t e3, uint16_t e2, uint16_t e1, uint16_t e0) {
  simde__m64_private r_;

#if defined(SIMDE_X86_MMX_NATIVE)
  r_.n = _mm_set_pi16(
      HEDLEY_STATIC_CAST(int16_t, e3),
      HEDLEY_STATIC_CAST(int16_t, e2),
      HEDLEY_STATIC_CAST(int16_t, e1),
      HEDLEY_STATIC_CAST(int16_t, e0)
    );
#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)
  const uint16_t v[sizeof(r_.u16) / sizeof(r_.u16[0])] = { e0, e1, e2, e3 };
  r_.neon_u16 = vld1_u16(v);
#else
  r_.u16[0] = e0;
  r_.u16[1] = e1;
  r_.u16[2] = e2;
  r_.u16[3] = e3;
#endif

  return simde__m64_from_private(r_);
}

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_x_mm_set_pu32 (uint32_t e1, uint32_t e0) {
  simde__m64_private r_;

#if defined(SIMDE_X86_MMX_NATIVE)
  r_.n = _mm_set_pi32(
      HEDLEY_STATIC_CAST(int32_t, e1),
      HEDLEY_STATIC_CAST(int32_t, e0));
#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)
  const uint32_t v[sizeof(r_.u32) / sizeof(r_.u32[0])] = { e0, e1 };
  r_.neon_u32 = vld1_u32(v);
#else
  r_.u32[0] = e0;
  r_.u32[1] = e1;
#endif

  return simde__m64_from_private(r_);
}

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_set_pi32 (int32_t e1, int32_t e0) {
  simde__m64_private r_;

#if defined(SIMDE_X86_MMX_NATIVE)
  r_.n = _mm_set_pi32(e1, e0);
#elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)
  const int32_t v[sizeof(r_.i32) / sizeof(r_.i32[0])] = { e0, e1 };
  r_.neon_i32 = vld1_s32(v);
#else
  r_.i32[0] = e0;
  r_.i32[1] = e1;
#endif

  return simde__m64_from_private(r_);
}
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_set_pi32(e1, e0) simde_mm_set_pi32(e1, e0)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_x_mm_set_pi64 (int64_t e0) {
  simde__m64_private r_;

#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
  const int64_t v[sizeof(r_.i64) / sizeof(r_.i64[0])] = { e0 };
  r_.neon_i64 = vld1_s64(v);
#else
  r_.i64[0] = e0;
#endif

  return simde__m64_from_private(r_);
}


SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_x_mm_set_f32x2 (simde_float32 e1, simde_float32 e0) {
  simde__m64_private r_;

#if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
  const simde_float32 v[sizeof(r_.f32) / sizeof(r_.f32[0])] = { e0, e1 };
  r_.neon_f32 = vld1_f32(v);
#else
  r_.f32[0] = e0;
  r_.f32[1] = e1;
#endif

  return simde__m64_from_private(r_);
}

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_set1_pi8 (int8_t a) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_set1_pi8(a);
  #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)
    simde__m64_private r_;
    r_.neon_i8 = vmov_n_s8(a);
    return simde__m64_from_private(r_);
  #else
    return simde_mm_set_pi8(a, a, a, a, a, a, a, a);
  #endif
}
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_set1_pi8(a) simde_mm_set1_pi8(a)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_set1_pi16 (int16_t a) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_set1_pi16(a);
  #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)
    simde__m64_private r_;
    r_.neon_i16 = vmov_n_s16(a);
    return simde__m64_from_private(r_);
  #else
    return simde_mm_set_pi16(a, a, a, a);
  #endif
}
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_set1_pi16(a) simde_mm_set1_pi16(a)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_set1_pi32 (int32_t a) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_set1_pi32(a);
  #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)
    simde__m64_private r_;
    r_.neon_i32 = vmov_n_s32(a);
    return simde__m64_from_private(r_);
  #else
    return simde_mm_set_pi32(a, a);
  #endif
}
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_set1_pi32(a) simde_mm_set1_pi32(a)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_setr_pi8 (int8_t e7, int8_t e6, int8_t e5, int8_t e4, int8_t e3, int8_t e2, int8_t e1, int8_t e0) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_setr_pi8(e7, e6, e5, e4, e3, e2, e1, e0);
  #else
    return simde_mm_set_pi8(e0, e1, e2, e3, e4, e5, e6, e7);
  #endif
}
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_setr_pi8(e7, e6, e5, e4, e3, e2, e1, e0) simde_mm_setr_pi8(e7, e6, e5, e4, e3, e2, e1, e0)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_setr_pi16 (int16_t e3, int16_t e2, int16_t e1, int16_t e0) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_setr_pi16(e3, e2, e1, e0);
  #else
    return simde_mm_set_pi16(e0, e1, e2, e3);
  #endif
}
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_setr_pi16(e3, e2, e1, e0) simde_mm_setr_pi16(e3, e2, e1, e0)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_setr_pi32 (int32_t e1, int32_t e0) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_setr_pi32(e1, e0);
  #else
    return simde_mm_set_pi32(e0, e1);
  #endif
}
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_setr_pi32(e1, e0) simde_mm_setr_pi32(e1, e0)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_setzero_si64 (void) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_setzero_si64();
  #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)
    simde__m64_private r_;
    r_.neon_u32 = vmov_n_u32(0);
    return simde__m64_from_private(r_);
  #else
    return simde_mm_set_pi32(0, 0);
  #endif
}
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_setzero_si64() simde_mm_setzero_si64()
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_x_mm_load_si64 (const void* mem_addr) {
  simde__m64 r;
  simde_memcpy(&r, SIMDE_ALIGN_ASSUME_LIKE(mem_addr, simde__m64), sizeof(r));
  return r;
}

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_x_mm_loadu_si64 (const void* mem_addr) {
  simde__m64 r;
  simde_memcpy(&r, mem_addr, sizeof(r));
  return r;
}

SIMDE_FUNCTION_ATTRIBUTES
void
simde_x_mm_store_si64 (void* mem_addr, simde__m64 value) {
  simde_memcpy(SIMDE_ALIGN_ASSUME_LIKE(mem_addr, simde__m64), &value, sizeof(value));
}

SIMDE_FUNCTION_ATTRIBUTES
void
simde_x_mm_storeu_si64 (void* mem_addr, simde__m64 value) {
  simde_memcpy(mem_addr, &value, sizeof(value));
}

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_x_mm_setone_si64 (void) {
  return simde_mm_set1_pi32(~INT32_C(0));
}

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_sll_pi16 (simde__m64 a, simde__m64 count) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_sll_pi16(a, count);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private count_ = simde__m64_to_private(count);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      HEDLEY_DIAGNOSTIC_PUSH
      #if HEDLEY_HAS_WARNING("-Wvector-conversion") && SIMDE_DETECT_CLANG_VERSION_NOT(10,0,0)
        #pragma clang diagnostic ignored "-Wvector-conversion"
      #endif
      r_.neon_i16 = vshl_s16(a_.neon_i16, vmov_n_s16(HEDLEY_STATIC_CAST(int16_t, vget_lane_u64(count_.neon_u64, 0))));
      HEDLEY_DIAGNOSTIC_POP
    #elif defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR) && defined(SIMDE_BUG_CLANG_POWER9_16x4_BAD_SHIFT)
      if (HEDLEY_UNLIKELY(count_.u64[0] > 15))
        return simde_mm_setzero_si64();

      r_.i16 = a_.i16 << HEDLEY_STATIC_CAST(int16_t, count_.u64[0]);
    #elif defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
      r_.i16 = a_.i16 << count_.u64[0];
    #else
      if (HEDLEY_UNLIKELY(count_.u64[0] > 15)) {
        simde_memset(&r_, 0, sizeof(r_));
        return simde__m64_from_private(r_);
      }

      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.u16) / sizeof(r_.u16[0])) ; i++) {
        r_.u16[i] = HEDLEY_STATIC_CAST(uint16_t, a_.u16[i] << count_.u64[0]);
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_psllw(a, count) simde_mm_sll_pi16(a, count)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_sll_pi16(a, count) simde_mm_sll_pi16(a, count)
#  define _m_psllw(a, count) simde_mm_sll_pi16(a, count)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_sll_pi32 (simde__m64 a, simde__m64 count) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_sll_pi32(a, count);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private count_ = simde__m64_to_private(count);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      HEDLEY_DIAGNOSTIC_PUSH
      #if HEDLEY_HAS_WARNING("-Wvector-conversion") && SIMDE_DETECT_CLANG_VERSION_NOT(10,0,0)
        #pragma clang diagnostic ignored "-Wvector-conversion"
      #endif
      r_.neon_i32 = vshl_s32(a_.neon_i32, vmov_n_s32(HEDLEY_STATIC_CAST(int32_t, vget_lane_u64(count_.neon_u64, 0))));
      HEDLEY_DIAGNOSTIC_POP
    #elif defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
      r_.i32 = a_.i32 << count_.u64[0];
    #else
      if (HEDLEY_UNLIKELY(count_.u64[0] > 31)) {
        simde_memset(&r_, 0, sizeof(r_));
        return simde__m64_from_private(r_);
      }

      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.u32) / sizeof(r_.u32[0])) ; i++) {
        r_.u32[i] = a_.u32[i] << count_.u64[0];
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_pslld(a, count) simde_mm_sll_pi32(a, count)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_sll_pi32(a, count) simde_mm_sll_pi32(a, count)
#  define _m_pslld(a, count) simde_mm_sll_pi32(a, count)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_slli_pi16 (simde__m64 a, int count) {
  #if defined(SIMDE_X86_MMX_NATIVE) && !defined(__PGI)
    return _mm_slli_pi16(a, count);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);

    #if defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i16 = psllh_s(a_.mmi_i16, count);
    #elif defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR) && defined(SIMDE_BUG_CLANG_POWER9_16x4_BAD_SHIFT)
      if (HEDLEY_UNLIKELY(count > 15))
        return simde_mm_setzero_si64();

      r_.i16 = a_.i16 << HEDLEY_STATIC_CAST(int16_t, count);
    #elif defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
      r_.i16 = a_.i16 << count;
    #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_i16 = vshl_s16(a_.neon_i16, vmov_n_s16((int16_t) count));
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.u16) / sizeof(r_.u16[0])) ; i++) {
        r_.u16[i] = HEDLEY_STATIC_CAST(uint16_t, a_.u16[i] << count);
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_psllwi(a, count) simde_mm_slli_pi16(a, count)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_slli_pi16(a, count) simde_mm_slli_pi16(a, count)
#  define _m_psllwi(a, count) simde_mm_slli_pi16(a, count)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_slli_pi32 (simde__m64 a, int count) {
  #if defined(SIMDE_X86_MMX_NATIVE) && !defined(__PGI)
    return _mm_slli_pi32(a, count);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);

    #if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
      r_.i32 = a_.i32 << count;
    #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_i32 = vshl_s32(a_.neon_i32, vmov_n_s32((int32_t) count));
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i32 = psllw_s(a_.mmi_i32, b_.mmi_i32);
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.u32) / sizeof(r_.u32[0])) ; i++) {
        r_.u32[i] = a_.u32[i] << count;
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_pslldi(a, b) simde_mm_slli_pi32(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_slli_pi32(a, count) simde_mm_slli_pi32(a, count)
#  define _m_pslldi(a, count) simde_mm_slli_pi32(a, count)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_slli_si64 (simde__m64 a, int count) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_slli_si64(a, count);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);

    #if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
      r_.i64 = a_.i64 << count;
    #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_i64 = vshl_s64(a_.neon_i64, vmov_n_s64((int64_t) count));
    #else
      r_.u64[0] = a_.u64[0] << count;
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_psllqi(a, count) simde_mm_slli_si64(a, count)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_slli_si64(a, count) simde_mm_slli_si64(a, count)
#  define _m_psllqi(a, count) simde_mm_slli_si64(a, count)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_sll_si64 (simde__m64 a, simde__m64 count) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_sll_si64(a, count);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private count_ = simde__m64_to_private(count);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_i64 = vshl_s64(a_.neon_i64, count_.neon_i64);
    #elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)
      r_.i64 = a_.i64 << count_.i64;
    #else
      if (HEDLEY_UNLIKELY(count_.u64[0] > 63)) {
        simde_memset(&r_, 0, sizeof(r_));
        return simde__m64_from_private(r_);
      }

      r_.u64[0] = a_.u64[0] << count_.u64[0];
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_psllq(a, count) simde_mm_sll_si64(a, count)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_sll_si64(a, count) simde_mm_sll_si64(a, count)
#  define _m_psllq(a, count) simde_mm_sll_si64(a, count)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_srl_pi16 (simde__m64 a, simde__m64 count) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_srl_pi16(a, count);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private count_ = simde__m64_to_private(count);

    #if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR) && defined(SIMDE_BUG_CLANG_POWER9_16x4_BAD_SHIFT)
      if (HEDLEY_UNLIKELY(count_.u64[0] > 15))
        return simde_mm_setzero_si64();

      r_.u16 = a_.u16 >> HEDLEY_STATIC_CAST(uint16_t, count_.u64[0]);
    #elif defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
      r_.u16 = a_.u16 >> count_.u64[0];
    #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_u16 = vshl_u16(a_.neon_u16, vmov_n_s16(-((int16_t) vget_lane_u64(count_.neon_u64, 0))));
    #else
      if (HEDLEY_UNLIKELY(count_.u64[0] > 15)) {
        simde_memset(&r_, 0, sizeof(r_));
        return simde__m64_from_private(r_);
      }

      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < sizeof(r_.u16) / sizeof(r_.u16[0]) ; i++) {
        r_.u16[i] = a_.u16[i] >> count_.u64[0];
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_psrlw(a, count) simde_mm_srl_pi16(a, count)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_srl_pi16(a, count) simde_mm_srl_pi16(a, count)
#  define _m_psrlw(a, count) simde_mm_srl_pi16(a, count)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_srl_pi32 (simde__m64 a, simde__m64 count) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_srl_pi32(a, count);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private count_ = simde__m64_to_private(count);

    #if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
      r_.u32 = a_.u32 >> count_.u64[0];
    #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_u32 = vshl_u32(a_.neon_u32, vmov_n_s32(-((int32_t) vget_lane_u64(count_.neon_u64, 0))));
    #else
      if (HEDLEY_UNLIKELY(count_.u64[0] > 31)) {
        simde_memset(&r_, 0, sizeof(r_));
        return simde__m64_from_private(r_);
      }

      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < sizeof(r_.u32) / sizeof(r_.u32[0]) ; i++) {
        r_.u32[i] = a_.u32[i] >> count_.u64[0];
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_psrld(a, count) simde_mm_srl_pi32(a, count)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_srl_pi32(a, count) simde_mm_srl_pi32(a, count)
#  define _m_psrld(a, count) simde_mm_srl_pi32(a, count)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_srli_pi16 (simde__m64 a, int count) {
  #if defined(SIMDE_X86_MMX_NATIVE) && !defined(__PGI)
    return _mm_srli_pi16(a, count);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);

    #if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
      r_.u16 = a_.u16 >> count;
    #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_u16 = vshl_u16(a_.neon_u16, vmov_n_s16(-((int16_t) count)));
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i16 = psrlh_s(a_.mmi_i16, b_.mmi_i16);
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.u16) / sizeof(r_.u16[0])) ; i++) {
        r_.u16[i] = a_.u16[i] >> count;
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_psrlwi(a, count) simde_mm_srli_pi16(a, count)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_srli_pi16(a, count) simde_mm_srli_pi16(a, count)
#  define _m_psrlwi(a, count) simde_mm_srli_pi16(a, count)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_srli_pi32 (simde__m64 a, int count) {
  #if defined(SIMDE_X86_MMX_NATIVE) && !defined(__PGI)
    return _mm_srli_pi32(a, count);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);

    #if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
      r_.u32 = a_.u32 >> count;
    #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_u32 = vshl_u32(a_.neon_u32, vmov_n_s32(-((int32_t) count)));
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i32 = psrlw_s(a_.mmi_i32, b_.mmi_i32);
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.u32) / sizeof(r_.u32[0])) ; i++) {
        r_.u32[i] = a_.u32[i] >> count;
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_psrldi(a, count) simde_mm_srli_pi32(a, count)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_srli_pi32(a, count) simde_mm_srli_pi32(a, count)
#  define _m_psrldi(a, count) simde_mm_srli_pi32(a, count)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_srli_si64 (simde__m64 a, int count) {
  #if defined(SIMDE_X86_MMX_NATIVE) && !defined(__PGI)
    return _mm_srli_si64(a, count);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_u64 = vshl_u64(a_.neon_u64, vmov_n_s64(-count));
    #elif defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
      r_.u64 = a_.u64 >> count;
    #else
      r_.u64[0] = a_.u64[0] >> count;
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_psrlqi(a, count) simde_mm_srli_si64(a, count)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_srli_si64(a, count) simde_mm_srli_si64(a, count)
#  define _m_psrlqi(a, count) simde_mm_srli_si64(a, count)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_srl_si64 (simde__m64 a, simde__m64 count) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_srl_si64(a, count);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private count_ = simde__m64_to_private(count);

    #if defined(SIMDE_ARM_NEON_A64V8_NATIVE)
      r_.neon_u64 = vshl_u64(a_.neon_u64, vneg_s64(count_.neon_i64));
    #elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)
      r_.u64 = a_.u64 >> count_.u64;
    #else
      if (HEDLEY_UNLIKELY(count_.u64[0] > 63)) {
        simde_memset(&r_, 0, sizeof(r_));
        return simde__m64_from_private(r_);
      }

      r_.u64[0] = a_.u64[0] >> count_.u64[0];
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_psrlq(a, count) simde_mm_srl_si64(a, count)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_srl_si64(a, count) simde_mm_srl_si64(a, count)
#  define _m_psrlq(a, count) simde_mm_srl_si64(a, count)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_srai_pi16 (simde__m64 a, int count) {
  #if defined(SIMDE_X86_MMX_NATIVE) && !defined(__PGI)
    return _mm_srai_pi16(a, count);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);

    #if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
      r_.i16 = a_.i16 >> (count & 0xff);
    #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_i16 = vshl_s16(a_.neon_i16, vmov_n_s16(-HEDLEY_STATIC_CAST(int16_t, count)));
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i16 = psrah_s(a_.mmi_i16, count);
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i16) / sizeof(r_.i16[0])) ; i++) {
        r_.i16[i] = a_.i16[i] >> (count & 0xff);
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_psrawi(a, count) simde_mm_srai_pi16(a, count)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_srai_pi16(a, count) simde_mm_srai_pi16(a, count)
#  define _m_psrawi(a, count) simde_mm_srai_pi16(a, count)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_srai_pi32 (simde__m64 a, int count) {
  #if defined(SIMDE_X86_MMX_NATIVE) && !defined(__PGI)
    return _mm_srai_pi32(a, count);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);

    #if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
      r_.i32 = a_.i32 >> (count & 0xff);
    #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_i32 = vshl_s32(a_.neon_i32, vmov_n_s32(-HEDLEY_STATIC_CAST(int32_t, count)));
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i32 = psraw_s(a_.mmi_i32, count);
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i32) / sizeof(r_.i32[0])) ; i++) {
        r_.i32[i] = a_.i32[i] >> (count & 0xff);
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_psradi(a, count) simde_mm_srai_pi32(a, count)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_srai_pi32(a, count) simde_mm_srai_pi32(a, count)
#  define _m_psradi(a, count) simde_mm_srai_pi32(a, count)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_sra_pi16 (simde__m64 a, simde__m64 count) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_sra_pi16(a, count);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private count_ = simde__m64_to_private(count);
    const int cnt = HEDLEY_STATIC_CAST(int, (count_.i64[0] > 15 ? 15 : count_.i64[0]));

    #if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
      r_.i16 = a_.i16 >> cnt;
    #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_i16 = vshl_s16(a_.neon_i16, vmov_n_s16(-HEDLEY_STATIC_CAST(int16_t, vget_lane_u64(count_.neon_u64, 0))));
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i16) / sizeof(r_.i16[0])) ; i++) {
        r_.i16[i] = a_.i16[i] >> cnt;
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_psraw(a, count) simde_mm_sra_pi16(a, count)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_sra_pi16(a, count) simde_mm_sra_pi16(a, count)
#  define _m_psraw(a, count) simde_mm_sra_pi16(a, count)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_sra_pi32 (simde__m64 a, simde__m64 count) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_sra_pi32(a, count);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private count_ = simde__m64_to_private(count);
    const int32_t cnt = (count_.u64[0] > 31) ? 31 : HEDLEY_STATIC_CAST(int32_t, count_.u64[0]);

    #if defined(SIMDE_VECTOR_SUBSCRIPT_SCALAR)
      r_.i32 = a_.i32 >> cnt;
    #elif defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_i32 = vshl_s32(a_.neon_i32, vmov_n_s32(-HEDLEY_STATIC_CAST(int32_t, vget_lane_u64(count_.neon_u64, 0))));
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i32) / sizeof(r_.i32[0])) ; i++) {
        r_.i32[i] = a_.i32[i] >> cnt;
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_psrad(a, b) simde_mm_sra_pi32(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_sra_pi32(a, count) simde_mm_sra_pi32(a, count)
#  define _m_psrad(a, count) simde_mm_sra_pi32(a, count)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_sub_pi8 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_sub_pi8(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_i8 = vsub_s8(a_.neon_i8, b_.neon_i8);
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i8 = psubb_s(a_.mmi_i8, b_.mmi_i8);
    #elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)
      r_.i8 = a_.i8 - b_.i8;
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i8) / sizeof(r_.i8[0])) ; i++) {
        r_.i8[i] = a_.i8[i] - b_.i8[i];
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_psubb(a, b) simde_mm_sub_pi8(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_sub_pi8(a, b) simde_mm_sub_pi8(a, b)
#  define _m_psubb(a, b) simde_mm_sub_pi8(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_sub_pi16 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_sub_pi16(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_i16 = vsub_s16(a_.neon_i16, b_.neon_i16);
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i16 = psubh_s(a_.mmi_i16, b_.mmi_i16);
    #elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)
      r_.i16 = a_.i16 - b_.i16;
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i16) / sizeof(r_.i16[0])) ; i++) {
        r_.i16[i] = a_.i16[i] - b_.i16[i];
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_psubw(a, b) simde_mm_sub_pi16(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_sub_pi16(a, b) simde_mm_sub_pi16(a, b)
#  define _m_psubw(a, b) simde_mm_sub_pi16(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_sub_pi32 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_sub_pi32(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_i32 = vsub_s32(a_.neon_i32, b_.neon_i32);
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i32 = psubw_s(a_.mmi_i32, b_.mmi_i32);
    #elif defined(SIMDE_VECTOR_SUBSCRIPT_OPS)
      r_.i32 = a_.i32 - b_.i32;
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i32) / sizeof(r_.i32[0])) ; i++) {
        r_.i32[i] = a_.i32[i] - b_.i32[i];
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_psubd(a, b) simde_mm_sub_pi32(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_sub_pi32(a, b) simde_mm_sub_pi32(a, b)
#  define _m_psubd(a, b) simde_mm_sub_pi32(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_subs_pi8 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_subs_pi8(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

    #if defined(SIMDE_ARM_NEON_A32V7_NATIVE)
      r_.neon_i8 = vqsub_s8(a_.neon_i8, b_.neon_i8);
    #elif defined(SIMDE_MIPS_LOONGSON_MMI_NATIVE)
      r_.mmi_i8 = psubsb(a_.mmi_i8, b_.mmi_i8);
    #else
      SIMDE_VECTORIZE
      for (size_t i = 0 ; i < (sizeof(r_.i8) / sizeof(r_.i8[0])) ; i++) {
        if (((b_.i8[i]) > 0 && (a_.i8[i]) < INT8_MIN + (b_.i8[i]))) {
          r_.i8[i] = INT8_MIN;
        } else if ((b_.i8[i]) < 0 && (a_.i8[i]) > INT8_MAX + (b_.i8[i])) {
          r_.i8[i] = INT8_MAX;
        } else {
          r_.i8[i] = (a_.i8[i]) - (b_.i8[i]);
        }
      }
    #endif

    return simde__m64_from_private(r_);
  #endif
}
#define simde_m_psubsb(a, b) simde_mm_subs_pi8(a, b)
#if defined(SIMDE_X86_MMX_ENABLE_NATIVE_ALIASES)
#  define _mm_subs_pi8(a, b) simde_mm_subs_pi8(a, b)
#  define _m_psubsb(a, b) simde_mm_subs_pi8(a, b)
#endif

SIMDE_FUNCTION_ATTRIBUTES
simde__m64
simde_mm_subs_pu8 (simde__m64 a, simde__m64 b) {
  #if defined(SIMDE_X86_MMX_NATIVE)
    return _mm_subs_pu8(a, b);
  #else
    simde__m64_private r_;
    simde__m64_private a_ = simde__m64_to_private(a);
    simde__m64_private b_ = simde__m64_to_private(b);

--> --------------------

--> maximum size reached

--> --------------------