1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140

#![cfg(target_feature="sse")]
#![cfg(target_feature="sse2")]
#![cfg(target_feature="sse3")]

use super::*;

/// # SSE3 Operations
impl m128 {
  /// Adds odd lanes (3 and 1) and subtracts even lanes (2 and 0).
  ///
  /// ```txt
  /// out[0]= self[0] - rhs[0]
  /// out[1]= self[1] + rhs[1]
  /// out[2]= self[2] - rhs[2]
  /// out[3]= self[3] + rhs[3]
  /// ```
  #[inline(always)]
  pub fn add_sub(self, rhs: Self) -> Self {
    Self(unsafe { _mm_addsub_ps(self.0, rhs.0) })
  }

  /// Horizontal add both `self` and `rhs`, then pack together.
  ///
  /// ```txt
  /// out[0]= self[0] + self[1]
  /// out[1]= self[2] + self[3]
  /// out[2]= rhs[0] + rhs[1]
  /// out[3]= rhs[2] + rhs[3]
  /// ```
  #[inline(always)]
  pub fn horizontal_add(self, rhs: Self) -> Self {
    Self(unsafe { _mm_hadd_ps(self.0, rhs.0) })
  }

  /// Horizontal subtract both `self` and `rhs`, then pack together.
  ///
  /// ```txt
  /// out[0]= self[0] - self[1]
  /// out[1]= self[2] - self[3]
  /// out[2]= rhs[0] - rhs[1]
  /// out[3]= rhs[2] - rhs[3]
  /// ```
  #[inline(always)]
  pub fn horizontal_sub(self, rhs: Self) -> Self {
    Self(unsafe { _mm_hsub_ps(self.0, rhs.0) })
  }

  /// Duplicate odd indexed lanes into a new `m128`.
  ///
  /// ```txt
  /// out[0]= self[1]
  /// out[1]= self[1]
  /// out[2]= self[3]
  /// out[3]= self[3]
  /// ```
  #[inline(always)]
  pub fn duplicate_odd(self) -> Self {
    Self(unsafe { _mm_movehdup_ps(self.0) })
  }

  /// Duplicate even indexed lanes into a new `m128`.
  ///
  /// ```txt
  /// out[0]= self[0]
  /// out[1]= self[0]
  /// out[2]= self[2]
  /// out[3]= self[2]
  /// ```
  #[inline(always)]
  pub fn duplicate_even(self) -> Self {
    Self(unsafe { _mm_moveldup_ps(self.0) })
  }
}

/// # SSE3 Operations
impl m128d {
  /// Adds the high lane (1) and subtracts the low lane (0).
  ///
  /// ```txt
  /// out[0]= self[0] - rhs[0]
  /// out[1]= self[1] + rhs[1]
  /// ```
  #[inline(always)]
  pub fn add_sub(self, rhs: Self) -> Self {
    Self(unsafe { _mm_addsub_pd(self.0, rhs.0) })
  }

  /// Horizontal add both `self` and `rhs`, then pack together.
  ///
  /// ```txt
  /// out[0]= self[0] + self[1]
  /// out[1]= rhs[0] + rhs[1]
  /// ```
  #[inline(always)]
  pub fn horizontal_add(self, rhs: Self) -> Self {
    Self(unsafe { _mm_hadd_pd(self.0, rhs.0) })
  }

  /// Horizontal subtract both `self` and `rhs`, then pack together.
  ///
  /// ```txt
  /// out[0]= self[0] - self[1]
  /// out[1]= rhs[0] - rhs[1]
  /// ```
  #[inline(always)]
  pub fn horizontal_sub(self, rhs: Self) -> Self {
    Self(unsafe { _mm_hsub_pd(self.0, rhs.0) })
  }

  /// Load the given `f64` address, duplicating it into both lanes.
  #[inline(always)]
  #[allow(clippy::trivially_copy_pass_by_ref)]
  pub fn load_splat(addr: &f64) -> Self {
    Self(unsafe { _mm_loaddup_pd(addr) })
  }

  /// Duplicate the low lane of `self` into both lanes of a new `m128d`.
  ///
  /// ```txt
  /// out[0]= self[0]
  /// out[1]= self[0]
  /// ```
  #[inline(always)]
  pub fn duplicate_low(self) -> Self {
    Self(unsafe { _mm_movedup_pd(self.0) })
  }
}

/// # SSE3 Operations
impl m128i {
  /// Loads 128-bits of integer data without alignment requirements.
  ///
  /// This can perform faster than [`m128i::load_unaligned`] if the data would
  /// cross a cache line boundary.
  #[inline(always)]
  pub fn load_quick_unaligned(addr: *const i128) -> Self {
    #[allow(clippy::cast_ptr_alignment)]
    Self(unsafe { _mm_lddqu_si128(addr as *const _) })
  }
}