diskann-wide 0.50.1

DiskANN is a fast approximate nearest neighbor search library for high dimensional data
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

/*
 * Copyright (c) Microsoft Corporation.
 * Licensed under the MIT license.
 */

// x86 intrinsics
use std::arch::x86_64::*;

use crate::{
    BitMask,
    arch::x86_64::{
        V4,
        common::AllOnes,
        macros::{self, X86Default, X86LoadStore, X86Splat},
        v4::{i16x32_::i16x32, u8x16_::u8x16},
    },
    constant::Const,
    helpers,
    traits::{SIMDMask, SIMDMulAdd, SIMDVector},
};

////////////////////
// 8-bit unsigned //
////////////////////

macros::x86_define_register!(u8x32, __m256i, BitMask<32, V4>, u8, 32, V4);
macros::x86_define_splat!(u8x32 as i8, _mm256_set1_epi8, "avx");
macros::x86_define_default!(u8x32, _mm256_setzero_si256, "avx");
macros::x86_splitjoin!(
    u8x32,
    u8x16,
    _mm256_extracti128_si256,
    _mm256_set_m128i,
    "avx2"
);
macros::x86_zipunzip_crosslane!(
    u8x32,
    _mm256_permutexvar_epi8,
    _mm256_setr_epi8(
        0, 16, 1, 17, 2, 18, 3, 19, 4, 20, 5, 21, 6, 22, 7, 23, 8, 24, 9, 25, 10, 26, 11, 27, 12,
        28, 13, 29, 14, 30, 15, 31
    ),
    _mm256_setr_epi8(
        0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 1, 3, 5, 7, 9, 11, 13, 15, 17,
        19, 21, 23, 25, 27, 29, 31
    )
);

impl std::ops::Mul for u8x32 {
    type Output = Self;
    #[inline(always)]
    fn mul(self, rhs: Self) -> Self::Output {
        let x: i16x32 = self.into();
        let y: i16x32 = rhs.into();
        (x * y).cast::<u8>()
    }
}

helpers::unsafe_map_binary_op!(u8x32, std::ops::Add, add, _mm256_add_epi8, "avx2");
helpers::unsafe_map_binary_op!(u8x32, std::ops::Sub, sub, _mm256_sub_epi8, "avx2");

helpers::unsafe_map_binary_op!(u8x32, std::ops::BitAnd, bitand, _mm256_and_si256, "avx2");
helpers::unsafe_map_binary_op!(u8x32, std::ops::BitOr, bitor, _mm256_or_si256, "avx2");
helpers::unsafe_map_binary_op!(u8x32, std::ops::BitXor, bitxor, _mm256_xor_si256, "avx2");

impl std::ops::Shr for u8x32 {
    type Output = Self;
    #[inline(always)]
    fn shr(self, rhs: Self) -> Self {
        let s: i16x32 = self.into();
        let rhs: i16x32 = rhs.into();
        (s.shr(rhs)).cast::<u8>()
    }
}

impl std::ops::Shl for u8x32 {
    type Output = Self;
    #[inline(always)]
    fn shl(self, rhs: Self) -> Self {
        let s: i16x32 = self.into();
        let rhs: i16x32 = rhs.into();
        (s.shl(rhs)).cast::<u8>()
    }
}

helpers::scalar_shift_by_splat!(u8x32, u8);

impl std::ops::Not for u8x32 {
    type Output = Self;
    #[inline(always)]
    fn not(self) -> Self {
        self ^ Self::from_underlying(self.arch(), <Self as SIMDVector>::Underlying::all_ones())
    }
}

impl SIMDMulAdd for u8x32 {
    #[inline(always)]
    fn mul_add_simd(self, rhs: Self, accumulator: Self) -> Self {
        self * rhs + accumulator
    }
}

macros::x86_avx512_load_store!(
    u8x32,
    _mm256_loadu_epi8,
    _mm256_maskz_loadu_epi8,
    _mm256_storeu_epi8,
    _mm256_mask_storeu_epi8,
    i8,
    "avx512bw,avx512vl"
);

macros::x86_avx512_int_comparisons!(u8x32, _mm256_cmp_epu8_mask, "avx512bw,avx512vl");

///////////
// Tests //
///////////

#[cfg(test)]
mod test_x86_u8 {
    use super::*;
    use crate::{reference::ReferenceScalarOps, test_utils};

    #[test]
    fn miri_test_load() {
        if let Some(arch) = V4::new_checked_uncached() {
            test_utils::test_load_simd::<u8, 32, u8x32>(arch);
        }
    }

    #[test]
    fn miri_test_store() {
        if let Some(arch) = V4::new_checked_uncached() {
            test_utils::test_store_simd::<u8, 32, u8x32>(arch);
        }
    }

    // constructors
    #[test]
    fn test_constructors() {
        if let Some(arch) = V4::new_checked_uncached() {
            test_utils::ops::test_splat::<u8, 32, u8x32>(arch);
        }
    }

    // Ops
    test_utils::ops::test_add!(u8x32, 0x3017fd73c99cc633, V4::new_checked_uncached());
    test_utils::ops::test_sub!(u8x32, 0xfc627f10b5f8db8a, V4::new_checked_uncached());
    test_utils::ops::test_mul!(u8x32, 0x0f4caa80eceaa523, V4::new_checked_uncached());
    test_utils::ops::test_fma!(u8x32, 0xb8f702ba85375041, V4::new_checked_uncached());

    test_utils::ops::test_splitjoin!(u8x32 => u8x16, 0x475a19e80c2f3977, V4::new_checked_uncached());
    test_utils::ops::test_zipunzip!(u8x32 => u8x16, 0xd4f0a2e8159b6c73, V4::new_checked_uncached());

    test_utils::ops::test_cmp!(u8x32, 0x941757bd5cc641a1, V4::new_checked_uncached());

    // Bit ops
    test_utils::ops::test_bitops!(u8x32, 0xd62d8de09f82ed4e, V4::new_checked_uncached());
}