wide/
u16x16_.rs

1use super::*;
2
3pick! {
4  if #[cfg(target_feature="avx2")] {
5    #[derive(Default, Clone, Copy, PartialEq, Eq)]
6    #[repr(C, align(32))]
7    pub struct u16x16 { pub(crate) avx2: m256i }
8  } else {
9    #[derive(Default, Clone, Copy, PartialEq, Eq)]
10    #[repr(C, align(32))]
11    pub struct u16x16 { pub(crate) a : u16x8, pub(crate) b : u16x8 }
12  }
13}
14
15int_uint_consts!(u16, 16, u16x16, 256);
16
17unsafe impl Zeroable for u16x16 {}
18unsafe impl Pod for u16x16 {}
19
20impl Add for u16x16 {
21  type Output = Self;
22  #[inline]
23  #[must_use]
24  fn add(self, rhs: Self) -> Self::Output {
25    pick! {
26      if #[cfg(target_feature="avx2")] {
27        Self { avx2: add_i16_m256i(self.avx2, rhs.avx2) }
28      } else {
29        Self {
30          a : self.a.add(rhs.a),
31          b : self.b.add(rhs.b),
32        }
33      }
34    }
35  }
36}
37
38impl Sub for u16x16 {
39  type Output = Self;
40  #[inline]
41  #[must_use]
42  fn sub(self, rhs: Self) -> Self::Output {
43    pick! {
44      if #[cfg(target_feature="avx2")] {
45        Self { avx2: sub_i16_m256i(self.avx2, rhs.avx2) }
46      } else {
47        Self {
48          a : self.a.sub(rhs.a),
49          b : self.b.sub(rhs.b),
50        }
51      }
52    }
53  }
54}
55
56impl Add<u16> for u16x16 {
57  type Output = Self;
58  #[inline]
59  #[must_use]
60  fn add(self, rhs: u16) -> Self::Output {
61    self.add(Self::splat(rhs))
62  }
63}
64
65impl Sub<u16> for u16x16 {
66  type Output = Self;
67  #[inline]
68  #[must_use]
69  fn sub(self, rhs: u16) -> Self::Output {
70    self.sub(Self::splat(rhs))
71  }
72}
73
74impl Add<u16x16> for u16 {
75  type Output = u16x16;
76  #[inline]
77  #[must_use]
78  fn add(self, rhs: u16x16) -> Self::Output {
79    u16x16::splat(self).add(rhs)
80  }
81}
82
83impl Sub<u16x16> for u16 {
84  type Output = u16x16;
85  #[inline]
86  #[must_use]
87  fn sub(self, rhs: u16x16) -> Self::Output {
88    u16x16::splat(self).sub(rhs)
89  }
90}
91
92impl BitAnd for u16x16 {
93  type Output = Self;
94  #[inline]
95  #[must_use]
96  fn bitand(self, rhs: Self) -> Self::Output {
97    pick! {
98      if #[cfg(target_feature="avx2")] {
99        Self { avx2: bitand_m256i(self.avx2, rhs.avx2) }
100      } else {
101        Self {
102          a : self.a.bitand(rhs.a),
103          b : self.b.bitand(rhs.b),
104        }
105      }
106    }
107  }
108}
109
110impl BitOr for u16x16 {
111  type Output = Self;
112  #[inline]
113  #[must_use]
114  fn bitor(self, rhs: Self) -> Self::Output {
115    pick! {
116      if #[cfg(target_feature="avx2")] {
117        Self { avx2: bitor_m256i(self.avx2, rhs.avx2) }
118      } else {
119        Self {
120          a : self.a.bitor(rhs.a),
121          b : self.b.bitor(rhs.b),
122        }
123      }
124    }
125  }
126}
127
128impl BitXor for u16x16 {
129  type Output = Self;
130  #[inline]
131  #[must_use]
132  fn bitxor(self, rhs: Self) -> Self::Output {
133    pick! {
134      if #[cfg(target_feature="avx2")] {
135        Self { avx2: bitxor_m256i(self.avx2, rhs.avx2) }
136      } else {
137        Self {
138          a : self.a.bitxor(rhs.a),
139          b : self.b.bitxor(rhs.b),
140        }
141      }
142    }
143  }
144}
145
146impl Not for u16x16 {
147  type Output = Self;
148  #[inline]
149  fn not(self) -> Self {
150    pick! {
151      if #[cfg(target_feature="avx2")] {
152        Self { avx2: self.avx2.not()  }
153      } else {
154        Self {
155          a : self.a.not(),
156          b : self.b.not(),
157        }
158      }
159    }
160  }
161}
162
163macro_rules! impl_shl_t_for_u16x16 {
164  ($($shift_type:ty),+ $(,)?) => {
165    $(impl Shl<$shift_type> for u16x16 {
166      type Output = Self;
167      /// Shifts all lanes by the value given.
168      #[inline]
169      #[must_use]
170      fn shl(self, rhs: $shift_type) -> Self::Output {
171        pick! {
172          if #[cfg(target_feature="avx2")] {
173            let shift = cast([rhs as u64, 0]);
174            Self { avx2: shl_all_u16_m256i(self.avx2, shift) }
175          } else {
176            Self {
177              a : self.a.shl(rhs),
178              b : self.b.shl(rhs),
179            }
180          }
181       }
182     }
183    })+
184  };
185}
186impl_shl_t_for_u16x16!(i8, u8, i16, u16, i32, u32, i64, u64, i128, u128);
187
188macro_rules! impl_shr_t_for_u16x16 {
189  ($($shift_type:ty),+ $(,)?) => {
190    $(impl Shr<$shift_type> for u16x16 {
191      type Output = Self;
192      /// Shifts all lanes by the value given.
193      #[inline]
194      #[must_use]
195      fn shr(self, rhs: $shift_type) -> Self::Output {
196        pick! {
197          if #[cfg(target_feature="avx2")] {
198            let shift = cast([rhs as u64, 0]);
199            Self { avx2: shr_all_u16_m256i(self.avx2, shift) }
200          } else {
201            Self {
202              a : self.a.shr(rhs),
203              b : self.b.shr(rhs),
204            }
205          }
206        }
207      }
208    })+
209  };
210}
211impl_shr_t_for_u16x16!(i8, u8, i16, u16, i32, u32, i64, u64, i128, u128);
212
213impl CmpEq for u16x16 {
214  type Output = Self;
215  #[inline]
216  #[must_use]
217  fn cmp_eq(self, rhs: Self) -> Self::Output {
218    pick! {
219      if #[cfg(target_feature="avx2")] {
220        Self { avx2: cmp_eq_mask_i16_m256i(self.avx2, rhs.avx2) }
221      } else {
222        Self {
223          a : self.a.cmp_eq(rhs.a),
224          b : self.b.cmp_eq(rhs.b),
225        }
226      }
227    }
228  }
229}
230
231impl Mul for u16x16 {
232  type Output = Self;
233  #[inline]
234  #[must_use]
235  fn mul(self, rhs: Self) -> Self::Output {
236    pick! {
237      if #[cfg(target_feature="avx2")] {
238        // non-widening multiplication is the same for unsigned and signed
239        Self { avx2: mul_i16_keep_low_m256i(self.avx2, rhs.avx2) }
240      } else {
241        Self {
242          a : self.a.mul(rhs.a),
243          b : self.b.mul(rhs.b),
244        }
245      }
246    }
247  }
248}
249
250impl From<u8x16> for u16x16 {
251  /// widens and sign extends to u16x16
252  #[inline]
253  #[must_use]
254  fn from(v: u8x16) -> Self {
255    pick! {
256      if #[cfg(target_feature="avx2")] {
257        u16x16 { avx2:convert_to_i16_m256i_from_u8_m128i(v.sse) }
258      } else if #[cfg(target_feature="sse2")] {
259        u16x16 {
260          a: u16x8 { sse: shr_imm_u16_m128i::<8>( unpack_low_i8_m128i(v.sse, v.sse)) },
261          b: u16x8 { sse: shr_imm_u16_m128i::<8>( unpack_high_i8_m128i(v.sse, v.sse)) },
262        }
263      } else {
264
265        u16x16::new([
266          v.as_array_ref()[0] as u16,
267          v.as_array_ref()[1] as u16,
268          v.as_array_ref()[2] as u16,
269          v.as_array_ref()[3] as u16,
270          v.as_array_ref()[4] as u16,
271          v.as_array_ref()[5] as u16,
272          v.as_array_ref()[6] as u16,
273          v.as_array_ref()[7] as u16,
274          v.as_array_ref()[8] as u16,
275          v.as_array_ref()[9] as u16,
276          v.as_array_ref()[10] as u16,
277          v.as_array_ref()[11] as u16,
278          v.as_array_ref()[12] as u16,
279          v.as_array_ref()[13] as u16,
280          v.as_array_ref()[14] as u16,
281          v.as_array_ref()[15] as u16,
282          ])
283      }
284    }
285  }
286}
287
288impl u16x16 {
289  #[inline]
290  #[must_use]
291  pub const fn new(array: [u16; 16]) -> Self {
292    unsafe { core::mem::transmute(array) }
293  }
294
295  #[inline]
296  #[must_use]
297  pub fn blend(self, t: Self, f: Self) -> Self {
298    pick! {
299      if #[cfg(target_feature="avx2")] {
300        Self { avx2: blend_varying_i8_m256i(f.avx2, t.avx2, self.avx2) }
301      } else {
302        Self {
303          a : self.a.blend(t.a, f.a),
304          b : self.b.blend(t.b, f.b),
305        }
306      }
307    }
308  }
309
310  #[inline]
311  #[must_use]
312  pub fn max(self, rhs: Self) -> Self {
313    pick! {
314      if #[cfg(target_feature="avx2")] {
315        Self { avx2: max_u16_m256i(self.avx2, rhs.avx2) }
316      } else {
317        Self {
318          a : self.a.max(rhs.a),
319          b : self.b.max(rhs.b),
320        }
321      }
322    }
323  }
324  #[inline]
325  #[must_use]
326  pub fn min(self, rhs: Self) -> Self {
327    pick! {
328      if #[cfg(target_feature="avx2")] {
329        Self { avx2: min_u16_m256i(self.avx2, rhs.avx2) }
330      } else {
331        Self {
332          a : self.a.min(rhs.a),
333          b : self.b.min(rhs.b),
334        }
335      }
336    }
337  }
338
339  #[inline]
340  #[must_use]
341  pub fn saturating_add(self, rhs: Self) -> Self {
342    pick! {
343      if #[cfg(target_feature="avx2")] {
344        Self { avx2: add_saturating_u16_m256i(self.avx2, rhs.avx2) }
345      } else {
346        Self {
347          a : self.a.saturating_add(rhs.a),
348          b : self.b.saturating_add(rhs.b),
349        }
350      }
351    }
352  }
353  #[inline]
354  #[must_use]
355  pub fn saturating_sub(self, rhs: Self) -> Self {
356    pick! {
357      if #[cfg(target_feature="avx2")] {
358        Self { avx2: sub_saturating_u16_m256i(self.avx2, rhs.avx2) }
359      } else {
360        Self {
361          a : self.a.saturating_sub(rhs.a),
362          b : self.b.saturating_sub(rhs.b),
363        }
364      }
365    }
366  }
367
368  #[inline]
369  pub fn to_array(self) -> [u16; 16] {
370    cast(self)
371  }
372
373  #[inline]
374  pub fn as_array_ref(&self) -> &[u16; 16] {
375    cast_ref(self)
376  }
377
378  #[inline]
379  pub fn as_array_mut(&mut self) -> &mut [u16; 16] {
380    cast_mut(self)
381  }
382}