xmmintrin.h

mingw32.rar

static __inline int
_mm_comile_ss (__m128 __A, __m128 __B)
{
  return __builtin_ia32_comile ((__v4sf)__A, (__v4sf)__B);
}

static __inline int
_mm_comigt_ss (__m128 __A, __m128 __B)
{
  return __builtin_ia32_comigt ((__v4sf)__A, (__v4sf)__B);
}

static __inline int
_mm_comige_ss (__m128 __A, __m128 __B)
{
  return __builtin_ia32_comige ((__v4sf)__A, (__v4sf)__B);
}

static __inline int
_mm_comineq_ss (__m128 __A, __m128 __B)
{
  return __builtin_ia32_comineq ((__v4sf)__A, (__v4sf)__B);
}

static __inline int
_mm_ucomieq_ss (__m128 __A, __m128 __B)
{
  return __builtin_ia32_ucomieq ((__v4sf)__A, (__v4sf)__B);
}

static __inline int
_mm_ucomilt_ss (__m128 __A, __m128 __B)
{
  return __builtin_ia32_ucomilt ((__v4sf)__A, (__v4sf)__B);
}

static __inline int
_mm_ucomile_ss (__m128 __A, __m128 __B)
{
  return __builtin_ia32_ucomile ((__v4sf)__A, (__v4sf)__B);
}

static __inline int
_mm_ucomigt_ss (__m128 __A, __m128 __B)
{
  return __builtin_ia32_ucomigt ((__v4sf)__A, (__v4sf)__B);
}

static __inline int
_mm_ucomige_ss (__m128 __A, __m128 __B)
{
  return __builtin_ia32_ucomige ((__v4sf)__A, (__v4sf)__B);
}

static __inline int
_mm_ucomineq_ss (__m128 __A, __m128 __B)
{
  return __builtin_ia32_ucomineq ((__v4sf)__A, (__v4sf)__B);
}

/* Convert the lower SPFP value to a 32-bit integer according to the current
   rounding mode.  */
static __inline int
_mm_cvtss_si32 (__m128 __A)
{
  return __builtin_ia32_cvtss2si ((__v4sf) __A);
}

static __inline int
_mm_cvt_ss2si (__m128 __A)
{
  return _mm_cvtss_si32 (__A);
}

#ifdef __x86_64__
/* Convert the lower SPFP value to a 32-bit integer according to the current
   rounding mode.  */
static __inline long long
_mm_cvtss_si64x (__m128 __A)
{
  return __builtin_ia32_cvtss2si64 ((__v4sf) __A);
}
#endif

/* Convert the two lower SPFP values to 32-bit integers according to the
   current rounding mode.  Return the integers in packed form.  */
static __inline __m64
_mm_cvtps_pi32 (__m128 __A)
{
  return (__m64) __builtin_ia32_cvtps2pi ((__v4sf) __A);
}

static __inline __m64
_mm_cvt_ps2pi (__m128 __A)
{
  return _mm_cvtps_pi32 (__A);
}

/* Truncate the lower SPFP value to a 32-bit integer.  */
static __inline int
_mm_cvttss_si32 (__m128 __A)
{
  return __builtin_ia32_cvttss2si ((__v4sf) __A);
}

static __inline int
_mm_cvtt_ss2si (__m128 __A)
{
  return _mm_cvttss_si32 (__A);
}

#ifdef __x86_64__
/* Truncate the lower SPFP value to a 32-bit integer.  */
static __inline long long
_mm_cvttss_si64x (__m128 __A)
{
  return __builtin_ia32_cvttss2si64 ((__v4sf) __A);
}
#endif

/* Truncate the two lower SPFP values to 32-bit integers.  Return the
   integers in packed form.  */
static __inline __m64
_mm_cvttps_pi32 (__m128 __A)
{
  return (__m64) __builtin_ia32_cvttps2pi ((__v4sf) __A);
}

static __inline __m64
_mm_cvtt_ps2pi (__m128 __A)
{
  return _mm_cvttps_pi32 (__A);
}

/* Convert B to a SPFP value and insert it as element zero in A.  */
static __inline __m128
_mm_cvtsi32_ss (__m128 __A, int __B)
{
  return (__m128) __builtin_ia32_cvtsi2ss ((__v4sf) __A, __B);
}

static __inline __m128
_mm_cvt_si2ss (__m128 __A, int __B)
{
  return _mm_cvtsi32_ss (__A, __B);
}

#ifdef __x86_64__
/* Convert B to a SPFP value and insert it as element zero in A.  */
static __inline __m128
_mm_cvtsi64x_ss (__m128 __A, long long __B)
{
  return (__m128) __builtin_ia32_cvtsi642ss ((__v4sf) __A, __B);
}
#endif

/* Convert the two 32-bit values in B to SPFP form and insert them
   as the two lower elements in A.  */
static __inline __m128
_mm_cvtpi32_ps (__m128 __A, __m64 __B)
{
  return (__m128) __builtin_ia32_cvtpi2ps ((__v4sf) __A, (__v2si)__B);
}

static __inline __m128
_mm_cvt_pi2ps (__m128 __A, __m64 __B)
{
  return _mm_cvtpi32_ps (__A, __B);
}

/* Convert the four signed 16-bit values in A to SPFP form.  */
static __inline __m128
_mm_cvtpi16_ps (__m64 __A)
{
  __v4hi __sign;
  __v2si __hisi, __losi;
  __v4sf __r;

  /* This comparison against zero gives us a mask that can be used to
     fill in the missing sign bits in the unpack operations below, so
     that we get signed values after unpacking.  */
  __sign = (__v4hi) __builtin_ia32_mmx_zero ();
  __sign = __builtin_ia32_pcmpgtw (__sign, (__v4hi)__A);

  /* Convert the four words to doublewords.  */
  __hisi = (__v2si) __builtin_ia32_punpckhwd ((__v4hi)__A, __sign);
  __losi = (__v2si) __builtin_ia32_punpcklwd ((__v4hi)__A, __sign);

  /* Convert the doublewords to floating point two at a time.  */
  __r = (__v4sf) __builtin_ia32_setzerops ();
  __r = __builtin_ia32_cvtpi2ps (__r, __hisi);
  __r = __builtin_ia32_movlhps (__r, __r);
  __r = __builtin_ia32_cvtpi2ps (__r, __losi);

  return (__m128) __r;
}

/* Convert the four unsigned 16-bit values in A to SPFP form.  */
static __inline __m128
_mm_cvtpu16_ps (__m64 __A)
{
  __v4hi __zero = (__v4hi) __builtin_ia32_mmx_zero ();
  __v2si __hisi, __losi;
  __v4sf __r;

  /* Convert the four words to doublewords.  */
  __hisi = (__v2si) __builtin_ia32_punpckhwd ((__v4hi)__A, __zero);
  __losi = (__v2si) __builtin_ia32_punpcklwd ((__v4hi)__A, __zero);

  /* Convert the doublewords to floating point two at a time.  */
  __r = (__v4sf) __builtin_ia32_setzerops ();
  __r = __builtin_ia32_cvtpi2ps (__r, __hisi);
  __r = __builtin_ia32_movlhps (__r, __r);
  __r = __builtin_ia32_cvtpi2ps (__r, __losi);

  return (__m128) __r;
}

/* Convert the low four signed 8-bit values in A to SPFP form.  */
static __inline __m128
_mm_cvtpi8_ps (__m64 __A)
{
  __v8qi __sign;

  /* This comparison against zero gives us a mask that can be used to
     fill in the missing sign bits in the unpack operations below, so
     that we get signed values after unpacking.  */
  __sign = (__v8qi) __builtin_ia32_mmx_zero ();
  __sign = __builtin_ia32_pcmpgtb (__sign, (__v8qi)__A);

  /* Convert the four low bytes to words.  */
  __A = (__m64) __builtin_ia32_punpcklbw ((__v8qi)__A, __sign);

  return _mm_cvtpi16_ps(__A);
}

/* Convert the low four unsigned 8-bit values in A to SPFP form.  */
static __inline __m128
_mm_cvtpu8_ps(__m64 __A)
{
  __v8qi __zero = (__v8qi) __builtin_ia32_mmx_zero ();
  __A = (__m64) __builtin_ia32_punpcklbw ((__v8qi)__A, __zero);
  return _mm_cvtpu16_ps(__A);
}

/* Convert the four signed 32-bit values in A and B to SPFP form.  */
static __inline __m128
_mm_cvtpi32x2_ps(__m64 __A, __m64 __B)
{
  __v4sf __zero = (__v4sf) __builtin_ia32_setzerops ();
  __v4sf __sfa = __builtin_ia32_cvtpi2ps (__zero, (__v2si)__A);
  __v4sf __sfb = __builtin_ia32_cvtpi2ps (__zero, (__v2si)__B);
  return (__m128) __builtin_ia32_movlhps (__sfa, __sfb);
}

/* Convert the four SPFP values in A to four signed 16-bit integers.  */
static __inline __m64
_mm_cvtps_pi16(__m128 __A)
{
  __v4sf __hisf = (__v4sf)__A;
  __v4sf __losf = __builtin_ia32_movhlps (__hisf, __hisf);
  __v2si __hisi = __builtin_ia32_cvtps2pi (__hisf);
  __v2si __losi = __builtin_ia32_cvtps2pi (__losf);
  return (__m64) __builtin_ia32_packssdw (__hisi, __losi);
}

/* Convert the four SPFP values in A to four signed 8-bit integers.  */
static __inline __m64
_mm_cvtps_pi8(__m128 __A)
{
  __v4hi __tmp = (__v4hi) _mm_cvtps_pi16 (__A);
  __v4hi __zero = (__v4hi) __builtin_ia32_mmx_zero ();
  return (__m64) __builtin_ia32_packsswb (__tmp, __zero);
}

/* Selects four specific SPFP values from A and B based on MASK.  */
#if 0
static __inline __m128
_mm_shuffle_ps (__m128 __A, __m128 __B, int __mask)
{
  return (__m128) __builtin_ia32_shufps ((__v4sf)__A, (__v4sf)__B, __mask);
}
#else
#define _mm_shuffle_ps(A, B, MASK) \
 ((__m128) __builtin_ia32_shufps ((__v4sf)(A), (__v4sf)(B), (MASK)))
#endif


/* Selects and interleaves the upper two SPFP values from A and B.  */
static __inline __m128
_mm_unpackhi_ps (__m128 __A, __m128 __B)
{
  return (__m128) __builtin_ia32_unpckhps ((__v4sf)__A, (__v4sf)__B);
}

/* Selects and interleaves the lower two SPFP values from A and B.  */
static __inline __m128
_mm_unpacklo_ps (__m128 __A, __m128 __B)
{
  return (__m128) __builtin_ia32_unpcklps ((__v4sf)__A, (__v4sf)__B);
}

/* Sets the upper two SPFP values with 64-bits of data loaded from P;
   the lower two values are passed through from A.  */
static __inline __m128
_mm_loadh_pi (__m128 __A, __m64 const *__P)
{
  return (__m128) __builtin_ia32_loadhps ((__v4sf)__A, (__v2si *)__P);
}

/* Stores the upper two SPFP values of A into P.  */
static __inline void
_mm_storeh_pi (__m64 *__P, __m128 __A)
{
  __builtin_ia32_storehps ((__v2si *)__P, (__v4sf)__A);
}

/* Moves the upper two values of B into the lower two values of A.  */
static __inline __m128
_mm_movehl_ps (__m128 __A, __m128 __B)
{
  return (__m128) __builtin_ia32_movhlps ((__v4sf)__A, (__v4sf)__B);
}

/* Moves the lower two values of B into the upper two values of A.  */
static __inline __m128
_mm_movelh_ps (__m128 __A, __m128 __B)
{
  return (__m128) __builtin_ia32_movlhps ((__v4sf)__A, (__v4sf)__B);
}

/* Sets the lower two SPFP values with 64-bits of data loaded from P;
   the upper two values are passed through from A.  */
static __inline __m128
_mm_loadl_pi (__m128 __A, __m64 const *__P)
{
  return (__m128) __builtin_ia32_loadlps ((__v4sf)__A, (__v2si *)__P);
}

/* Stores the lower two SPFP values of A into P.  */
static __inline void
_mm_storel_pi (__m64 *__P, __m128 __A)
{
  __builtin_ia32_storelps ((__v2si *)__P, (__v4sf)__A);
}

/* Creates a 4-bit mask from the most significant bits of the SPFP values.  */
static __inline int
_mm_movemask_ps (__m128 __A)
{
  return __builtin_ia32_movmskps ((__v4sf)__A);
}

/* Return the contents of the control register.  */
static __inline unsigned int
_mm_getcsr (void)
{
  return __builtin_ia32_stmxcsr ();
}

/* Read exception bits from the control register.  */
static __inline unsigned int
_MM_GET_EXCEPTION_STATE (void)
{
  return _mm_getcsr() & _MM_EXCEPT_MASK;
}

static __inline unsigned int
_MM_GET_EXCEPTION_MASK (void)
{
  return _mm_getcsr() & _MM_MASK_MASK;
}

static __inline unsigned int
_MM_GET_ROUNDING_MODE (void)
{
  return _mm_getcsr() & _MM_ROUND_MASK;
}

static __inline unsigned int
_MM_GET_FLUSH_ZERO_MODE (void)
{
  return _mm_getcsr() & _MM_FLUSH_ZERO_MASK;
}

/* Set the control register to I.  */
static __inline void
_mm_setcsr (unsigned int __I)
{
  __builtin_ia32_ldmxcsr (__I);
}

/* Set exception bits in the control register.  */
static __inline void
_MM_SET_EXCEPTION_STATE(unsigned int __mask)
{
  _mm_setcsr((_mm_getcsr() & ~_MM_EXCEPT_MASK) | __mask);
}

static __inline void
_MM_SET_EXCEPTION_MASK (unsigned int __mask)
{
  _mm_setcsr((_mm_getcsr() & ~_MM_MASK_MASK) | __mask);
}

static __inline void