p256/arithmetic/scalar/
scalar64.rs

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
//! 64-bit secp256r1 scalar field algorithms.

use super::{MODULUS, MU};
use crate::{
    arithmetic::util::{adc, mac, sbb},
    U256,
};

/// Barrett Reduction
///
/// The general algorithm is:
/// ```text
/// p = n = order of group
/// b = 2^64 = 64bit machine word
/// k = 4
/// a \in [0, 2^512]
/// mu := floor(b^{2k} / p)
/// q1 := floor(a / b^{k - 1})
/// q2 := q1 * mu
/// q3 := <- floor(a / b^{k - 1})
/// r1 := a mod b^{k + 1}
/// r2 := q3 * m mod b^{k + 1}
/// r := r1 - r2
///
/// if r < 0: r := r + b^{k + 1}
/// while r >= p: do r := r - p (at most twice)
/// ```
///
/// References:
/// - Handbook of Applied Cryptography, Chapter 14
///   Algorithm 14.42
///   http://cacr.uwaterloo.ca/hac/about/chap14.pdf
///
/// - Efficient and Secure Elliptic Curve Cryptography Implementation of Curve P-256
///   Algorithm 6) Barrett Reduction modulo p
///   https://csrc.nist.gov/csrc/media/events/workshop-on-elliptic-curve-cryptography-standards/documents/papers/session6-adalier-mehmet.pdf
#[inline]
#[allow(clippy::too_many_arguments)]
pub(super) const fn barrett_reduce(lo: U256, hi: U256) -> U256 {
    let lo = lo.as_words();
    let hi = hi.as_words();
    let a0 = lo[0];
    let a1 = lo[1];
    let a2 = lo[2];
    let a3 = lo[3];
    let a4 = hi[0];
    let a5 = hi[1];
    let a6 = hi[2];
    let a7 = hi[3];
    let q1: [u64; 5] = [a3, a4, a5, a6, a7];
    let q3 = q1_times_mu_shift_five(&q1);

    let r1: [u64; 5] = [a0, a1, a2, a3, a4];
    let r2: [u64; 5] = q3_times_n_keep_five(&q3);
    let r: [u64; 5] = sub_inner_five(r1, r2);

    // Result is in range (0, 3*n - 1),
    // and 90% of the time, no subtraction will be needed.
    let r = subtract_n_if_necessary(r[0], r[1], r[2], r[3], r[4]);
    let r = subtract_n_if_necessary(r[0], r[1], r[2], r[3], r[4]);
    U256::from_words([r[0], r[1], r[2], r[3]])
}

const fn q1_times_mu_shift_five(q1: &[u64; 5]) -> [u64; 5] {
    // Schoolbook multiplication.

    let (_w0, carry) = mac(0, q1[0], MU[0], 0);
    let (w1, carry) = mac(0, q1[0], MU[1], carry);
    let (w2, carry) = mac(0, q1[0], MU[2], carry);
    let (w3, carry) = mac(0, q1[0], MU[3], carry);
    let (w4, w5) = mac(0, q1[0], MU[4], carry);

    let (_w1, carry) = mac(w1, q1[1], MU[0], 0);
    let (w2, carry) = mac(w2, q1[1], MU[1], carry);
    let (w3, carry) = mac(w3, q1[1], MU[2], carry);
    let (w4, carry) = mac(w4, q1[1], MU[3], carry);
    let (w5, w6) = mac(w5, q1[1], MU[4], carry);

    let (_w2, carry) = mac(w2, q1[2], MU[0], 0);
    let (w3, carry) = mac(w3, q1[2], MU[1], carry);
    let (w4, carry) = mac(w4, q1[2], MU[2], carry);
    let (w5, carry) = mac(w5, q1[2], MU[3], carry);
    let (w6, w7) = mac(w6, q1[2], MU[4], carry);

    let (_w3, carry) = mac(w3, q1[3], MU[0], 0);
    let (w4, carry) = mac(w4, q1[3], MU[1], carry);
    let (w5, carry) = mac(w5, q1[3], MU[2], carry);
    let (w6, carry) = mac(w6, q1[3], MU[3], carry);
    let (w7, w8) = mac(w7, q1[3], MU[4], carry);

    let (_w4, carry) = mac(w4, q1[4], MU[0], 0);
    let (w5, carry) = mac(w5, q1[4], MU[1], carry);
    let (w6, carry) = mac(w6, q1[4], MU[2], carry);
    let (w7, carry) = mac(w7, q1[4], MU[3], carry);
    let (w8, w9) = mac(w8, q1[4], MU[4], carry);

    // let q2 = [_w0, _w1, _w2, _w3, _w4, w5, w6, w7, w8, w9];
    [w5, w6, w7, w8, w9]
}

const fn q3_times_n_keep_five(q3: &[u64; 5]) -> [u64; 5] {
    // Schoolbook multiplication.

    let modulus = MODULUS.as_words();

    let (w0, carry) = mac(0, q3[0], modulus[0], 0);
    let (w1, carry) = mac(0, q3[0], modulus[1], carry);
    let (w2, carry) = mac(0, q3[0], modulus[2], carry);
    let (w3, carry) = mac(0, q3[0], modulus[3], carry);
    let (w4, _) = mac(0, q3[0], 0, carry);

    let (w1, carry) = mac(w1, q3[1], modulus[0], 0);
    let (w2, carry) = mac(w2, q3[1], modulus[1], carry);
    let (w3, carry) = mac(w3, q3[1], modulus[2], carry);
    let (w4, _) = mac(w4, q3[1], modulus[3], carry);

    let (w2, carry) = mac(w2, q3[2], modulus[0], 0);
    let (w3, carry) = mac(w3, q3[2], modulus[1], carry);
    let (w4, _) = mac(w4, q3[2], modulus[2], carry);

    let (w3, carry) = mac(w3, q3[3], modulus[0], 0);
    let (w4, _) = mac(w4, q3[3], modulus[1], carry);

    let (w4, _) = mac(w4, q3[4], modulus[0], 0);

    [w0, w1, w2, w3, w4]
}

#[inline]
#[allow(clippy::too_many_arguments)]
const fn sub_inner_five(l: [u64; 5], r: [u64; 5]) -> [u64; 5] {
    let (w0, borrow) = sbb(l[0], r[0], 0);
    let (w1, borrow) = sbb(l[1], r[1], borrow);
    let (w2, borrow) = sbb(l[2], r[2], borrow);
    let (w3, borrow) = sbb(l[3], r[3], borrow);
    let (w4, _borrow) = sbb(l[4], r[4], borrow);

    // If underflow occurred on the final limb - don't care (= add b^{k+1}).
    [w0, w1, w2, w3, w4]
}

#[inline]
#[allow(clippy::too_many_arguments)]
const fn subtract_n_if_necessary(r0: u64, r1: u64, r2: u64, r3: u64, r4: u64) -> [u64; 5] {
    let modulus = MODULUS.as_words();

    let (w0, borrow) = sbb(r0, modulus[0], 0);
    let (w1, borrow) = sbb(r1, modulus[1], borrow);
    let (w2, borrow) = sbb(r2, modulus[2], borrow);
    let (w3, borrow) = sbb(r3, modulus[3], borrow);
    let (w4, borrow) = sbb(r4, 0, borrow);

    // If underflow occurred on the final limb, borrow = 0xfff...fff, otherwise
    // borrow = 0x000...000. Thus, we use it as a mask to conditionally add the
    // modulus.
    let (w0, carry) = adc(w0, modulus[0] & borrow, 0);
    let (w1, carry) = adc(w1, modulus[1] & borrow, carry);
    let (w2, carry) = adc(w2, modulus[2] & borrow, carry);
    let (w3, carry) = adc(w3, modulus[3] & borrow, carry);
    let (w4, _carry) = adc(w4, 0, carry);

    [w0, w1, w2, w3, w4]
}