xmmintrin.h

mingw32.rar

_MM_SET_ROUNDING_MODE (unsigned int __mode)
{
  _mm_setcsr((_mm_getcsr() & ~_MM_ROUND_MASK) | __mode);
}

static __inline void
_MM_SET_FLUSH_ZERO_MODE (unsigned int __mode)
{
  _mm_setcsr((_mm_getcsr() & ~_MM_FLUSH_ZERO_MASK) | __mode);
}

/* Create a vector with element 0 as *P and the rest zero.  */
static __inline __m128
_mm_load_ss (float const *__P)
{
  return (__m128) __builtin_ia32_loadss (__P);
}

/* Create a vector with all four elements equal to *P.  */
static __inline __m128
_mm_load1_ps (float const *__P)
{
  __v4sf __tmp = __builtin_ia32_loadss (__P);
  return (__m128) __builtin_ia32_shufps (__tmp, __tmp, _MM_SHUFFLE (0,0,0,0));
}

static __inline __m128
_mm_load_ps1 (float const *__P)
{
  return _mm_load1_ps (__P);
}

/* Load four SPFP values from P.  The address must be 16-byte aligned.  */
static __inline __m128
_mm_load_ps (float const *__P)
{
  return (__m128) __builtin_ia32_loadaps (__P);
}

/* Load four SPFP values from P.  The address need not be 16-byte aligned.  */
static __inline __m128
_mm_loadu_ps (float const *__P)
{
  return (__m128) __builtin_ia32_loadups (__P);
}

/* Load four SPFP values in reverse order.  The address must be aligned.  */
static __inline __m128
_mm_loadr_ps (float const *__P)
{
  __v4sf __tmp = __builtin_ia32_loadaps (__P);
  return (__m128) __builtin_ia32_shufps (__tmp, __tmp, _MM_SHUFFLE (0,1,2,3));
}

/* Create a vector with element 0 as F and the rest zero.  */
static __inline __m128
_mm_set_ss (float __F)
{
  return (__m128) __builtin_ia32_loadss (&__F);
}

/* Create a vector with all four elements equal to F.  */
static __inline __m128
_mm_set1_ps (float __F)
{
  __v4sf __tmp = __builtin_ia32_loadss (&__F);
  return (__m128) __builtin_ia32_shufps (__tmp, __tmp, _MM_SHUFFLE (0,0,0,0));
}

static __inline __m128
_mm_set_ps1 (float __F)
{
  return _mm_set1_ps (__F);
}

/* Create the vector [Z Y X W].  */
static __inline __m128
_mm_set_ps (const float __Z, const float __Y, const float __X, const float __W)
{
  return (__v4sf) {__W, __X, __Y, __Z};
}

/* Create the vector [W X Y Z].  */
static __inline __m128
_mm_setr_ps (float __Z, float __Y, float __X, float __W)
{
  return _mm_set_ps (__W, __X, __Y, __Z);
}

/* Create a vector of zeros.  */
static __inline __m128
_mm_setzero_ps (void)
{
  return (__m128) __builtin_ia32_setzerops ();
}

/* Stores the lower SPFP value.  */
static __inline void
_mm_store_ss (float *__P, __m128 __A)
{
  __builtin_ia32_storess (__P, (__v4sf)__A);
}

/* Store the lower SPFP value across four words.  */
static __inline void
_mm_store1_ps (float *__P, __m128 __A)
{
  __v4sf __va = (__v4sf)__A;
  __v4sf __tmp = __builtin_ia32_shufps (__va, __va, _MM_SHUFFLE (0,0,0,0));
  __builtin_ia32_storeaps (__P, __tmp);
}

static __inline void
_mm_store_ps1 (float *__P, __m128 __A)
{
  _mm_store1_ps (__P, __A);
}

/* Store four SPFP values.  The address must be 16-byte aligned.  */
static __inline void
_mm_store_ps (float *__P, __m128 __A)
{
  __builtin_ia32_storeaps (__P, (__v4sf)__A);
}

/* Store four SPFP values.  The address need not be 16-byte aligned.  */
static __inline void
_mm_storeu_ps (float *__P, __m128 __A)
{
  __builtin_ia32_storeups (__P, (__v4sf)__A);
}

/* Store four SPFP values in reverse order.  The address must be aligned.  */
static __inline void
_mm_storer_ps (float *__P, __m128 __A)
{
  __v4sf __va = (__v4sf)__A;
  __v4sf __tmp = __builtin_ia32_shufps (__va, __va, _MM_SHUFFLE (0,1,2,3));
  __builtin_ia32_storeaps (__P, __tmp);
}

/* Sets the low SPFP value of A from the low value of B.  */
static __inline __m128
_mm_move_ss (__m128 __A, __m128 __B)
{
  return (__m128) __builtin_ia32_movss ((__v4sf)__A, (__v4sf)__B);
}

/* Extracts one of the four words of A.  The selector N must be immediate.  */
#if 0
static __inline int
_mm_extract_pi16 (__m64 __A, int __N)
{
  return __builtin_ia32_pextrw ((__v4hi)__A, __N);
}

static __inline int
_m_pextrw (__m64 __A, int __N)
{
  return _mm_extract_pi16 (__A, __N);
}
#else
#define _mm_extract_pi16(A, N) \
  __builtin_ia32_pextrw ((__v4hi)(A), (N))
#define _m_pextrw(A, N)		_mm_extract_pi16((A), (N))
#endif

/* Inserts word D into one of four words of A.  The selector N must be
   immediate.  */
#if 0
static __inline __m64
_mm_insert_pi16 (__m64 __A, int __D, int __N)
{
  return (__m64)__builtin_ia32_pinsrw ((__v4hi)__A, __D, __N);
}

static __inline __m64
_m_pinsrw (__m64 __A, int __D, int __N)
{
  return _mm_insert_pi16 (__A, __D, __N);
}
#else
#define _mm_insert_pi16(A, D, N) \
  ((__m64) __builtin_ia32_pinsrw ((__v4hi)(A), (D), (N)))
#define _m_pinsrw(A, D, N)	 _mm_insert_pi16((A), (D), (N))
#endif

/* Compute the element-wise maximum of signed 16-bit values.  */
static __inline __m64
_mm_max_pi16 (__m64 __A, __m64 __B)
{
  return (__m64) __builtin_ia32_pmaxsw ((__v4hi)__A, (__v4hi)__B);
}

static __inline __m64
_m_pmaxsw (__m64 __A, __m64 __B)
{
  return _mm_max_pi16 (__A, __B);
}

/* Compute the element-wise maximum of unsigned 8-bit values.  */
static __inline __m64
_mm_max_pu8 (__m64 __A, __m64 __B)
{
  return (__m64) __builtin_ia32_pmaxub ((__v8qi)__A, (__v8qi)__B);
}

static __inline __m64
_m_pmaxub (__m64 __A, __m64 __B)
{
  return _mm_max_pu8 (__A, __B);
}

/* Compute the element-wise minimum of signed 16-bit values.  */
static __inline __m64
_mm_min_pi16 (__m64 __A, __m64 __B)
{
  return (__m64) __builtin_ia32_pminsw ((__v4hi)__A, (__v4hi)__B);
}

static __inline __m64
_m_pminsw (__m64 __A, __m64 __B)
{
  return _mm_min_pi16 (__A, __B);
}

/* Compute the element-wise minimum of unsigned 8-bit values.  */
static __inline __m64
_mm_min_pu8 (__m64 __A, __m64 __B)
{
  return (__m64) __builtin_ia32_pminub ((__v8qi)__A, (__v8qi)__B);
}

static __inline __m64
_m_pminub (__m64 __A, __m64 __B)
{
  return _mm_min_pu8 (__A, __B);
}

/* Create an 8-bit mask of the signs of 8-bit values.  */
static __inline int
_mm_movemask_pi8 (__m64 __A)
{
  return __builtin_ia32_pmovmskb ((__v8qi)__A);
}

static __inline int
_m_pmovmskb (__m64 __A)
{
  return _mm_movemask_pi8 (__A);
}

/* Multiply four unsigned 16-bit values in A by four unsigned 16-bit values
   in B and produce the high 16 bits of the 32-bit results.  */
static __inline __m64
_mm_mulhi_pu16 (__m64 __A, __m64 __B)
{
  return (__m64) __builtin_ia32_pmulhuw ((__v4hi)__A, (__v4hi)__B);
}

static __inline __m64
_m_pmulhuw (__m64 __A, __m64 __B)
{
  return _mm_mulhi_pu16 (__A, __B);
}

/* Return a combination of the four 16-bit values in A.  The selector
   must be an immediate.  */
#if 0
static __inline __m64
_mm_shuffle_pi16 (__m64 __A, int __N)
{
  return (__m64) __builtin_ia32_pshufw ((__v4hi)__A, __N);
}

static __inline __m64
_m_pshufw (__m64 __A, int __N)
{
  return _mm_shuffle_pi16 (__A, __N);
}
#else
#define _mm_shuffle_pi16(A, N) \
  ((__m64) __builtin_ia32_pshufw ((__v4hi)(A), (N)))
#define _m_pshufw(A, N)		_mm_shuffle_pi16 ((A), (N))
#endif

/* Conditionally store byte elements of A into P.  The high bit of each
   byte in the selector N determines whether the corresponding byte from
   A is stored.  */
static __inline void
_mm_maskmove_si64 (__m64 __A, __m64 __N, char *__P)
{
  __builtin_ia32_maskmovq ((__v8qi)__A, (__v8qi)__N, __P);
}

static __inline void
_m_maskmovq (__m64 __A, __m64 __N, char *__P)
{
  _mm_maskmove_si64 (__A, __N, __P);
}

/* Compute the rounded averages of the unsigned 8-bit values in A and B.  */
static __inline __m64
_mm_avg_pu8 (__m64 __A, __m64 __B)
{
  return (__m64) __builtin_ia32_pavgb ((__v8qi)__A, (__v8qi)__B);
}

static __inline __m64
_m_pavgb (__m64 __A, __m64 __B)
{
  return _mm_avg_pu8 (__A, __B);
}

/* Compute the rounded averages of the unsigned 16-bit values in A and B.  */
static __inline __m64
_mm_avg_pu16 (__m64 __A, __m64 __B)
{
  return (__m64) __builtin_ia32_pavgw ((__v4hi)__A, (__v4hi)__B);
}

static __inline __m64
_m_pavgw (__m64 __A, __m64 __B)
{
  return _mm_avg_pu16 (__A, __B);
}

/* Compute the sum of the absolute differences of the unsigned 8-bit
   values in A and B.  Return the value in the lower 16-bit word; the
   upper words are cleared.  */
static __inline __m64
_mm_sad_pu8 (__m64 __A, __m64 __B)
{
  return (__m64) __builtin_ia32_psadbw ((__v8qi)__A, (__v8qi)__B);
}

static __inline __m64
_m_psadbw (__m64 __A, __m64 __B)
{
  return _mm_sad_pu8 (__A, __B);
}

/* Loads one cache line from address P to a location "closer" to the
   processor.  The selector I specifies the type of prefetch operation.  */
#if 0
static __inline void
_mm_prefetch (void *__P, enum _mm_hint __I)
{
  __builtin_prefetch (__P, 0, __I);
}
#else
#define _mm_prefetch(P, I) \
  __builtin_prefetch ((P), 0, (I))
#endif

/* Stores the data in A to the address P without polluting the caches.  */
static __inline void
_mm_stream_pi (__m64 *__P, __m64 __A)
{
  __builtin_ia32_movntq ((unsigned long long *)__P, (unsigned long long)__A);
}

/* Likewise.  The address must be 16-byte aligned.  */
static __inline void
_mm_stream_ps (float *__P, __m128 __A)
{
  __builtin_ia32_movntps (__P, (__v4sf)__A);
}

/* Guarantees that every preceding store is globally visible before
   any subsequent store.  */
static __inline void
_mm_sfence (void)
{
  __builtin_ia32_sfence ();
}

/* The execution of the next instruction is delayed by an implementation
   specific amount of time.  The instruction does not modify the
   architectural state.  */
static __inline void
_mm_pause (void)
{
  __asm__ __volatile__ ("rep; nop" : : );
}

/* Transpose the 4x4 matrix composed of row[0-3].  */
#define _MM_TRANSPOSE4_PS(row0, row1, row2, row3)			\
do {									\
  __v4sf __r0 = (row0), __r1 = (row1), __r2 = (row2), __r3 = (row3);	\
  __v4sf __t0 = __builtin_ia32_shufps (__r0, __r1, 0x44);		\
  __v4sf __t2 = __builtin_ia32_shufps (__r0, __r1, 0xEE);		\
  __v4sf __t1 = __builtin_ia32_shufps (__r2, __r3, 0x44);		\
  __v4sf __t3 = __builtin_ia32_shufps (__r2, __r3, 0xEE);		\
  (row0) = __builtin_ia32_shufps (__t0, __t1, 0x88);			\
  (row1) = __builtin_ia32_shufps (__t0, __t1, 0xDD);			\
  (row2) = __builtin_ia32_shufps (__t2, __t3, 0x88);			\
  (row3) = __builtin_ia32_shufps (__t2, __t3, 0xDD);			\
} while (0)

/* For backward source compatibility.  */
#include <emmintrin.h>

#endif /* __SSE__ */
#endif /* _XMMINTRIN_H_INCLUDED */