simdeez 0.2.0

SIMD library to abstract over different instruction sets and widths
Documentation 
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

use super::*;
use std::mem;
pub struct Scalar;

impl Simd for Scalar {
    type Vi32 = i32;
    type Vf32 = f32;

    const WIDTH_BYTES: usize = 4;

    unsafe fn set_lane_epi32(a: &mut Self::Vi32, value: i32, i: usize) {
        debug_assert!(i < Self::WIDTH_BYTES / 4);
        *a = value;
    }
    unsafe fn set_lane_ps(a: &mut Self::Vf32, value: f32, i: usize) {
        debug_assert!(i < Self::WIDTH_BYTES / 4);
        *a = value;
    }
    unsafe fn get_lane_epi32(a: Self::Vi32, i: usize) -> i32 {
        debug_assert!(i < Self::WIDTH_BYTES / 4);
        a
    }
    unsafe fn get_lane_ps(a: Self::Vf32, i: usize) -> f32 {
        debug_assert!(i < Self::WIDTH_BYTES / 4);
        a
    }
    unsafe fn abs_ps(a: Self::Vf32) -> Self::Vf32 {
        a.abs()
    }
    unsafe fn add_epi32(a: Self::Vi32, b: Self::Vi32) -> Self::Vi32 {
        a + b
    }
    unsafe fn add_ps(a: Self::Vf32, b: Self::Vf32) -> Self::Vf32 {
        a + b
    }
    unsafe fn and_si(a: Self::Vi32, b: Self::Vi32) -> Self::Vi32 {
        a & b
    }
    unsafe fn andnot_ps(a: Self::Vf32, b: Self::Vf32) -> Self::Vf32 {
        let ai = mem::transmute::<f32, i32>(a);
        let bi = mem::transmute::<f32, i32>(b);

        mem::transmute::<i32, f32>((!ai) & bi)
    }
    unsafe fn andnot_si(a: Self::Vi32, b: Self::Vi32) -> Self::Vi32 {
        (!a) & b
    }
    unsafe fn blendv_epi32(a: Self::Vi32, b: Self::Vi32, mask: Self::Vi32) -> Self::Vi32 {
        ((!mask) & a) | (mask & b)
    }
    unsafe fn blendv_ps(a: Self::Vf32, b: Self::Vf32, mask: Self::Vf32) -> Self::Vf32 {
        let ai = mem::transmute::<f32, i32>(a);
        let bi = mem::transmute::<f32, i32>(b);
        let maski = mem::transmute::<f32, i32>(mask);
        mem::transmute::<i32, f32>(((!maski) & ai) | (maski & bi))
    }
    unsafe fn castps_si(a: Self::Vf32) -> Self::Vi32 {
        mem::transmute::<f32, i32>(a)
    }
    unsafe fn castsi_ps(a: Self::Vi32) -> Self::Vf32 {
        mem::transmute::<i32, f32>(a)
    }
    unsafe fn ceil_ps(a: Self::Vf32) -> Self::Vf32 {
        a.ceil()
    }
    unsafe fn cmpeq_epi32(a: Self::Vi32, b: Self::Vi32) -> Self::Vi32 {
        mem::transmute::<bool, i8>(a == b) as i32 * -1
    }
    unsafe fn cmpge_ps(a: Self::Vf32, b: Self::Vf32) -> Self::Vf32 {
        mem::transmute::<i32, f32>(mem::transmute::<bool, i8>(a >= b) as i32 * -1)
    }
    unsafe fn cmpgt_epi32(a: Self::Vi32, b: Self::Vi32) -> Self::Vi32 {
        mem::transmute::<bool, i8>(a > b) as i32 * -1
    }
    unsafe fn cmpgt_ps(a: Self::Vf32, b: Self::Vf32) -> Self::Vf32 {
        mem::transmute::<i32, f32>(mem::transmute::<bool, i8>(a > b) as i32 * -1)
    }
    unsafe fn cmplt_ps(a: Self::Vf32, b: Self::Vf32) -> Self::Vf32 {
        mem::transmute::<i32, f32>(mem::transmute::<bool, i8>(a < b) as i32 * -1)
    }
    unsafe fn cvtepi32_ps(a: Self::Vi32) -> Self::Vf32 {
        a as f32
    }
    unsafe fn cvtps_epi32(a: Self::Vf32) -> Self::Vi32 {
        a as i32
    }
    unsafe fn floor_ps(a: Self::Vf32) -> Self::Vf32 {
        a.floor()
    }
    unsafe fn fastfloor_ps(a: Self::Vf32) -> Self::Vf32 {
        a.floor()
    }
    unsafe fn fmadd_ps(a: Self::Vf32, b: Self::Vf32, c: Self::Vf32) -> Self::Vf32 {
        a * b + c
    }
    unsafe fn fnmadd_ps(a: Self::Vf32, b: Self::Vf32, c: Self::Vf32) -> Self::Vf32 {
        (-(a * b)) + c
    }
    unsafe fn i32gather_epi32(arr: &[i32], index: Self::Vi32) -> Self::Vi32 {
        arr[index as usize]
    }
    unsafe fn i32gather_ps(arr: &[f32], index: Self::Vi32) -> Self::Vf32 {
        arr[index as usize]
    }
    unsafe fn loadu_ps(a: &f32) -> Self::Vf32 {
        *a
    }
    unsafe fn loadu_si(a: &i32) -> Self::Vi32 {
        *a
    }
    unsafe fn storeu_ps(a: &mut f32, b: Self::Vf32) {
        *a = b;
    }
    unsafe fn max_ps(a: Self::Vf32, b: Self::Vf32) -> Self::Vf32 {
        if a > b {
            a
        } else {
            b
        }
    }
    unsafe fn min_ps(a: Self::Vf32, b: Self::Vf32) -> Self::Vf32 {
        if a < b {
            a
        } else {
            b
        }
    }
    unsafe fn mul_ps(a: Self::Vf32, b: Self::Vf32) -> Self::Vf32 {
        a * b
    }
    unsafe fn div_ps(a: Self::Vf32, b: Self::Vf32) -> Self::Vf32 {
        a / b
    }
    unsafe fn mullo_epi32(a: Self::Vi32, b: Self::Vi32) -> Self::Vi32 {
        a * b
    }
    unsafe fn or_si(a: Self::Vi32, b: Self::Vi32) -> Self::Vi32 {
        a | b
    }
    unsafe fn round_ps(a: Self::Vf32) -> Self::Vf32 {
        a.round()
    }
    unsafe fn set1_epi32(a: i32) -> Self::Vi32 {
        a
    }
    unsafe fn set1_ps(a: f32) -> Self::Vf32 {
        a
    }
    unsafe fn setzero_ps() -> Self::Vf32 {
        0.0
    }
    unsafe fn setzero_si() -> Self::Vi32 {
        0
    }
    unsafe fn srai_epi32(a: Self::Vi32, imm8: i32) -> Self::Vi32 {
        a >> imm8
    }
    unsafe fn sub_epi32(a: Self::Vi32, b: Self::Vi32) -> Self::Vi32 {
        a - b
    }
    unsafe fn sub_ps(a: Self::Vf32, b: Self::Vf32) -> Self::Vf32 {
        a - b
    }
    unsafe fn xor_si(a: Self::Vi32, b: Self::Vi32) -> Self::Vi32 {
        a ^ b
    }
}