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

// Copyright 2020 Yevhenii Reizner
//
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.

// Based on https://github.com/Lokathor/wide (Zlib)

cfg_if::cfg_if! {
    if #[cfg(all(feature = "simd", target_feature = "sse2"))] {
        #[cfg(target_arch = "x86")]
        use core::arch::x86::*;
        #[cfg(target_arch = "x86_64")]
        use core::arch::x86_64::*;

        // unused when AVX is available
        #[cfg(not(all(feature = "simd", target_feature = "avx2")))]
        use bytemuck::cast;

        #[derive(Clone, Copy, Debug)]
        #[repr(C, align(16))]
        pub struct u32x4(__m128i);
    } else if #[cfg(all(feature = "simd", target_feature = "simd128"))] {
        use core::arch::wasm32::*;

        #[derive(Clone, Copy, Debug)]
        #[repr(C, align(16))]
        pub struct u32x4(v128);
    } else if #[cfg(all(feature = "simd", target_arch = "aarch64", target_feature = "neon"))] {
        use core::arch::aarch64::*;

        #[derive(Clone, Copy, Debug)]
        #[repr(C, align(16))]
        pub struct u32x4(uint32x4_t);
    } else {
        #[derive(Clone, Copy, Debug)]
        #[repr(C, align(16))]
        pub struct u32x4([u32; 4]);
    }
}

unsafe impl bytemuck::Zeroable for u32x4 {}
unsafe impl bytemuck::Pod for u32x4 {}

impl Default for u32x4 {
    fn default() -> Self {
        Self::splat(0)
    }
}

impl u32x4 {
    pub fn splat(n: u32) -> Self {
        bytemuck::cast([n, n, n, n])
    }

    // unused when AVX is available
    #[cfg(not(all(feature = "simd", target_feature = "avx2")))]
    pub fn cmp_eq(self, rhs: Self) -> Self {
        cfg_if::cfg_if! {
            if #[cfg(all(feature = "simd", target_feature = "sse2"))] {
                Self(unsafe { _mm_cmpeq_epi32(self.0, rhs.0) })
            } else if #[cfg(all(feature = "simd", target_feature = "simd128"))] {
                Self(u32x4_eq(self.0, rhs.0))
            } else if #[cfg(all(feature = "simd", target_arch = "aarch64", target_feature = "neon"))] {
                Self(unsafe { vceqq_u32(self.0, rhs.0) })
            } else {
                Self([
                    if self.0[0] == rhs.0[0] { u32::MAX } else { 0 },
                    if self.0[1] == rhs.0[1] { u32::MAX } else { 0 },
                    if self.0[2] == rhs.0[2] { u32::MAX } else { 0 },
                    if self.0[3] == rhs.0[3] { u32::MAX } else { 0 },
                ])
            }
        }
    }

    // unused when AVX is available
    #[cfg(not(all(feature = "simd", target_feature = "avx2")))]
    pub fn shl<const RHS: i32>(self) -> Self {
        cfg_if::cfg_if! {
            if #[cfg(all(feature = "simd", target_feature = "sse2"))] {
                let shift = cast([RHS as u64, 0]);
                Self(unsafe { _mm_sll_epi32(self.0, shift) })
            } else if #[cfg(all(feature = "simd", target_feature = "simd128"))] {
                Self(u32x4_shl(self.0, RHS as _))
            } else if #[cfg(all(feature = "simd", target_arch = "aarch64", target_feature = "neon"))] {
                Self(unsafe { vshlq_n_u32::<RHS>(self.0) })
            } else {
                let u = RHS as u64;
                Self([
                    self.0[0] << u,
                    self.0[1] << u,
                    self.0[2] << u,
                    self.0[3] << u,
                ])
            }
        }
    }

    // unused when AVX is available
    #[cfg(not(all(feature = "simd", target_feature = "avx2")))]
    pub fn shr<const RHS: i32>(self) -> Self {
        cfg_if::cfg_if! {
            if #[cfg(all(feature = "simd", target_feature = "sse2"))] {
                let shift: __m128i = cast([RHS as u64, 0]);
                Self(unsafe { _mm_srl_epi32(self.0, shift) })
            } else if #[cfg(all(feature = "simd", target_feature = "simd128"))] {
                Self(u32x4_shr(self.0, RHS as _))
            } else if #[cfg(all(feature = "simd", target_arch = "aarch64", target_feature = "neon"))] {
                Self(unsafe { vshrq_n_u32::<RHS>(self.0) })
            } else {
                let u = RHS as u64;
                Self([
                    self.0[0] >> u,
                    self.0[1] >> u,
                    self.0[2] >> u,
                    self.0[3] >> u,
                ])
            }
        }
    }
}

impl core::ops::Not for u32x4 {
    type Output = Self;

    fn not(self) -> Self {
        cfg_if::cfg_if! {
            if #[cfg(all(feature = "simd", target_feature = "sse2"))] {
                let all_bits = unsafe { _mm_set1_epi32(-1) };
                Self(unsafe { _mm_xor_si128(self.0, all_bits) })
            } else if #[cfg(all(feature = "simd", target_feature = "simd128"))] {
                Self(v128_not(self.0))
            } else if #[cfg(all(feature = "simd", target_arch = "aarch64", target_feature = "neon"))] {
                Self(unsafe { vmvnq_u32(self.0) })
            } else {
                Self([
                    !self.0[0],
                    !self.0[1],
                    !self.0[2],
                    !self.0[3],
                ])
            }
        }
    }
}

impl core::ops::Add for u32x4 {
    type Output = Self;

    fn add(self, rhs: Self) -> Self::Output {
        cfg_if::cfg_if! {
            if #[cfg(all(feature = "simd", target_feature = "sse2"))] {
                Self(unsafe { _mm_add_epi32(self.0, rhs.0) })
            } else if #[cfg(all(feature = "simd", target_feature = "simd128"))] {
                Self(u32x4_add(self.0, rhs.0))
            } else if #[cfg(all(feature = "simd", target_arch = "aarch64", target_feature = "neon"))] {
                Self(unsafe { vaddq_u32(self.0, rhs.0) })
            } else {
                Self([
                    self.0[0].wrapping_add(rhs.0[0]),
                    self.0[1].wrapping_add(rhs.0[1]),
                    self.0[2].wrapping_add(rhs.0[2]),
                    self.0[3].wrapping_add(rhs.0[3]),
                ])
            }
        }
    }
}

impl core::ops::BitAnd for u32x4 {
    type Output = Self;

    fn bitand(self, rhs: Self) -> Self::Output {
        cfg_if::cfg_if! {
            if #[cfg(all(feature = "simd", target_feature = "sse2"))] {
                Self(unsafe { _mm_and_si128(self.0, rhs.0) })
            } else if #[cfg(all(feature = "simd", target_feature = "simd128"))] {
                Self(v128_and(self.0, rhs.0))
            } else if #[cfg(all(feature = "simd", target_arch = "aarch64", target_feature = "neon"))] {
                Self(unsafe { vandq_u32(self.0, rhs.0) })
            } else {
                Self([
                    self.0[0] & rhs.0[0],
                    self.0[1] & rhs.0[1],
                    self.0[2] & rhs.0[2],
                    self.0[3] & rhs.0[3],
                ])
            }
        }
    }
}