simd 0.2.5

OBSOLETE: See `core::simd` instead. 
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
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290

use super::super::*;
use sixty_four::*;

use super::super::bitcast;

pub use v256::{
    f64x4, bool64fx4, u64x4, i64x4, bool64ix4,
    f32x8, bool32fx8, u32x8, i32x8, bool32ix8,
    u16x16, i16x16, bool16ix16,
    u8x32, i8x32, bool8ix32,
    LowHigh128
};

#[allow(dead_code)]
extern "platform-intrinsic" {
    fn x86_mm256_addsub_ps(x: f32x8, y: f32x8) -> f32x8;
    fn x86_mm256_addsub_pd(x: f64x4, y: f64x4) -> f64x4;
    fn x86_mm256_dp_ps(x: f32x8, y: f32x8, z: i32) -> f32x8;
    fn x86_mm256_hadd_ps(x: f32x8, y: f32x8) -> f32x8;
    fn x86_mm256_hadd_pd(x: f64x4, y: f64x4) -> f64x4;
    fn x86_mm256_hsub_ps(x: f32x8, y: f32x8) -> f32x8;
    fn x86_mm256_hsub_pd(x: f64x4, y: f64x4) -> f64x4;
    fn x86_mm256_max_ps(x: f32x8, y: f32x8) -> f32x8;
    fn x86_mm256_max_pd(x: f64x4, y: f64x4) -> f64x4;
    fn x86_mm256_min_ps(x: f32x8, y: f32x8) -> f32x8;
    fn x86_mm256_min_pd(x: f64x4, y: f64x4) -> f64x4;
    fn x86_mm256_movemask_ps(x: f32x8) -> i32;
    fn x86_mm256_movemask_pd(x: f64x4) -> i32;
    fn x86_mm_permutevar_ps(x: f32x4, y: i32x4) -> f32x4;
    fn x86_mm_permutevar_pd(x: f64x2, y: i64x2) -> f64x2;
    fn x86_mm256_permutevar_ps(x: f32x8, y: i32x8) -> f32x8;
    fn x86_mm256_permutevar_pd(x: f64x4, y: i64x4) -> f64x4;
    fn x86_mm256_rcp_ps(x: f32x8) -> f32x8;
    fn x86_mm256_rsqrt_ps(x: f32x8) -> f32x8;
    fn x86_mm256_sqrt_ps(x: f32x8) -> f32x8;
    fn x86_mm256_sqrt_pd(x: f64x4) -> f64x4;
    fn x86_mm_testc_ps(x: f32x4, y: f32x4) -> i32;
    fn x86_mm256_testc_ps(x: f32x8, y: f32x8) -> i32;
    fn x86_mm_testc_pd(x: f64x2, y: f64x2) -> i32;
    fn x86_mm256_testc_pd(x: f64x4, y: f64x4) -> i32;
    fn x86_mm256_testc_si256(x: u64x4, y: u64x4) -> i32;
    fn x86_mm_testnzc_ps(x: f32x4, y: f32x4) -> i32;
    fn x86_mm256_testnzc_ps(x: f32x8, y: f32x8) -> i32;
    fn x86_mm_testnzc_pd(x: f64x2, y: f64x2) -> i32;
    fn x86_mm256_testnzc_pd(x: f64x4, y: f64x4) -> i32;
    fn x86_mm256_testnzc_si256(x: u64x4, y: u64x4) -> i32;
    fn x86_mm_testz_ps(x: f32x4, y: f32x4) -> i32;
    fn x86_mm256_testz_ps(x: f32x8, y: f32x8) -> i32;
    fn x86_mm_testz_pd(x: f64x2, y: f64x2) -> i32;
    fn x86_mm256_testz_pd(x: f64x4, y: f64x4) -> i32;
    fn x86_mm256_testz_si256(x: u64x4, y: u64x4) -> i32;
}

#[doc(hidden)]
pub mod common {
    use super::*;
    use core::mem;

    macro_rules! bools {
        ($($ty: ty, $all: ident, $any: ident, $testc: ident, $testz: ident;)*) => {
            $(
                #[inline]
                pub fn $all(x: $ty) -> bool {
                    unsafe {
                        super::$testc(mem::transmute(x), mem::transmute(<$ty>::splat(true))) != 0
                    }
                }
                #[inline]
                pub fn $any(x: $ty) -> bool {
                    unsafe {
                        super::$testz(mem::transmute(x), mem::transmute(x)) == 0
                    }
                }
                )*
        }
    }

    bools! {
        bool32fx8, bool32fx8_all, bool32fx8_any, x86_mm256_testc_ps, x86_mm256_testz_ps;
        bool64fx4, bool64fx4_all, bool64fx4_any, x86_mm256_testc_pd, x86_mm256_testz_pd;
        bool8ix32, bool8ix32_all, bool8ix32_any, x86_mm256_testc_si256, x86_mm256_testz_si256;
        bool16ix16, bool16ix16_all, bool16ix16_any, x86_mm256_testc_si256, x86_mm256_testz_si256;
        bool32ix8, bool32ix8_all, bool32ix8_any, x86_mm256_testc_si256, x86_mm256_testz_si256;
        bool64ix4, bool64ix4_all, bool64ix4_any, x86_mm256_testc_si256, x86_mm256_testz_si256;
    }
}

// 128-bit vectors:

// 32 bit floats

pub trait AvxF32x4 {
    fn permutevar(self, other: i32x4) -> f32x4;
}
impl AvxF32x4 for f32x4 {
    fn permutevar(self, other: i32x4) -> f32x4 {
        unsafe { x86_mm_permutevar_ps(self, other) }
    }
}

pub trait AvxF64x4 {
    fn sqrt(self) -> Self;
    fn addsub(self, other: Self) -> Self;
    fn hadd(self, other: Self) -> Self;
    fn hsub(self, other: Self) -> Self;
    fn max(self, other: Self) -> Self;
    fn min(self, other: Self) -> Self;
    fn move_mask(self) -> u32;
}

impl AvxF64x4 for f64x4 {
    #[inline]
    fn sqrt(self) -> Self {
        unsafe { x86_mm256_sqrt_pd(self) }
    }

    #[inline]
    fn addsub(self, other: Self) -> Self {
        unsafe { x86_mm256_addsub_pd(self, other) }
    }

    #[inline]
    fn hadd(self, other: Self) -> Self {
        unsafe { x86_mm256_hadd_pd(self, other) }
    }

    #[inline]
    fn hsub(self, other: Self) -> Self {
        unsafe { x86_mm256_hsub_pd(self, other) }
    }

    #[inline]
    fn max(self, other: Self) -> Self {
        unsafe { x86_mm256_max_pd(self, other) }
    }

    #[inline]
    fn min(self, other: Self) -> Self {
        unsafe { x86_mm256_min_pd(self, other) }
    }

    #[inline]
    fn move_mask(self) -> u32 {
        unsafe { x86_mm256_movemask_pd(self) as u32 }
    }
}

pub trait AvxBool64fx4 {
    fn move_mask(self) -> u32;
}
impl AvxBool64fx4 for bool64fx4 {
    #[inline]
    fn move_mask(self) -> u32 {
        unsafe { x86_mm256_movemask_pd(bitcast(self)) as u32 }
    }
}

pub trait AvxF32x8 {
    fn sqrt(self) -> Self;
    fn addsub(self, other: Self) -> Self;
    fn hadd(self, other: Self) -> Self;
    fn hsub(self, other: Self) -> Self;
    fn max(self, other: Self) -> Self;
    fn min(self, other: Self) -> Self;
    fn move_mask(self) -> u32;
    /// Compute an approximation to the reciprocal of the square root
    /// of `self`, that is, `f32x8::splat(1.0) / self.sqrt()`.
    ///
    /// The accuracy of this approximation is platform dependent.
    fn approx_rsqrt(self) -> Self;
    /// Compute an approximation to the reciprocal of `self`, that is,
    /// `f32x8::splat(1.0) / self`.
    ///
    /// The accuracy of this approximation is platform dependent.
    fn approx_reciprocal(self) -> Self;
}

impl AvxF32x8 for f32x8 {
    #[inline]
    fn sqrt(self) -> Self {
        unsafe { x86_mm256_sqrt_ps(self) }
    }

    #[inline]
    fn addsub(self, other: Self) -> Self {
        unsafe { x86_mm256_addsub_ps(self, other) }
    }

    #[inline]
    fn hadd(self, other: Self) -> Self {
        unsafe { x86_mm256_hadd_ps(self, other) }
    }

    #[inline]
    fn hsub(self, other: Self) -> Self {
        unsafe { x86_mm256_hsub_ps(self, other) }
    }

    #[inline]
    fn max(self, other: Self) -> Self {
        unsafe { x86_mm256_max_ps(self, other) }
    }

    #[inline]
    fn min(self, other: Self) -> Self {
        unsafe { x86_mm256_min_ps(self, other) }
    }

    #[inline]
    fn move_mask(self) -> u32 {
        unsafe { x86_mm256_movemask_ps(self) as u32 }
    }

    #[inline]
    fn approx_reciprocal(self) -> Self {
        unsafe { x86_mm256_rcp_ps(self) }
    }

    #[inline]
    fn approx_rsqrt(self) -> Self {
        unsafe { x86_mm256_rsqrt_ps(self) }
    }
}

pub trait AvxBool32fx8 {
    fn move_mask(self) -> u32;
}
impl AvxBool32fx8 for bool32fx8 {
    #[inline]
    fn move_mask(self) -> u32 {
        unsafe { x86_mm256_movemask_ps(bitcast(self)) as u32 }
    }
}

pub trait AvxBool32fx4 {}
impl AvxBool32fx4 for bool32fx4 {}

// 64 bit floats

pub trait AvxF64x2 {
    fn permutevar(self, other: i64x2) -> f64x2;
}
impl AvxF64x2 for f64x2 {
    fn permutevar(self, other: i64x2) -> f64x2 {
        unsafe { x86_mm_permutevar_pd(self, other) }
    }
}

pub trait AvxBool64fx2 {}
impl AvxBool64fx2 for bool64fx2 {}

// 64 bit integers

pub trait AvxU64x2 {}
impl AvxU64x2 for u64x2 {}
pub trait AvxI64x2 {}
impl AvxI64x2 for i64x2 {}

pub trait AvxBool64ix2 {}
impl AvxBool64ix2 for bool64ix2 {}

// 32 bit integers

pub trait AvxU32x4 {}
impl AvxU32x4 for u32x4 {}
pub trait AvxI32x4 {}
impl AvxI32x4 for i32x4 {}

pub trait AvxBool32ix4 {}
impl AvxBool32ix4 for bool32ix4 {}

// 16 bit integers

pub trait AvxU16x8 {}
impl AvxU16x8 for u16x8 {}
pub trait AvxI16x8 {}
impl AvxI16x8 for i16x8 {}

pub trait AvxBool16ix8 {}
impl AvxBool16ix8 for bool16ix8 {}

// 8 bit integers

pub trait AvxU8x16 {}
impl AvxU8x16 for u8x16 {}
pub trait AvxI8x16 {}
impl AvxI8x16 for i8x16 {}

pub trait AvxBool8ix16 {}
impl AvxBool8ix16 for bool8ix16 {}