xmmintrin.h

Go to the documentation of this file.

/*===---- xmmintrin.h - SSE intrinsics -------------------------------------===
 *
 * Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 * See https://llvm.org/LICENSE.txt for license information.
 * SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 *
 *===-----------------------------------------------------------------------===
 */
 
#ifndef __XMMINTRIN_H
#define __XMMINTRIN_H
 
#include <mmintrin.h>
 
typedef int __v4si __attribute__((__vector_size__(16)));
typedef float __v4sf __attribute__((__vector_size__(16)));
typedef float __m128 __attribute__((__vector_size__(16), __aligned__(16)));
 
typedef float __m128_u __attribute__((__vector_size__(16), __aligned__(1)));
 
/* Unsigned types */
typedef unsigned int __v4su __attribute__((__vector_size__(16)));
 
/* This header should only be included in a hosted environment as it depends on
 * a standard library to provide allocation routines. */
#if __STDC_HOSTED__
#include <mm_malloc.h>
#endif
 
/* Define the default attributes for the functions in this file. */
#define __DEFAULT_FN_ATTRS __attribute__((__always_inline__, __nodebug__, __target__("sse"), __min_vector_width__(128)))
#define __DEFAULT_FN_ATTRS_MMX __attribute__((__always_inline__, __nodebug__, __target__("mmx,sse"), __min_vector_width__(64)))
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_add_ss(__m128 __a, __m128 __b)
{
  __a[0] += __b[0];
  return __a;
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_add_ps(__m128 __a, __m128 __b)
{
  return (__m128)((__v4sf)__a + (__v4sf)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_sub_ss(__m128 __a, __m128 __b)
{
  __a[0] -= __b[0];
  return __a;
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_sub_ps(__m128 __a, __m128 __b)
{
  return (__m128)((__v4sf)__a - (__v4sf)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_mul_ss(__m128 __a, __m128 __b)
{
  __a[0] *= __b[0];
  return __a;
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_mul_ps(__m128 __a, __m128 __b)
{
  return (__m128)((__v4sf)__a * (__v4sf)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_div_ss(__m128 __a, __m128 __b)
{
  __a[0] /= __b[0];
  return __a;
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_div_ps(__m128 __a, __m128 __b)
{
  return (__m128)((__v4sf)__a / (__v4sf)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_sqrt_ss(__m128 __a)
{
  return (__m128)__builtin_ia32_sqrtss((__v4sf)__a);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_sqrt_ps(__m128 __a)
{
  return __builtin_ia32_sqrtps((__v4sf)__a);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_rcp_ss(__m128 __a)
{
  return (__m128)__builtin_ia32_rcpss((__v4sf)__a);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_rcp_ps(__m128 __a)
{
  return (__m128)__builtin_ia32_rcpps((__v4sf)__a);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_rsqrt_ss(__m128 __a)
{
  return __builtin_ia32_rsqrtss((__v4sf)__a);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_rsqrt_ps(__m128 __a)
{
  return __builtin_ia32_rsqrtps((__v4sf)__a);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_min_ss(__m128 __a, __m128 __b)
{
  return __builtin_ia32_minss((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_min_ps(__m128 __a, __m128 __b)
{
  return __builtin_ia32_minps((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_max_ss(__m128 __a, __m128 __b)
{
  return __builtin_ia32_maxss((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_max_ps(__m128 __a, __m128 __b)
{
  return __builtin_ia32_maxps((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_and_ps(__m128 __a, __m128 __b)
{
  return (__m128)((__v4su)__a & (__v4su)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_andnot_ps(__m128 __a, __m128 __b)
{
  return (__m128)(~(__v4su)__a & (__v4su)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_or_ps(__m128 __a, __m128 __b)
{
  return (__m128)((__v4su)__a | (__v4su)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_xor_ps(__m128 __a, __m128 __b)
{
  return (__m128)((__v4su)__a ^ (__v4su)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmpeq_ss(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_ia32_cmpeqss((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmpeq_ps(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_ia32_cmpeqps((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmplt_ss(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_ia32_cmpltss((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmplt_ps(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_ia32_cmpltps((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmple_ss(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_ia32_cmpless((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmple_ps(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_ia32_cmpleps((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmpgt_ss(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_shufflevector((__v4sf)__a,
                                         (__v4sf)__builtin_ia32_cmpltss((__v4sf)__b, (__v4sf)__a),
                                         4, 1, 2, 3);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmpgt_ps(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_ia32_cmpltps((__v4sf)__b, (__v4sf)__a);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmpge_ss(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_shufflevector((__v4sf)__a,
                                         (__v4sf)__builtin_ia32_cmpless((__v4sf)__b, (__v4sf)__a),
                                         4, 1, 2, 3);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmpge_ps(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_ia32_cmpleps((__v4sf)__b, (__v4sf)__a);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmpneq_ss(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_ia32_cmpneqss((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmpneq_ps(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_ia32_cmpneqps((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmpnlt_ss(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_ia32_cmpnltss((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmpnlt_ps(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_ia32_cmpnltps((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmpnle_ss(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_ia32_cmpnless((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmpnle_ps(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_ia32_cmpnleps((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmpngt_ss(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_shufflevector((__v4sf)__a,
                                         (__v4sf)__builtin_ia32_cmpnltss((__v4sf)__b, (__v4sf)__a),
                                         4, 1, 2, 3);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmpngt_ps(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_ia32_cmpnltps((__v4sf)__b, (__v4sf)__a);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmpnge_ss(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_shufflevector((__v4sf)__a,
                                         (__v4sf)__builtin_ia32_cmpnless((__v4sf)__b, (__v4sf)__a),
                                         4, 1, 2, 3);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmpnge_ps(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_ia32_cmpnleps((__v4sf)__b, (__v4sf)__a);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmpord_ss(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_ia32_cmpordss((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmpord_ps(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_ia32_cmpordps((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmpunord_ss(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_ia32_cmpunordss((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cmpunord_ps(__m128 __a, __m128 __b)
{
  return (__m128)__builtin_ia32_cmpunordps((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ int __DEFAULT_FN_ATTRS
_mm_comieq_ss(__m128 __a, __m128 __b)
{
  return __builtin_ia32_comieq((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ int __DEFAULT_FN_ATTRS
_mm_comilt_ss(__m128 __a, __m128 __b)
{
  return __builtin_ia32_comilt((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ int __DEFAULT_FN_ATTRS
_mm_comile_ss(__m128 __a, __m128 __b)
{
  return __builtin_ia32_comile((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ int __DEFAULT_FN_ATTRS
_mm_comigt_ss(__m128 __a, __m128 __b)
{
  return __builtin_ia32_comigt((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ int __DEFAULT_FN_ATTRS
_mm_comige_ss(__m128 __a, __m128 __b)
{
  return __builtin_ia32_comige((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ int __DEFAULT_FN_ATTRS
_mm_comineq_ss(__m128 __a, __m128 __b)
{
  return __builtin_ia32_comineq((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ int __DEFAULT_FN_ATTRS
_mm_ucomieq_ss(__m128 __a, __m128 __b)
{
  return __builtin_ia32_ucomieq((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ int __DEFAULT_FN_ATTRS
_mm_ucomilt_ss(__m128 __a, __m128 __b)
{
  return __builtin_ia32_ucomilt((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ int __DEFAULT_FN_ATTRS
_mm_ucomile_ss(__m128 __a, __m128 __b)
{
  return __builtin_ia32_ucomile((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ int __DEFAULT_FN_ATTRS
_mm_ucomigt_ss(__m128 __a, __m128 __b)
{
  return __builtin_ia32_ucomigt((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ int __DEFAULT_FN_ATTRS
_mm_ucomige_ss(__m128 __a, __m128 __b)
{
  return __builtin_ia32_ucomige((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ int __DEFAULT_FN_ATTRS
_mm_ucomineq_ss(__m128 __a, __m128 __b)
{
  return __builtin_ia32_ucomineq((__v4sf)__a, (__v4sf)__b);
}
 
static __inline__ int __DEFAULT_FN_ATTRS
_mm_cvtss_si32(__m128 __a)
{
  return __builtin_ia32_cvtss2si((__v4sf)__a);
}
 
static __inline__ int __DEFAULT_FN_ATTRS
_mm_cvt_ss2si(__m128 __a)
{
  return _mm_cvtss_si32(__a);
}
 
#ifdef __x86_64__
 
static __inline__ long long __DEFAULT_FN_ATTRS
_mm_cvtss_si64(__m128 __a)
{
  return __builtin_ia32_cvtss2si64((__v4sf)__a);
}
 
#endif
 
static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX
_mm_cvtps_pi32(__m128 __a)
{
  return (__m64)__builtin_ia32_cvtps2pi((__v4sf)__a);
}
 
static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX
_mm_cvt_ps2pi(__m128 __a)
{
  return _mm_cvtps_pi32(__a);
}
 
static __inline__ int __DEFAULT_FN_ATTRS
_mm_cvttss_si32(__m128 __a)
{
  return __builtin_ia32_cvttss2si((__v4sf)__a);
}
 
static __inline__ int __DEFAULT_FN_ATTRS
_mm_cvtt_ss2si(__m128 __a)
{
  return _mm_cvttss_si32(__a);
}
 
#ifdef __x86_64__
static __inline__ long long __DEFAULT_FN_ATTRS
_mm_cvttss_si64(__m128 __a)
{
  return __builtin_ia32_cvttss2si64((__v4sf)__a);
}
#endif
 
static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX
_mm_cvttps_pi32(__m128 __a)
{
  return (__m64)__builtin_ia32_cvttps2pi((__v4sf)__a);
}
 
static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX
_mm_cvtt_ps2pi(__m128 __a)
{
  return _mm_cvttps_pi32(__a);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cvtsi32_ss(__m128 __a, int __b)
{
  __a[0] = __b;
  return __a;
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cvt_si2ss(__m128 __a, int __b)
{
  return _mm_cvtsi32_ss(__a, __b);
}
 
#ifdef __x86_64__
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_cvtsi64_ss(__m128 __a, long long __b)
{
  __a[0] = __b;
  return __a;
}
 
#endif
 
static __inline__ __m128 __DEFAULT_FN_ATTRS_MMX
_mm_cvtpi32_ps(__m128 __a, __m64 __b)
{
  return __builtin_ia32_cvtpi2ps((__v4sf)__a, (__v2si)__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS_MMX
_mm_cvt_pi2ps(__m128 __a, __m64 __b)
{
  return _mm_cvtpi32_ps(__a, __b);
}
 
static __inline__ float __DEFAULT_FN_ATTRS
_mm_cvtss_f32(__m128 __a)
{
  return __a[0];
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_loadh_pi(__m128 __a, const __m64 *__p)
{
  typedef float __mm_loadh_pi_v2f32 __attribute__((__vector_size__(8)));
  struct __mm_loadh_pi_struct {
    __mm_loadh_pi_v2f32 __u;
  } __attribute__((__packed__, __may_alias__));
  __mm_loadh_pi_v2f32 __b = ((const struct __mm_loadh_pi_struct*)__p)->__u;
  __m128 __bb = __builtin_shufflevector(__b, __b, 0, 1, 0, 1);
  return __builtin_shufflevector(__a, __bb, 0, 1, 4, 5);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_loadl_pi(__m128 __a, const __m64 *__p)
{
  typedef float __mm_loadl_pi_v2f32 __attribute__((__vector_size__(8)));
  struct __mm_loadl_pi_struct {
    __mm_loadl_pi_v2f32 __u;
  } __attribute__((__packed__, __may_alias__));
  __mm_loadl_pi_v2f32 __b = ((const struct __mm_loadl_pi_struct*)__p)->__u;
  __m128 __bb = __builtin_shufflevector(__b, __b, 0, 1, 0, 1);
  return __builtin_shufflevector(__a, __bb, 4, 5, 2, 3);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_load_ss(const float *__p)
{
  struct __mm_load_ss_struct {
    float __u;
  } __attribute__((__packed__, __may_alias__));
  float __u = ((const struct __mm_load_ss_struct*)__p)->__u;
  return __extension__ (__m128){ __u, 0, 0, 0 };
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_load1_ps(const float *__p)
{
  struct __mm_load1_ps_struct {
    float __u;
  } __attribute__((__packed__, __may_alias__));
  float __u = ((const struct __mm_load1_ps_struct*)__p)->__u;
  return __extension__ (__m128){ __u, __u, __u, __u };
}
 
#define        _mm_load_ps1(p) _mm_load1_ps(p)
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_load_ps(const float *__p)
{
  return *(const __m128*)__p;
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_loadu_ps(const float *__p)
{
  struct __loadu_ps {
    __m128_u __v;
  } __attribute__((__packed__, __may_alias__));
  return ((const struct __loadu_ps*)__p)->__v;
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_loadr_ps(const float *__p)
{
  __m128 __a = _mm_load_ps(__p);
  return __builtin_shufflevector((__v4sf)__a, (__v4sf)__a, 3, 2, 1, 0);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_undefined_ps(void)
{
  return (__m128)__builtin_ia32_undef128();
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_set_ss(float __w)
{
  return __extension__ (__m128){ __w, 0, 0, 0 };
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_set1_ps(float __w)
{
  return __extension__ (__m128){ __w, __w, __w, __w };
}
 
/* Microsoft specific. */
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_set_ps1(float __w)
{
    return _mm_set1_ps(__w);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_set_ps(float __z, float __y, float __x, float __w)
{
  return __extension__ (__m128){ __w, __x, __y, __z };
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_setr_ps(float __z, float __y, float __x, float __w)
{
  return __extension__ (__m128){ __z, __y, __x, __w };
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_setzero_ps(void)
{
  return __extension__ (__m128){ 0, 0, 0, 0 };
}
 
static __inline__ void __DEFAULT_FN_ATTRS
_mm_storeh_pi(__m64 *__p, __m128 __a)
{
  typedef float __mm_storeh_pi_v2f32 __attribute__((__vector_size__(8)));
  struct __mm_storeh_pi_struct {
    __mm_storeh_pi_v2f32 __u;
  } __attribute__((__packed__, __may_alias__));
  ((struct __mm_storeh_pi_struct*)__p)->__u = __builtin_shufflevector(__a, __a, 2, 3);
}
 
static __inline__ void __DEFAULT_FN_ATTRS
_mm_storel_pi(__m64 *__p, __m128 __a)
{
  typedef float __mm_storeh_pi_v2f32 __attribute__((__vector_size__(8)));
  struct __mm_storeh_pi_struct {
    __mm_storeh_pi_v2f32 __u;
  } __attribute__((__packed__, __may_alias__));
  ((struct __mm_storeh_pi_struct*)__p)->__u = __builtin_shufflevector(__a, __a, 0, 1);
}
 
static __inline__ void __DEFAULT_FN_ATTRS
_mm_store_ss(float *__p, __m128 __a)
{
  struct __mm_store_ss_struct {
    float __u;
  } __attribute__((__packed__, __may_alias__));
  ((struct __mm_store_ss_struct*)__p)->__u = __a[0];
}
 
static __inline__ void __DEFAULT_FN_ATTRS
_mm_storeu_ps(float *__p, __m128 __a)
{
  struct __storeu_ps {
    __m128_u __v;
  } __attribute__((__packed__, __may_alias__));
  ((struct __storeu_ps*)__p)->__v = __a;
}
 
static __inline__ void __DEFAULT_FN_ATTRS
_mm_store_ps(float *__p, __m128 __a)
{
  *(__m128*)__p = __a;
}
 
static __inline__ void __DEFAULT_FN_ATTRS
_mm_store1_ps(float *__p, __m128 __a)
{
  __a = __builtin_shufflevector((__v4sf)__a, (__v4sf)__a, 0, 0, 0, 0);
  _mm_store_ps(__p, __a);
}
 
static __inline__ void __DEFAULT_FN_ATTRS
_mm_store_ps1(float *__p, __m128 __a)
{
  _mm_store1_ps(__p, __a);
}
 
static __inline__ void __DEFAULT_FN_ATTRS
_mm_storer_ps(float *__p, __m128 __a)
{
  __a = __builtin_shufflevector((__v4sf)__a, (__v4sf)__a, 3, 2, 1, 0);
  _mm_store_ps(__p, __a);
}
 
#define _MM_HINT_ET0 7
#define _MM_HINT_ET1 6
#define _MM_HINT_T0  3
#define _MM_HINT_T1  2
#define _MM_HINT_T2  1
#define _MM_HINT_NTA 0
 
#ifndef _MSC_VER
/* FIXME: We have to #define this because "sel" must be a constant integer, and
   Sema doesn't do any form of constant propagation yet. */
 
#define _mm_prefetch(a, sel) (__builtin_prefetch((const void *)(a), \
                                                 ((sel) >> 2) & 1, (sel) & 0x3))
#endif
 
static __inline__ void __DEFAULT_FN_ATTRS_MMX
_mm_stream_pi(__m64 *__p, __m64 __a)
{
  __builtin_ia32_movntq(__p, __a);
}
 
static __inline__ void __DEFAULT_FN_ATTRS
_mm_stream_ps(float *__p, __m128 __a)
{
  __builtin_nontemporal_store((__v4sf)__a, (__v4sf*)__p);
}
 
#if defined(__cplusplus)
extern "C" {
#endif
 
void _mm_sfence(void);
 
#if defined(__cplusplus)
} // extern "C"
#endif
 
#define _mm_extract_pi16(a, n) \
  (int)__builtin_ia32_vec_ext_v4hi((__v4hi)a, (int)n)
 
#define _mm_insert_pi16(a, d, n) \
  (__m64)__builtin_ia32_vec_set_v4hi((__v4hi)a, (int)d, (int)n)
 
static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX
_mm_max_pi16(__m64 __a, __m64 __b)
{
  return (__m64)__builtin_ia32_pmaxsw((__v4hi)__a, (__v4hi)__b);
}
 
static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX
_mm_max_pu8(__m64 __a, __m64 __b)
{
  return (__m64)__builtin_ia32_pmaxub((__v8qi)__a, (__v8qi)__b);
}
 
static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX
_mm_min_pi16(__m64 __a, __m64 __b)
{
  return (__m64)__builtin_ia32_pminsw((__v4hi)__a, (__v4hi)__b);
}
 
static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX
_mm_min_pu8(__m64 __a, __m64 __b)
{
  return (__m64)__builtin_ia32_pminub((__v8qi)__a, (__v8qi)__b);
}
 
static __inline__ int __DEFAULT_FN_ATTRS_MMX
_mm_movemask_pi8(__m64 __a)
{
  return __builtin_ia32_pmovmskb((__v8qi)__a);
}
 
static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX
_mm_mulhi_pu16(__m64 __a, __m64 __b)
{
  return (__m64)__builtin_ia32_pmulhuw((__v4hi)__a, (__v4hi)__b);
}
 
#define _mm_shuffle_pi16(a, n) \
  (__m64)__builtin_ia32_pshufw((__v4hi)(__m64)(a), (n))
 
static __inline__ void __DEFAULT_FN_ATTRS_MMX
_mm_maskmove_si64(__m64 __d, __m64 __n, char *__p)
{
  __builtin_ia32_maskmovq((__v8qi)__d, (__v8qi)__n, __p);
}
 
static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX
_mm_avg_pu8(__m64 __a, __m64 __b)
{
  return (__m64)__builtin_ia32_pavgb((__v8qi)__a, (__v8qi)__b);
}
 
static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX
_mm_avg_pu16(__m64 __a, __m64 __b)
{
  return (__m64)__builtin_ia32_pavgw((__v4hi)__a, (__v4hi)__b);
}
 
static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX
_mm_sad_pu8(__m64 __a, __m64 __b)
{
  return (__m64)__builtin_ia32_psadbw((__v8qi)__a, (__v8qi)__b);
}
 
#if defined(__cplusplus)
extern "C" {
#endif
 
unsigned int _mm_getcsr(void);
 
void _mm_setcsr(unsigned int __i);
 
#if defined(__cplusplus)
} // extern "C"
#endif
 
#define _mm_shuffle_ps(a, b, mask) \
  (__m128)__builtin_ia32_shufps((__v4sf)(__m128)(a), (__v4sf)(__m128)(b), \
                                (int)(mask))
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_unpackhi_ps(__m128 __a, __m128 __b)
{
  return __builtin_shufflevector((__v4sf)__a, (__v4sf)__b, 2, 6, 3, 7);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_unpacklo_ps(__m128 __a, __m128 __b)
{
  return __builtin_shufflevector((__v4sf)__a, (__v4sf)__b, 0, 4, 1, 5);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_move_ss(__m128 __a, __m128 __b)
{
  __a[0] = __b[0];
  return __a;
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_movehl_ps(__m128 __a, __m128 __b)
{
  return __builtin_shufflevector((__v4sf)__a, (__v4sf)__b, 6, 7, 2, 3);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS
_mm_movelh_ps(__m128 __a, __m128 __b)
{
  return __builtin_shufflevector((__v4sf)__a, (__v4sf)__b, 0, 1, 4, 5);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS_MMX
_mm_cvtpi16_ps(__m64 __a)
{
  __m64 __b, __c;
  __m128 __r;
 
  __b = _mm_setzero_si64();
  __b = _mm_cmpgt_pi16(__b, __a);
  __c = _mm_unpackhi_pi16(__a, __b);
  __r = _mm_setzero_ps();
  __r = _mm_cvtpi32_ps(__r, __c);
  __r = _mm_movelh_ps(__r, __r);
  __c = _mm_unpacklo_pi16(__a, __b);
  __r = _mm_cvtpi32_ps(__r, __c);
 
  return __r;
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS_MMX
_mm_cvtpu16_ps(__m64 __a)
{
  __m64 __b, __c;
  __m128 __r;
 
  __b = _mm_setzero_si64();
  __c = _mm_unpackhi_pi16(__a, __b);
  __r = _mm_setzero_ps();
  __r = _mm_cvtpi32_ps(__r, __c);
  __r = _mm_movelh_ps(__r, __r);
  __c = _mm_unpacklo_pi16(__a, __b);
  __r = _mm_cvtpi32_ps(__r, __c);
 
  return __r;
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS_MMX
_mm_cvtpi8_ps(__m64 __a)
{
  __m64 __b;
 
  __b = _mm_setzero_si64();
  __b = _mm_cmpgt_pi8(__b, __a);
  __b = _mm_unpacklo_pi8(__a, __b);
 
  return _mm_cvtpi16_ps(__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS_MMX
_mm_cvtpu8_ps(__m64 __a)
{
  __m64 __b;
 
  __b = _mm_setzero_si64();
  __b = _mm_unpacklo_pi8(__a, __b);
 
  return _mm_cvtpi16_ps(__b);
}
 
static __inline__ __m128 __DEFAULT_FN_ATTRS_MMX
_mm_cvtpi32x2_ps(__m64 __a, __m64 __b)
{
  __m128 __c;
 
  __c = _mm_setzero_ps();
  __c = _mm_cvtpi32_ps(__c, __b);
  __c = _mm_movelh_ps(__c, __c);
 
  return _mm_cvtpi32_ps(__c, __a);
}
 
static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX
_mm_cvtps_pi16(__m128 __a)
{
  __m64 __b, __c;
 
  __b = _mm_cvtps_pi32(__a);
  __a = _mm_movehl_ps(__a, __a);
  __c = _mm_cvtps_pi32(__a);
 
  return _mm_packs_pi32(__b, __c);
}
 
static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX
_mm_cvtps_pi8(__m128 __a)
{
  __m64 __b, __c;
 
  __b = _mm_cvtps_pi16(__a);
  __c = _mm_setzero_si64();
 
  return _mm_packs_pi16(__b, __c);
}
 
static __inline__ int __DEFAULT_FN_ATTRS
_mm_movemask_ps(__m128 __a)
{
  return __builtin_ia32_movmskps((__v4sf)__a);
}
 
 
#define _MM_ALIGN16 __attribute__((aligned(16)))
 
#define _MM_SHUFFLE(z, y, x, w) (((z) << 6) | ((y) << 4) | ((x) << 2) | (w))
 
#define _MM_EXCEPT_INVALID    (0x0001U)
#define _MM_EXCEPT_DENORM     (0x0002U)
#define _MM_EXCEPT_DIV_ZERO   (0x0004U)
#define _MM_EXCEPT_OVERFLOW   (0x0008U)
#define _MM_EXCEPT_UNDERFLOW  (0x0010U)
#define _MM_EXCEPT_INEXACT    (0x0020U)
#define _MM_EXCEPT_MASK       (0x003fU)
 
#define _MM_MASK_INVALID      (0x0080U)
#define _MM_MASK_DENORM       (0x0100U)
#define _MM_MASK_DIV_ZERO     (0x0200U)
#define _MM_MASK_OVERFLOW     (0x0400U)
#define _MM_MASK_UNDERFLOW    (0x0800U)
#define _MM_MASK_INEXACT      (0x1000U)
#define _MM_MASK_MASK         (0x1f80U)
 
#define _MM_ROUND_NEAREST     (0x0000U)
#define _MM_ROUND_DOWN        (0x2000U)
#define _MM_ROUND_UP          (0x4000U)
#define _MM_ROUND_TOWARD_ZERO (0x6000U)
#define _MM_ROUND_MASK        (0x6000U)
 
#define _MM_FLUSH_ZERO_MASK   (0x8000U)
#define _MM_FLUSH_ZERO_ON     (0x8000U)
#define _MM_FLUSH_ZERO_OFF    (0x0000U)
 
#define _MM_GET_EXCEPTION_MASK() (_mm_getcsr() & _MM_MASK_MASK)
#define _MM_GET_EXCEPTION_STATE() (_mm_getcsr() & _MM_EXCEPT_MASK)
#define _MM_GET_FLUSH_ZERO_MODE() (_mm_getcsr() & _MM_FLUSH_ZERO_MASK)
#define _MM_GET_ROUNDING_MODE() (_mm_getcsr() & _MM_ROUND_MASK)
 
#define _MM_SET_EXCEPTION_MASK(x) (_mm_setcsr((_mm_getcsr() & ~_MM_MASK_MASK) | (x)))
#define _MM_SET_EXCEPTION_STATE(x) (_mm_setcsr((_mm_getcsr() & ~_MM_EXCEPT_MASK) | (x)))
#define _MM_SET_FLUSH_ZERO_MODE(x) (_mm_setcsr((_mm_getcsr() & ~_MM_FLUSH_ZERO_MASK) | (x)))
#define _MM_SET_ROUNDING_MODE(x) (_mm_setcsr((_mm_getcsr() & ~_MM_ROUND_MASK) | (x)))
 
#define _MM_TRANSPOSE4_PS(row0, row1, row2, row3) \
do { \
  __m128 tmp3, tmp2, tmp1, tmp0; \
  tmp0 = _mm_unpacklo_ps((row0), (row1)); \
  tmp2 = _mm_unpacklo_ps((row2), (row3)); \
  tmp1 = _mm_unpackhi_ps((row0), (row1)); \
  tmp3 = _mm_unpackhi_ps((row2), (row3)); \
  (row0) = _mm_movelh_ps(tmp0, tmp2); \
  (row1) = _mm_movehl_ps(tmp2, tmp0); \
  (row2) = _mm_movelh_ps(tmp1, tmp3); \
  (row3) = _mm_movehl_ps(tmp3, tmp1); \
} while (0)
 
/* Aliases for compatibility. */
#define _m_pextrw _mm_extract_pi16
#define _m_pinsrw _mm_insert_pi16
#define _m_pmaxsw _mm_max_pi16
#define _m_pmaxub _mm_max_pu8
#define _m_pminsw _mm_min_pi16
#define _m_pminub _mm_min_pu8
#define _m_pmovmskb _mm_movemask_pi8
#define _m_pmulhuw _mm_mulhi_pu16
#define _m_pshufw _mm_shuffle_pi16
#define _m_maskmovq _mm_maskmove_si64
#define _m_pavgb _mm_avg_pu8
#define _m_pavgw _mm_avg_pu16
#define _m_psadbw _mm_sad_pu8
#define _m_ _mm_
#define _m_ _mm_
 
#undef __DEFAULT_FN_ATTRS
#undef __DEFAULT_FN_ATTRS_MMX
 
/* Ugly hack for backwards-compatibility (compatible with gcc) */
#if defined(__SSE2__) && !__building_module(_Builtin_intrinsics)
#include <emmintrin.h>
#endif
 
#endif /* __XMMINTRIN_H */