arcium-primitives 0.4.0

Arcium primitives
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

use std::num::Wrapping;

use itertools::Itertools;

// Carry-less multiplication of two n-limb numbers.
// Output has 2.n limbs.
pub fn carry_less_mul<const LIMBS: usize>(
    a: [u64; LIMBS],
    b: [u64; LIMBS],
) -> ([u64; LIMBS], [u64; LIMBS]) {
    let mut res = [[0u64; LIMBS]; 2];
    let slice = res.as_flattened_mut();

    (0..a.len())
        .cartesian_product(0..b.len())
        .for_each(|(i, j)| clfma_u64(&mut slice[i + j..i + j + 2], a[i], b[j]));

    (res[0], res[1])
}

pub fn carry_less_mul_1limb(a: [u64; 1], b: [u64; 1]) -> ([u64; 1], [u64; 1]) {
    #[cfg(all(
        any(target_arch = "x86", target_arch = "x86_64"),
        target_feature = "pclmulqdq",
        target_feature = "sse2",
    ))]
    return clmul_u64_sse2(a, b);

    #[allow(unreachable_code)]
    carry_less_mul(a, b)
}

pub fn carry_less_mul_2limbs(a: [u64; 2], b: [u64; 2]) -> ([u64; 2], [u64; 2]) {
    #[cfg(all(
        any(target_arch = "x86", target_arch = "x86_64"),
        target_feature = "pclmulqdq",
        target_feature = "sse2",
    ))]
    return clmul_2_u64_sse2(a, b);

    #[allow(unreachable_code)]
    carry_less_mul(a, b)
}

fn clfma_u64(res: &mut [u64], a: u64, b: u64) {
    let bit_k: u64 = (-Wrapping((a) & 0x01)).0;
    res[0] ^= b & bit_k;

    for k in 1..64 {
        let bit_k: u64 = (-Wrapping((a >> k) & 0x01)).0;
        let tmp = b & bit_k;
        res[0] ^= tmp << k;
        res[1] ^= tmp >> (64 - k);
    }
}

#[cfg(all(
    any(target_arch = "x86", target_arch = "x86_64"),
    target_feature = "pclmulqdq",
    target_feature = "sse2",
))]
fn clmul_2_u64_sse2(x: [u64; 2], y: [u64; 2]) -> ([u64; 2], [u64; 2]) {
    #[cfg(target_arch = "x86")]
    use core::arch::x86::*;
    #[cfg(target_arch = "x86_64")]
    use core::arch::x86_64::*;
    use std::mem::MaybeUninit;

    unsafe {
        let a_b = _mm_loadu_si128(x.as_ptr() as *const _);
        let c_d = _mm_loadu_si128(y.as_ptr() as *const _);

        let prod_a_c = _mm_clmulepi64_si128::<0>(a_b, c_d);
        let prod_b_d = _mm_clmulepi64_si128::<17>(a_b, c_d);

        let s_ab = _mm_xor_si128(_mm_broadcastq_epi64(a_b), a_b);
        let s_cd = _mm_xor_si128(_mm_broadcastq_epi64(c_d), c_d);
        let prod_ab_cd = _mm_clmulepi64_si128::<17>(s_ab, s_cd);

        let prod_ab_cd = _mm_xor_si128(prod_ab_cd, _mm_xor_si128(prod_a_c, prod_b_d));

        let res_l = _mm_xor_si128(prod_a_c, _mm_slli_si128::<8>(prod_ab_cd));
        let res_h = _mm_xor_si128(prod_b_d, _mm_srli_si128::<8>(prod_ab_cd));

        let mut low = MaybeUninit::<[u64; 2]>::uninit();
        _mm_storeu_si128(low.as_mut_ptr() as *mut _, res_l);
        let mut high = MaybeUninit::<[u64; 2]>::uninit();
        _mm_storeu_si128(high.as_mut_ptr() as *mut _, res_h);

        (low.assume_init(), high.assume_init())
    }
}

#[cfg(all(
    any(target_arch = "x86", target_arch = "x86_64"),
    target_feature = "pclmulqdq",
    target_feature = "sse2",
))]
fn clmul_u64_sse2(x: [u64; 1], y: [u64; 1]) -> ([u64; 1], [u64; 1]) {
    #[cfg(target_arch = "x86")]
    use core::arch::x86::*;
    #[cfg(target_arch = "x86_64")]
    use core::arch::x86_64::*;
    use std::mem::MaybeUninit;

    let t = unsafe {
        let x = _mm_loadu_si64(x.as_ptr() as *const _);
        let y = _mm_loadu_si64(y.as_ptr() as *const _);

        let prod = _mm_clmulepi64_si128::<0>(x, y);

        let mut res = MaybeUninit::<[u64; 2]>::uninit();
        _mm_storeu_si128(res.as_mut_ptr() as *mut _, prod);

        res.assume_init()
    };

    ([t[0]], [t[1]])
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn test_carry_less_mul_2limbs() {
        macro_rules! prod_test_case {
            ($aval:expr, $bval:expr, $prod:expr) => {{
                let prod_comp = carry_less_mul($aval, $bval);
                assert_eq!(prod_comp, $prod);
            }};
        }

        prod_test_case!(
            [0x9f418f3bffd84bba, 0x4a7c605645afdfb1],
            [0x80b7bd91cddc5be5, 0x3a97291035e41e1f],
            (
                [0xfffa315f244b1f92, 0x288de208c77fb2f9],
                [0x94aced70da538690, 0x0f0f341c05b65c5e,]
            )
        );

        prod_test_case!(
            [0x74ef862bc1b6d333, 0x3a88103b80d97b73],
            [0x753f4846eb020b5a, 0x8f108359ea25fa8f],
            (
                [0x419f60d2de880d0e, 0xfd4d74204161d27d],
                [0xe96db1bf781f351f, 0x1c36456adc21ac7a]
            )
        );

        prod_test_case!(
            [0x6447b3dcaed62649, 0x6e4af40b2ee1b4c1],
            [0xbd7a4e12fdb29840, 0x8950f56742015f25],
            (
                [0x010b1b56e559ca40, 0xf1ff4cfe33d20957],
                [0xed77c432e4701779, 0x342437199bebda57]
            )
        );
    }
}