wide/
lib.rs

Help
1#![no_std]
2#![allow(non_camel_case_types)]
3#![warn(clippy::doc_markdown)]
4#![warn(clippy::missing_inline_in_public_items)]
5#![allow(clippy::eq_op)]
6#![allow(clippy::excessive_precision)]
7#![allow(clippy::let_and_return)]
8#![allow(clippy::unusual_byte_groupings)]
9#![allow(clippy::misrefactored_assign_op)]
10#![cfg_attr(test, allow(clippy::approx_constant))]
11
12//! A crate to help you go wide.
13//!
14//! This crate provides SIMD-compatible data types.
15//!
16//! When possible, explicit SIMD is used with all the math operations here. As a
17//! fallback, the fact that all the lengths of a fixed length array are doing
18//! the same thing will often make LLVM notice that it should use SIMD
19//! instructions to complete the task. In the worst case, the code just becomes
20//! totally scalar (though the math is still correct, at least).
21//!
22//! ## Crate Features
23//!
24//! * `std`: This causes the feature to link to `std`.
25//!   * Currently this just improves the performance of `sqrt` when an explicit
26//!     SIMD `sqrt` isn't available.
27
28// Note(Lokathor): Due to standard library magic, the std-only methods for f32
29// and f64 will automatically be available simply by declaring this.
30#[cfg(feature = "std")]
31extern crate std;
32
33// TODO
34// Add/Sub/Mul/Div with constant
35// Shuffle left/right/by index
36
37use core::{
38  fmt::{
39    Binary, Debug, Display, LowerExp, LowerHex, Octal, UpperExp, UpperHex,
40  },
41  ops::*,
42};
43
44#[allow(unused_imports)]
45use safe_arch::*;
46
47use bytemuck::*;
48
49#[cfg(feature = "serde")]
50use serde::{ser::SerializeTuple, Deserialize, Serialize};
51
52#[macro_use]
53mod macros;
54
55macro_rules! pick {
56  ($(if #[cfg($($test:meta),*)] {
57      $($if_tokens:tt)*
58    })else+ else {
59      $($else_tokens:tt)*
60    }) => {
61    pick!{
62      @__forests [ ] ;
63      $( [ {$($test),*} {$($if_tokens)*} ], )*
64      [ { } {$($else_tokens)*} ],
65    }
66  };
67  (if #[cfg($($if_meta:meta),*)] {
68      $($if_tokens:tt)*
69    } $(else if #[cfg($($else_meta:meta),*)] {
70      $($else_tokens:tt)*
71    })*) => {
72    pick!{
73      @__forests [ ] ;
74      [ {$($if_meta),*} {$($if_tokens)*} ],
75      $( [ {$($else_meta),*} {$($else_tokens)*} ], )*
76    }
77  };
78  (@__forests [$($not:meta,)*];) => {
79    /* halt expansion */
80  };
81  (@__forests [$($not:meta,)*]; [{$($m:meta),*} {$($tokens:tt)*}], $($rest:tt)*) => {
82    #[cfg(all( $($m,)* not(any($($not),*)) ))]
83    pick!{ @__identity $($tokens)* }
84    pick!{ @__forests [ $($not,)* $($m,)* ] ; $($rest)* }
85  };
86  (@__identity $($tokens:tt)*) => {
87    $($tokens)*
88  };
89}
90
91// TODO: make these generic over `mul_add`? Worth it?
92
93macro_rules! polynomial_2 {
94  ($x:expr, $c0:expr, $c1:expr, $c2:expr $(,)?) => {{
95    let x = $x;
96    let x2 = x * x;
97    x2.mul_add($c2, x.mul_add($c1, $c0))
98  }};
99}
100
101macro_rules! polynomial_3 {
102  ($x:expr, $c0:expr, $c1:expr, $c2:expr, $c3:expr $(,)?) => {{
103    let x = $x;
104    let x2 = x * x;
105    $c3.mul_add(x, $c2).mul_add(x2, $c1.mul_add(x, $c0))
106  }};
107}
108
109macro_rules! polynomial_4 {
110  ($x:expr, $c0:expr, $c1:expr, $c2:expr ,$c3:expr, $c4:expr $(,)?) => {{
111    let x = $x;
112    let x2 = x * x;
113    let x4 = x2 * x2;
114    $c3.mul_add(x, $c2).mul_add(x2, $c1.mul_add(x, $c0)) + $c4 * x4
115  }};
116}
117
118macro_rules! polynomial_5 {
119  ($x:expr, $c0:expr, $c1:expr, $c2:expr, $c3:expr, $c4:expr, $c5:expr $(,)?) => {{
120    let x = $x;
121    let x2 = x * x;
122    let x4 = x2 * x2;
123    $c3
124      .mul_add(x, $c2)
125      .mul_add(x2, $c5.mul_add(x, $c4).mul_add(x4, $c1.mul_add(x, $c0)))
126  }};
127}
128
129macro_rules! polynomial_5n {
130  ($x:expr, $c0:expr, $c1:expr, $c2:expr, $c3:expr, $c4:expr $(,)?) => {{
131    let x = $x;
132    let x2 = x * x;
133    let x4 = x2 * x2;
134    x2.mul_add(x.mul_add($c3, $c2), (x4.mul_add($c4 + x, x.mul_add($c1, $c0))))
135  }};
136}
137
138macro_rules! polynomial_6 {
139  ($x:expr, $c0:expr, $c1:expr, $c2:expr, $c3:expr, $c4:expr, $c5:expr ,$c6:expr $(,)?) => {{
140    let x = $x;
141    let x2 = x * x;
142    let x4 = x2 * x2;
143    x4.mul_add(
144      x2.mul_add($c6, x.mul_add($c5, $c4)),
145      x2.mul_add(x.mul_add($c3, $c2), x.mul_add($c1, $c0)),
146    )
147  }};
148}
149
150macro_rules! polynomial_6n {
151  ($x:expr, $c0:expr, $c1:expr, $c2:expr, $c3:expr, $c4:expr, $c5:expr $(,)?) => {{
152    let x = $x;
153    let x2 = x * x;
154    let x4 = x2 * x2;
155    x4.mul_add(
156      x.mul_add($c5, x2 + $c4),
157      x2.mul_add(x.mul_add($c3, $c2), x.mul_add($c1, $c0)),
158    )
159  }};
160}
161
162macro_rules! polynomial_8 {
163  ($x:expr, $c0:expr, $c1:expr, $c2:expr, $c3:expr, $c4:expr, $c5:expr,  $c6:expr, $c7:expr, $c8:expr $(,)?) => {{
164    let x = $x;
165    let x2 = x * x;
166    let x4 = x2 * x2;
167    let x8 = x4 * x4;
168    x4.mul_add(
169      x2.mul_add($c7.mul_add(x, $c6), x.mul_add($c5, $c4)),
170      x8.mul_add($c8, x2.mul_add(x.mul_add($c3, $c2), x.mul_add($c1, $c0))),
171    )
172  }};
173}
174
175macro_rules! polynomial_13 {
176  // calculates polynomial c13*x^13 + c12*x^12 + ... + c1*x + c0
177  ($x:expr,  $c2:expr, $c3:expr, $c4:expr, $c5:expr,$c6:expr, $c7:expr, $c8:expr,$c9:expr, $c10:expr, $c11:expr, $c12:expr, $c13:expr  $(,)?) => {{
178    let x = $x;
179    let x2 = x * x;
180    let x4 = x2 * x2;
181    let x8 = x4 * x4;
182    x8.mul_add(
183      x4.mul_add(
184        x.mul_add($c13, $c12),
185        x2.mul_add(x.mul_add($c11, $c10), x.mul_add($c9, $c8)),
186      ),
187      x4.mul_add(
188        x2.mul_add(x.mul_add($c7, $c6), x.mul_add($c5, $c4)),
189        x2.mul_add(x.mul_add($c3, $c2), x),
190      ),
191    )
192  }};
193}
194
195macro_rules! polynomial_13m {
196  // return  ((c8+c9*x) + (c10+c11*x)*x2 + (c12+c13*x)*x4)*x8 + (((c6+c7*x)*x2 +
197  // (c4+c5*x))*x4 + ((c2+c3*x)*x2 + x));
198  ($x:expr,  $c2:expr, $c3:expr, $c4:expr, $c5:expr,$c6:expr, $c7:expr, $c8:expr,$c9:expr, $c10:expr, $c11:expr, $c12:expr, $c13:expr  $(,)?) => {{
199    let x = $x;
200    let x2 = x * x;
201    let x4 = x2 * x2;
202    let x8 = x4 * x4;
203
204    x8.mul_add(
205      x4.mul_add(
206        x.mul_add($c13, $c12),
207        x2.mul_add(x.mul_add($c11, $c10), x.mul_add($c9, $c8)),
208      ),
209      x4.mul_add(
210        x2.mul_add(x.mul_add($c7, $c6), x.mul_add($c5, $c4)),
211        x2.mul_add(x.mul_add($c3, $c2), x),
212      ),
213    )
214  }};
215}
216
217mod f32x8_;
218pub use f32x8_::*;
219
220mod f32x4_;
221pub use f32x4_::*;
222
223mod f64x4_;
224pub use f64x4_::*;
225
226mod f64x2_;
227pub use f64x2_::*;
228
229mod i8x16_;
230pub use i8x16_::*;
231
232mod i16x16_;
233pub use i16x16_::*;
234
235mod i8x32_;
236pub use i8x32_::*;
237
238mod i16x8_;
239pub use i16x8_::*;
240
241mod i32x4_;
242pub use i32x4_::*;
243
244mod i32x8_;
245pub use i32x8_::*;
246
247mod i64x2_;
248pub use i64x2_::*;
249
250mod i64x4_;
251pub use i64x4_::*;
252
253mod u8x16_;
254pub use u8x16_::*;
255
256mod u16x8_;
257pub use u16x8_::*;
258
259mod u16x16_;
260pub use u16x16_::*;
261
262mod u32x4_;
263pub use u32x4_::*;
264
265mod u32x8_;
266pub use u32x8_::*;
267
268mod u64x2_;
269pub use u64x2_::*;
270
271mod u64x4_;
272pub use u64x4_::*;
273
274#[allow(dead_code)]
275fn generic_bit_blend<T>(mask: T, y: T, n: T) -> T
276where
277  T: Copy + BitXor<Output = T> + BitAnd<Output = T>,
278{
279  n ^ ((n ^ y) & mask)
280}
281
282/// given `type.op(type)` and type is `Copy`, impls `type.op(&type)`
283macro_rules! bulk_impl_op_ref_self_for {
284  ($(($op:ident, $method:ident) => [$($t:ty),+]),+ $(,)?) => {
285    $( // do each trait/list matching given
286      $( // do the current trait for each type in its list.
287        impl $op<&Self> for $t {
288          type Output = Self;
289          #[inline]
290          #[must_use]
291          fn $method(self, rhs: &Self) -> Self::Output {
292            self.$method(*rhs)
293          }
294        }
295      )+
296    )+
297  };
298}
299
300bulk_impl_op_ref_self_for! {
301  (Add, add) => [f32x8, f32x4, f64x4, f64x2, i8x32, i8x16, i16x8, i16x16, i32x8, i32x4, i64x2, u8x16, u16x8, u16x16, u32x8, u32x4, u64x4, u64x2],
302  (Sub, sub) => [f32x8, f32x4, f64x4, f64x2, i8x32, i8x16, i16x8, i16x16, i32x8, i32x4, i64x2, u8x16, u16x8, u16x16, u32x8, u32x4, u64x4, u64x2],
303  (Mul, mul) => [f32x8, f32x4, f64x4, f64x2, i16x8, i16x16, i32x8, i32x4, u16x8, u16x16],
304  (Div, div) => [f32x8, f32x4, f64x4, f64x2],
305  (BitAnd, bitand) => [f32x8, f32x4, f64x4, f64x2, i8x32, i8x16, i16x8, i16x16, i32x8, i32x4, i64x2, u8x16, u16x8, u16x16,u32x8, u32x4, u64x4, u64x2],
306  (BitOr, bitor) => [f32x8, f32x4, f64x4, f64x2, i8x32, i8x16, i16x8, i16x16, i32x8, i32x4, i64x2, u8x16, u16x8, u16x16, u32x8, u32x4, u64x4, u64x2],
307  (BitXor, bitxor) => [f32x8, f32x4, f64x4, f64x2, i8x32, i8x16, i16x8, i16x16, i32x8, i32x4, i64x2, u8x16, u16x8, u16x16, u32x8, u32x4, u64x4, u64x2],
308}
309
310/// given `type.op(rhs)` and type is Copy, impls `type.op_assign(rhs)`
311macro_rules! bulk_impl_op_assign_for {
312  ($(($op:ident<$rhs:ty>, $method:ident, $method_assign:ident) => [$($t:ty),+]),+ $(,)?) => {
313    $( // do each trait/list matching given
314      $( // do the current trait for each type in its list.
315        impl $op<$rhs> for $t {
316          #[inline]
317          fn $method_assign(&mut self, rhs: $rhs) {
318            *self = self.$method(rhs);
319          }
320        }
321      )+
322    )+
323  };
324}
325
326// Note: remember to update bulk_impl_op_ref_self_for first or this will give
327// weird errors!
328bulk_impl_op_assign_for! {
329  (AddAssign<Self>, add, add_assign) => [f32x8, f32x4, f64x4, f64x2, i8x32, i8x16, i16x8, i16x16, i32x8, i32x4, i64x2, u8x16, u16x8, u16x16, u32x8, u32x4, u64x4, u64x2],
330  (AddAssign<&Self>, add, add_assign) => [f32x8, f32x4, f64x4, f64x2, i8x32, i8x16, i16x8, i16x16, i32x8, i32x4, i64x2, u8x16, u16x8, u16x16, u32x8, u32x4, u64x4, u64x2],
331  (SubAssign<Self>, sub, sub_assign) => [f32x8, f32x4, f64x4, f64x2, i8x32, i8x16, i16x8, i16x16, i32x8, i32x4, i64x2, u8x16, u16x8, u16x16, u32x8, u32x4, u64x4, u64x2],
332  (SubAssign<&Self>, sub, sub_assign) => [f32x8, f32x4, f64x4, f64x2, i8x32, i8x16, i16x8, i16x16, i32x8, i32x4, i64x2, u8x16, u16x8, u16x16, u32x8, u32x4, u64x4, u64x2],
333  (MulAssign<Self>, mul, mul_assign) => [f32x8, f32x4, f64x4, f64x2, i16x8, i16x16, i32x8, i32x4, u16x8, u16x16],
334  (MulAssign<&Self>, mul, mul_assign) => [f32x8, f32x4, f64x4, f64x2, i16x8, i16x16, i32x8, i32x4, u16x8, u16x16],
335  (DivAssign<Self>, div, div_assign) => [f32x8, f32x4, f64x4, f64x2],
336  (DivAssign<&Self>, div, div_assign) => [f32x8, f32x4, f64x4, f64x2],
337  (BitAndAssign<Self>, bitand, bitand_assign) => [f32x8, f32x4, f64x4, f64x2, i8x32, i8x16, i16x8, i16x16, u16x16, i32x8, i32x4, i64x2, u8x16, u16x8, u32x8, u32x4, u64x4, u64x2],
338  (BitAndAssign<&Self>, bitand, bitand_assign) => [f32x8, f32x4, f64x4, f64x2, i8x32, i8x16, i16x8, i16x16, u16x16, i32x8, i32x4, i64x2, u8x16, u16x8, u32x8, u32x4, u64x4, u64x2],
339  (BitOrAssign<Self>, bitor, bitor_assign) => [f32x8, f32x4, f64x4, f64x2, i8x32, i8x16, i16x8, i16x16, u16x16, i32x8, i32x4, i64x2, u8x16, u16x8, u32x8, u32x4, u64x4, u64x2],
340  (BitOrAssign<&Self>, bitor, bitor_assign) => [f32x8, f32x4, f64x4, f64x2, i8x32, i8x16, i16x8, i16x16, u16x16, i32x8, i32x4, i64x2, u8x16, u16x8, u32x8, u32x4, u64x4, u64x2],
341  (BitXorAssign<Self>, bitxor, bitxor_assign) => [f32x8, f32x4, f64x4, f64x2, i8x32, i8x16, i16x8, i16x16, u16x16, i32x8, i32x4, i64x2, u8x16, u16x8, u32x8, u32x4, u64x4, u64x2],
342  (BitXorAssign<&Self>, bitxor, bitxor_assign) => [f32x8, f32x4, f64x4, f64x2, i8x32, i8x16, i16x8, i16x16, u16x16, i32x8, i32x4, i64x2, u8x16, u16x8, u32x8, u32x4, u64x4, u64x2],
343}
344
345macro_rules! impl_simple_neg {
346  ($($t:ty),+ $(,)?) => {
347    $(
348      impl Neg for $t {
349        type Output = Self;
350        #[inline]
351        #[must_use]
352        fn neg(self) -> Self::Output {
353          Self::default() - self
354        }
355      }
356      impl Neg for &'_ $t {
357        type Output = $t;
358        #[inline]
359        #[must_use]
360        fn neg(self) -> Self::Output {
361          <$t>::default() - *self
362        }
363      }
364    )+
365  };
366}
367
368impl_simple_neg! {
369  f32x8, f32x4, f64x4, f64x2, i8x32, i8x16, i16x8, i16x16, i32x8, i32x4, i64x4, i64x2, u8x16, u16x8, u32x8, u32x4, u64x2, u64x4
370}
371
372macro_rules! impl_simple_not {
373  ($($t:ty),+ $(,)?) => {
374    $(
375      impl Not for $t {
376        type Output = Self;
377        #[inline]
378        #[must_use]
379        fn not(self) -> Self::Output {
380          self ^ cast::<u128, $t>(u128::MAX)
381        }
382      }
383      impl Not for &'_ $t {
384        type Output = $t;
385        #[inline]
386        #[must_use]
387        fn not(self) -> Self::Output {
388          *self ^ cast::<u128, $t>(u128::MAX)
389        }
390      }
391    )+
392  };
393}
394
395impl_simple_not! {
396  f32x4, i8x32, i8x16, i16x8, i16x16, i32x4, i64x2, u8x16, u16x8, u16x16, u32x4, u64x2,
397}
398
399macro_rules! impl_simple_sum {
400  ($($t:ty),+ $(,)?) => {
401    $(
402      impl<RHS> core::iter::Sum<RHS> for $t where $t: AddAssign<RHS> {
403        #[inline]
404        fn sum<I: Iterator<Item = RHS>>(iter: I) -> Self {
405          let mut total = Self::zeroed();
406          for val in iter {
407            total += val;
408          }
409          total
410        }
411      }
412    )+
413  };
414}
415
416impl_simple_sum! {
417  f32x4, f64x4, f64x2, i8x32, i8x16, i16x8, i16x16, i32x8, i32x4, i64x4, i64x2, u8x16, u16x8, u16x16, u32x8, u32x4, u64x2, u64x4
418}
419
420macro_rules! impl_floating_product {
421  ($($t:ty),+ $(,)?) => {
422    $(
423      impl<RHS> core::iter::Product<RHS> for $t where $t: MulAssign<RHS> {
424        #[inline]
425        fn product<I: Iterator<Item = RHS>>(iter: I) -> Self {
426          let mut total = Self::from(1.0);
427          for val in iter {
428            total *= val;
429          }
430          total
431        }
432      }
433    )+
434  };
435}
436
437impl_floating_product! {
438  f32x8, f32x4, f64x4, f64x2
439}
440
441macro_rules! impl_integer_product {
442  ($($t:ty),+ $(,)?) => {
443    $(
444      impl<RHS> core::iter::Product<RHS> for $t where $t: MulAssign<RHS> {
445        #[inline]
446        fn product<I: Iterator<Item = RHS>>(iter: I) -> Self {
447          let mut total = Self::from(1);
448          for val in iter {
449            total *= val;
450          }
451          total
452        }
453      }
454    )+
455  };
456}
457
458impl_integer_product! {
459  i16x8, i32x4, i32x8,
460}
461
462/// impls `From<a> for b` by just calling `cast`
463macro_rules! impl_from_a_for_b_with_cast {
464  ($(($arr:ty, $simd:ty)),+  $(,)?) => {
465    $(impl From<$arr> for $simd {
466      #[inline]
467      #[must_use]
468      fn from(arr: $arr) -> Self {
469        cast(arr)
470      }
471    }
472    impl From<$simd> for $arr {
473      #[inline]
474      #[must_use]
475      fn from(simd: $simd) -> Self {
476        cast(simd)
477      }
478    })+
479  };
480}
481
482impl_from_a_for_b_with_cast! {
483  ([f32;8], f32x8),
484  ([f32;4], f32x4), ([f64;4], f64x4), ([f64;2], f64x2),
485  ([i8;32], i8x32), ([i8;16], i8x16), ([i16;8], i16x8), ([i16;16], i16x16), ([i32;8], i32x8), ([i32;4], i32x4), ([i64;2], i64x2), ([i64;4], i64x4),
486  ([u8;16], u8x16), ([u16;8], u16x8), ([u16;16], u16x16), ([u32;8], u32x8), ([u32;4], u32x4), ([u64;2], u64x2), ([u64;4], u64x4),
487}
488
489macro_rules! impl_from_single_value {
490  ($(([$elem:ty;$len:expr], $simd:ty)),+  $(,)?) => {
491    $(impl From<$elem> for $simd {
492      /// Splats the single value given across all lanes.
493      #[inline]
494      #[must_use]
495      fn from(elem: $elem) -> Self {
496        cast([elem; $len])
497      }
498    }
499    impl $simd {
500      #[inline]
501      #[must_use]
502      pub fn splat(elem: $elem) -> $simd {
503        cast([elem; $len])
504      }
505    })+
506  };
507}
508
509impl_from_single_value! {
510  ([f32;8], f32x8),
511  ([f32;4], f32x4), ([f64;4], f64x4), ([f64;2], f64x2),
512  ([i8;32], i8x32), ([i8;16], i8x16), ([i16;8], i16x8), ([i16;16], i16x16), ([i32;8], i32x8), ([i32;4], i32x4), ([i64;2], i64x2), ([i64;4], i64x4),
513  ([u8;16], u8x16), ([u16;8], u16x8), ([u16;16], u16x16), ([u32;8], u32x8), ([u32;4], u32x4), ([u64;2], u64x2), ([u64;4], u64x4),
514}
515
516/// formatter => [(arr, simd)+],+
517macro_rules! impl_formatter_for {
518  ($($trait:ident => [$(($arr:ty, $simd:ty)),+]),+ $(,)?) => {
519    $( // do per trait
520      $( // do per simd type
521        impl $trait for $simd {
522          #[allow(clippy::missing_inline_in_public_items)]
523          fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
524            let a: $arr = cast(*self);
525            write!(f, "(")?;
526            for (x, a_ref) in a.iter().enumerate() {
527              if x > 0 {
528                write!(f, ", ")?;
529              }
530              $trait::fmt(a_ref, f)?;
531            }
532            write!(f, ")")
533          }
534        }
535      )+
536    )+
537  }
538}
539
540impl_formatter_for! {
541  Binary => [([u32;8], f32x8), ([u32;4], f32x4), ([u64;4], f64x4), ([u64;2], f64x2),
542  ([i8;32], i8x32), ([i8;16], i8x16), ([i16;8], i16x8), ([i16;16], i16x16), ([i32;8], i32x8), ([i32;4], i32x4), ([i64;2], i64x2),([i64;4], i64x4),
543  ([u8;16], u8x16), ([u16;8], u16x8), ([u16;16], u16x16), ([u32;8], u32x8), ([u32;4], u32x4), ([u64;2], u64x2),([u64;4], u64x4)],
544  Debug => [([f32;8], f32x8), ([f32;4], f32x4), ([f64;4], f64x4), ([f64;2], f64x2),
545  ([i8;32], i8x32), ([i8;16], i8x16), ([i16;8], i16x8), ([i16;16], i16x16), ([i32;8], i32x8), ([i32;4], i32x4), ([i64;2], i64x2),([i64;4], i64x4),
546  ([u8;16], u8x16), ([u16;8], u16x8), ([u16;16], u16x16), ([u32;8], u32x8), ([u32;4], u32x4), ([u64;2], u64x2),([u64;4], u64x4)],
547  Display => [([f32;8], f32x8), ([f32;4], f32x4), ([f64;4], f64x4), ([f64;2], f64x2),
548  ([i8;32], i8x32), ([i8;16], i8x16), ([i16;8], i16x8), ([i16;16], i16x16), ([i32;8], i32x8), ([i32;4], i32x4), ([i64;2], i64x2),([i64;4], i64x4),
549  ([u8;16], u8x16), ([u16;8], u16x8), ([u16;16], u16x16), ([u32;8], u32x8), ([u32;4], u32x4), ([u64;2], u64x2),([u64;4], u64x4)],
550  LowerExp => [([f32;8], f32x8), ([f32;4], f32x4), ([u64;4], f64x4), ([u64;2], f64x2),
551  ([i8;32], i8x32), ([i8;16], i8x16), ([i16;8], i16x8), ([i16;16], i16x16), ([i32;8], i32x8), ([i32;4], i32x4), ([i64;2], i64x2),([i64;4], i64x4),
552  ([u8;16], u8x16), ([u16;8], u16x8), ([u16;16], u16x16), ([u32;8], u32x8), ([u32;4], u32x4), ([u64;2], u64x2),([u64;4], u64x4)],
553  LowerHex => [([u32;8], f32x8), ([u32;4], f32x4), ([u64;4], f64x4), ([u64;2], f64x2),
554  ([i8;32], i8x32), ([i8;16], i8x16), ([i16;8], i16x8), ([i16;16], i16x16), ([i32;8], i32x8), ([i32;4], i32x4), ([i64;2], i64x2),([i64;4], i64x4),
555  ([u8;16], u8x16), ([u16;8], u16x8), ([u16;16], u16x16), ([u32;8], u32x8), ([u32;4], u32x4), ([u64;2], u64x2),([u64;4], u64x4)],
556  Octal => [([u32;8], f32x8), ([u32;4], f32x4), ([u64;4], f64x4), ([u64;2], f64x2),
557  ([i8;32], i8x32), ([i8;16], i8x16), ([i16;8], i16x8), ([i16;16], i16x16), ([i32;8], i32x8), ([i32;4], i32x4), ([i64;2], i64x2),([i64;4], i64x4),
558  ([u8;16], u8x16), ([u16;8], u16x8), ([u16;16], u16x16), ([u32;8], u32x8), ([u32;4], u32x4), ([u64;2], u64x2),([u64;4], u64x4)],
559  UpperExp => [([u32;8], f32x8), ([u32;4], f32x4), ([u64;4], f64x4), ([u64;2], f64x2),
560  ([i8;32], i8x32), ([i8;16], i8x16), ([i16;8], i16x8), ([i16;16], i16x16), ([i32;8], i32x8), ([i32;4], i32x4), ([i64;2], i64x2),([i64;4], i64x4),
561  ([u8;16], u8x16), ([u16;8], u16x8), ([u16;16], u16x16), ([u32;8], u32x8), ([u32;4], u32x4), ([u64;2], u64x2),([u64;4], u64x4)],
562  UpperHex => [([u32;8], f32x8), ([u32;4], f32x4), ([u64;4], f64x4), ([u64;2], f64x2),
563  ([i8;32], i8x32), ([i8;16], i8x16), ([i16;8], i16x8), ([i16;16], i16x16), ([i32;8], i32x8), ([i32;4], i32x4), ([i64;2], i64x2),([i64;4], i64x4),
564  ([u8;16], u8x16), ([u16;8], u16x8), ([u16;16], u16x16), ([u32;8], u32x8), ([u32;4], u32x4), ([u64;2], u64x2),([u64;4], u64x4)],
565}
566
567// With const generics this could be simplified I hope
568macro_rules! from_array {
569  ($ty:ty,$dst:ty,$dst_wide:ident,32) => {
570    impl From<&[$ty]> for $dst_wide {
571      #[inline]
572      fn from(src: &[$ty]) -> $dst_wide {
573        match src.len() {
574          32 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst, src[13] as $dst, src[14] as $dst, src[15] as $dst, src[16] as $dst, src[17] as $dst, src[18] as $dst, src[19] as $dst, src[20] as $dst, src[21] as $dst, src[22] as $dst, src[23] as $dst, src[24] as $dst, src[25] as $dst, src[26] as $dst, src[27] as $dst, src[28] as $dst, src[29] as $dst, src[30] as $dst, src[31] as $dst,]),
575          31 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst, src[13] as $dst, src[14] as $dst, src[15] as $dst, src[16] as $dst, src[17] as $dst, src[18] as $dst, src[19] as $dst, src[20] as $dst, src[21] as $dst, src[22] as $dst, src[23] as $dst, src[24] as $dst, src[25] as $dst, src[26] as $dst, src[27] as $dst, src[28] as $dst, src[29] as $dst, src[30] as $dst,0 as $dst,]),
576          30 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst, src[13] as $dst, src[14] as $dst, src[15] as $dst, src[16] as $dst, src[17] as $dst, src[18] as $dst, src[19] as $dst, src[20] as $dst, src[21] as $dst, src[22] as $dst, src[23] as $dst, src[24] as $dst, src[25] as $dst, src[26] as $dst, src[27] as $dst, src[28] as $dst, src[29] as $dst,0 as $dst,0 as $dst,]),
577          29 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst, src[13] as $dst, src[14] as $dst, src[15] as $dst, src[16] as $dst, src[17] as $dst, src[18] as $dst, src[19] as $dst, src[20] as $dst, src[21] as $dst, src[22] as $dst, src[23] as $dst, src[24] as $dst, src[25] as $dst, src[26] as $dst, src[27] as $dst, src[28] as $dst,0 as $dst,0 as $dst,0 as $dst,]),
578          28 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst, src[13] as $dst, src[14] as $dst, src[15] as $dst, src[16] as $dst, src[17] as $dst, src[18] as $dst, src[19] as $dst, src[20] as $dst, src[21] as $dst, src[22] as $dst, src[23] as $dst, src[24] as $dst, src[25] as $dst, src[26] as $dst, src[27] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
579          27 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst, src[13] as $dst, src[14] as $dst, src[15] as $dst, src[16] as $dst, src[17] as $dst, src[18] as $dst, src[19] as $dst, src[20] as $dst, src[21] as $dst, src[22] as $dst, src[23] as $dst, src[24] as $dst, src[25] as $dst, src[26] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
580          26 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst, src[13] as $dst, src[14] as $dst, src[15] as $dst, src[16] as $dst, src[17] as $dst, src[18] as $dst, src[19] as $dst, src[20] as $dst, src[21] as $dst, src[22] as $dst, src[23] as $dst, src[24] as $dst, src[25] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
581          25 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst, src[13] as $dst, src[14] as $dst, src[15] as $dst, src[16] as $dst, src[17] as $dst, src[18] as $dst, src[19] as $dst, src[20] as $dst, src[21] as $dst, src[22] as $dst, src[23] as $dst, src[24] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
582          24 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst, src[13] as $dst, src[14] as $dst, src[15] as $dst, src[16] as $dst, src[17] as $dst, src[18] as $dst, src[19] as $dst, src[20] as $dst, src[21] as $dst, src[22] as $dst, src[23] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
583          23 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst, src[13] as $dst, src[14] as $dst, src[15] as $dst, src[16] as $dst, src[17] as $dst, src[18] as $dst, src[19] as $dst, src[20] as $dst, src[21] as $dst, src[22] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
584          22 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst, src[13] as $dst, src[14] as $dst, src[15] as $dst, src[16] as $dst, src[17] as $dst, src[18] as $dst, src[19] as $dst, src[20] as $dst, src[21] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
585          21 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst, src[13] as $dst, src[14] as $dst, src[15] as $dst, src[16] as $dst, src[17] as $dst, src[18] as $dst, src[19] as $dst, src[20] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
586          20 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst, src[13] as $dst, src[14] as $dst, src[15] as $dst, src[16] as $dst, src[17] as $dst, src[18] as $dst, src[19] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
587          19 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst, src[13] as $dst, src[14] as $dst, src[15] as $dst, src[16] as $dst, src[17] as $dst, src[18] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
588          18 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst, src[13] as $dst, src[14] as $dst, src[15] as $dst, src[16] as $dst, src[17] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
589          17 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst, src[13] as $dst, src[14] as $dst, src[15] as $dst, src[16] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
590          16 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst, src[13] as $dst, src[14] as $dst, src[15] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
591          15 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst, src[13] as $dst, src[14] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
592          14 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst, src[13] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
593          13 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
594          12 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
595          11 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
596          10 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
597          9 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
598          8 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
599          7 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
600          6 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
601          5 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
602          4 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
603          3 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
604          2 => $dst_wide::from([src[0] as $dst, src[1] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
605          1 => $dst_wide::from([src[0] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
606          _ => panic!(
607            "Converting from an array larger than what can be stored in $dst_wide"
608          ),
609        }
610      }
611    }
612  };
613  ($ty:ty,$dst:ty,$dst_wide:ident,16) => {
614    impl From<&[$ty]> for $dst_wide {
615      #[inline]
616      fn from(src: &[$ty]) -> $dst_wide {
617        match src.len() {
618          16 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst, src[13] as $dst, src[14] as $dst, src[15] as $dst,]),
619          15 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst, src[13] as $dst, src[14] as $dst,0 as $dst,]),
620          14 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst, src[13] as $dst,0 as $dst,0 as $dst,]),
621          13 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst, src[12] as $dst,0 as $dst,0 as $dst,0 as $dst,]),
622          12 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst, src[11] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
623          11 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst, src[10] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
624          10 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst, src[9] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
625          9 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst, src[8] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
626          8 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
627          7 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
628          6 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
629          5 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
630          4 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
631          3 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
632          2 => $dst_wide::from([src[0] as $dst, src[1] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
633          1 => $dst_wide::from([src[0] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
634          _ => panic!(
635            "Converting from an array larger than what can be stored in $dst_wide"
636          ),
637        }
638      }
639    }
640  };
641  ($ty:ty,$dst:ty,$dst_wide:ident,8) => {
642    impl From<&[$ty]> for $dst_wide {
643      #[inline]
644      fn from(src: &[$ty]) -> $dst_wide {
645        match src.len() {
646          8 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst, src[7] as $dst,]),
647          7 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst, src[6] as $dst,0 as $dst,]),
648          6 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst, src[5] as $dst,0 as $dst,0 as $dst,]),
649          5 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst, src[4] as $dst,0 as $dst,0 as $dst,0 as $dst,]),
650          4 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
651          3 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
652          2 => $dst_wide::from([src[0] as $dst, src[1] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
653          1 => $dst_wide::from([src[0] as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
654          0 => $dst_wide::from([0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,0 as $dst,]),
655          _ => panic!(
656            "Converting from an array larger than what can be stored in $dst_wide"
657          ),
658        }
659      }
660    }
661  };
662  ($ty:ty,$dst:ty,$dst_wide:ident,4) => {
663    impl From<&[$ty]> for $dst_wide {
664      #[inline]
665      fn from(src: &[$ty]) -> $dst_wide {
666        match src.len() {
667          4 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst, src[3] as $dst,]),
668          3 => $dst_wide::from([src[0] as $dst, src[1] as $dst, src[2] as $dst,0 as $dst,]),
669          2 => $dst_wide::from([src[0] as $dst, src[1] as $dst,0 as $dst,0 as $dst,]),
670          1 => $dst_wide::from([src[0] as $dst,0 as $dst,0 as $dst,0 as $dst,]),
671          _ => panic!(
672            "Converting from an array larger than what can be stored in $dst_wide"
673          ),
674        }
675      }
676    }
677  };
678}
679
680from_array!(i8, i8, i8x32, 32);
681from_array!(i8, i8, i8x16, 16);
682from_array!(i8, i32, i32x8, 8);
683from_array!(u8, u8, u8x16, 16);
684from_array!(i16, i16, i16x16, 16);
685from_array!(u16, u16, u16x16, 16);
686from_array!(i32, i32, i32x8, 8);
687from_array!(f32, f32, f32x8, 8);
688from_array!(f32, f32, f32x4, 4);
689from_array!(f64, f64, f64x4, 4);
690from_array!(u64, u64, u64x4, 4);
691from_array!(i64, i64, i64x4, 4);
692
693#[allow(unused)]
694fn software_sqrt(x: f64) -> f64 {
695  use core::num::Wrapping;
696  type wu32 = Wrapping<u32>;
697  const fn w(u: u32) -> wu32 {
698    Wrapping(u)
699  }
700  let mut z: f64;
701  let sign: wu32 = w(0x80000000);
702  let mut ix0: i32;
703  let mut s0: i32;
704  let mut q: i32;
705  let mut m: i32;
706  let mut t: i32;
707  let mut i: i32;
708  let mut r: wu32;
709  let mut t1: wu32;
710  let mut s1: wu32;
711  let mut ix1: wu32;
712  let mut q1: wu32;
713  // extract data
714
715  pick! {
716    if #[cfg(target_endian = "little")]
717    {
718      let [low, high]: [u32; 2] = cast(x);
719      ix0 = high as i32;
720      ix1 = w(low);
721    }
722    else
723    {
724      let [high, low]: [u32; 2] = cast(x);
725      ix0 = high as i32;
726      ix1 = w(low);
727    }
728  }
729
730  // inf and nan
731  {
732    if x.is_nan() {
733      return f64::NAN;
734    }
735    if ix0 & 0x7ff00000 == 0x7ff00000 {
736      return x * x + x;
737    }
738  }
739  // handle zero
740  {
741    if ix0 <= 0 {
742      if ((ix0 & (!sign).0 as i32) | (ix1.0 as i32)) == 0 {
743        return x;
744      } else if ix0 < 0 {
745        return (x - x) / (x - x);
746      }
747    }
748  }
749  // normalize
750  {
751    m = ix0 >> 20;
752    if m == 0 {
753      // subnormal
754      while ix0 == 0 {
755        m -= 21;
756        ix0 |= (ix1 >> 11).0 as i32;
757        ix1 <<= 21;
758      }
759      i = 0;
760      while ix0 & 0x00100000 == 0 {
761        ix0 <<= 1;
762        i += 1;
763      }
764      m -= i - 1;
765      ix0 |= (ix1.0 >> (31 - i)) as i32;
766      ix1 <<= i as usize;
767    }
768    // un-bias exponent
769    m -= 1023;
770    ix0 = (ix0 & 0x000fffff) | 0x00100000;
771    if (m & 1) != 0 {
772      // odd m, double the input to make it even
773      ix0 += ix0 + ((ix1 & sign) >> 31).0 as i32;
774      ix1 += ix1;
775    }
776    m >>= 1;
777  }
778  // generate sqrt bit by bit
779  {
780    ix0 += ix0 + ((ix1 & sign) >> 31).0 as i32;
781    ix1 += ix1;
782    // q and q1 store the sqrt(x);
783    q = 0;
784    q1 = w(0);
785    s0 = 0;
786    s1 = w(0);
787    // our bit that moves from right to left
788    r = w(0x00200000);
789    while r != w(0) {
790      t = s0 + (r.0 as i32);
791      if t <= ix0 {
792        s0 = t + (r.0 as i32);
793        ix0 -= t;
794        q += (r.0 as i32);
795      }
796      ix0 += ix0 + ((ix1 & sign) >> 31).0 as i32;
797      ix1 += ix1;
798      r >>= 1;
799    }
800    r = sign;
801    while r != w(0) {
802      t1 = s1 + r;
803      t = s0;
804      if (t < ix0) || ((t == ix0) && (t1 <= ix1)) {
805        s1 = t1 + r;
806        if t1 & sign == sign && (s1 & sign) == w(0) {
807          s0 += 1;
808        }
809        ix0 -= t;
810        if ix1 < t1 {
811          ix0 -= 1;
812        }
813        ix1 -= t1;
814        q1 += r;
815      }
816      ix0 += ix0 + ((ix1 & sign) >> 31).0 as i32;
817      ix1 += ix1;
818      r >>= 1;
819    }
820  }
821  // use floating add to find out rounding direction
822  {
823    if ix0 | (ix1.0 as i32) != 0 {
824      z = 1.0 - 1.0e-300;
825      if z >= 1.0 {
826        z = 1.0 + 1.0e-300;
827        if q1 == w(0xffffffff) {
828          q1 = w(0);
829          q += 1;
830        } else if z > 1.0 {
831          if q1 == w(0xfffffffe) {
832            q += 1;
833          }
834          q1 += w(2);
835        } else {
836          q1 += q1 & w(1);
837        }
838      }
839    }
840  }
841  // finish up
842  ix0 = (q >> 1) + 0x3fe00000;
843  ix1 = q1 >> 1;
844  if q & 1 == 1 {
845    ix1 |= sign;
846  }
847  ix0 += m << 20;
848
849  pick! {
850    if #[cfg(target_endian = "little")]
851    {
852      cast::<[u32; 2], f64>([ix1.0, ix0 as u32])
853    }
854    else
855    {
856      cast::<[u32; 2], f64>([ix0 as u32, ix1.0])
857    }
858  }
859}
860
861#[test]
862fn test_software_sqrt() {
863  assert!(software_sqrt(f64::NAN).is_nan());
864  assert_eq!(software_sqrt(f64::INFINITY), f64::INFINITY);
865  assert_eq!(software_sqrt(0.0), 0.0);
866  assert_eq!(software_sqrt(-0.0), -0.0);
867  assert!(software_sqrt(-1.0).is_nan());
868  assert!(software_sqrt(f64::NEG_INFINITY).is_nan());
869  assert_eq!(software_sqrt(4.0), 2.0);
870  assert_eq!(software_sqrt(9.0), 3.0);
871  assert_eq!(software_sqrt(16.0), 4.0);
872  assert_eq!(software_sqrt(25.0), 5.0);
873  assert_eq!(software_sqrt(5000.0 * 5000.0), 5000.0);
874}
875
876pub trait CmpEq<Rhs = Self> {
877  type Output;
878  fn cmp_eq(self, rhs: Rhs) -> Self::Output;
879}
880
881pub trait CmpGt<Rhs = Self> {
882  type Output;
883  fn cmp_gt(self, rhs: Rhs) -> Self::Output;
884}
885
886pub trait CmpGe<Rhs = Self> {
887  type Output;
888  fn cmp_ge(self, rhs: Rhs) -> Self::Output;
889}
890
891pub trait CmpNe<Rhs = Self> {
892  type Output;
893  fn cmp_ne(self, rhs: Rhs) -> Self::Output;
894}
895
896pub trait CmpLt<Rhs = Self> {
897  type Output;
898  fn cmp_lt(self, rhs: Rhs) -> Self::Output;
899}
900
901pub trait CmpLe<Rhs = Self> {
902  type Output;
903  fn cmp_le(self, rhs: Rhs) -> Self::Output;
904}
905
906macro_rules! bulk_impl_const_rhs_op {
907  (($op:ident,$method:ident) => [$(($lhs:ty,$rhs:ty),)+]) => {
908    $(
909    impl $op<$rhs> for $lhs {
910      type Output = Self;
911      #[inline]
912      #[must_use]
913      fn $method(self, rhs: $rhs) -> Self::Output {
914        self.$method(<$lhs>::splat(rhs))
915      }
916    }
917    )+
918  };
919}
920
921bulk_impl_const_rhs_op!((CmpEq, cmp_eq) => [(f64x4, f64), (f64x2, f64), (f32x4,f32), (f32x8,f32),]);
922bulk_impl_const_rhs_op!((CmpLt, cmp_lt) => [(f64x4, f64), (f64x2, f64), (f32x4,f32), (f32x8,f32),]);
923bulk_impl_const_rhs_op!((CmpGt, cmp_gt) => [(f64x4, f64), (f64x2, f64), (f32x4,f32), (f32x8,f32),]);
924bulk_impl_const_rhs_op!((CmpNe, cmp_ne) => [(f64x4, f64), (f64x2, f64), (f32x4,f32), (f32x8,f32),]);
925bulk_impl_const_rhs_op!((CmpLe, cmp_le) => [(f64x4, f64), (f64x2, f64), (f32x4,f32), (f32x8,f32),]);
926bulk_impl_const_rhs_op!((CmpGe, cmp_ge) => [(f64x4, f64), (f64x2, f64), (f32x4,f32), (f32x8,f32),]);
927
928macro_rules! impl_serde {
929  ($i:ident, [$t:ty; $len:expr]) => {
930    #[cfg(feature = "serde")]
931    impl Serialize for $i {
932      #[inline]
933      fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>
934      where
935        S: serde::Serializer,
936      {
937        let array = self.as_array_ref();
938        let mut seq = serializer.serialize_tuple($len)?;
939        for e in array {
940          seq.serialize_element(e)?;
941        }
942        seq.end()
943      }
944    }
945
946    #[cfg(feature = "serde")]
947    impl<'de> Deserialize<'de> for $i {
948      #[inline]
949      fn deserialize<D>(deserializer: D) -> Result<Self, D::Error>
950      where
951        D: serde::Deserializer<'de>,
952      {
953        Ok(<[$t; $len]>::deserialize(deserializer)?.into())
954      }
955    }
956  };
957}
958
959impl_serde!(f32x8, [f32; 8]);
960impl_serde!(f32x4, [f32; 4]);
961impl_serde!(f64x4, [f64; 4]);
962impl_serde!(f64x2, [f64; 2]);
963impl_serde!(i8x16, [i8; 16]);
964impl_serde!(i16x16, [i16; 16]);
965impl_serde!(i8x32, [i8; 32]);
966impl_serde!(i16x8, [i16; 8]);
967impl_serde!(i32x4, [i32; 4]);
968impl_serde!(i32x8, [i32; 8]);
969impl_serde!(i64x2, [i64; 2]);
970impl_serde!(i64x4, [i64; 4]);
971impl_serde!(u8x16, [u8; 16]);
972impl_serde!(u16x8, [u16; 8]);
973impl_serde!(u16x16, [u16; 16]);
974impl_serde!(u32x4, [u32; 4]);
975impl_serde!(u32x8, [u32; 8]);
976impl_serde!(u64x2, [u64; 2]);
977impl_serde!(u64x4, [u64; 4]);
wide/lib.rs

wide/
lib.rs
