sha2 0.11.1

Pure Rust implementation of the SHA-2 hash function family including SHA-224, SHA-256, SHA-384, and SHA-512.
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
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198

// Implementation adapted from mbedtls.
use core::arch::aarch64::*;

#[cfg(not(target_arch = "aarch64"))]
compile_error!("aarch64-sha3 backend can be used only on aarch64 target arches");

#[target_feature(enable = "sha3")]
pub(super) fn compress(state: &mut [u64; 8], blocks: &[[u8; 128]]) {
    // Load state into vectors.
    let [mut ab, mut cd, mut ef, mut gh] = load_state(state);

    // Iterate through the message blocks.
    for block in blocks {
        // Keep original state values.
        let ab_orig = ab;
        let cd_orig = cd;
        let ef_orig = ef;
        let gh_orig = gh;

        // Load the message block into vectors, assuming little endianness.
        let [
            mut s0,
            mut s1,
            mut s2,
            mut s3,
            mut s4,
            mut s5,
            mut s6,
            mut s7,
        ] = load_block(block);

        // Rounds 0 and 1
        let mut initial_sum = vaddq_u64(s0, rk(0));
        let mut sum = vaddq_u64(vextq_u64(initial_sum, initial_sum, 1), gh);
        let mut intermed = vsha512hq_u64(sum, vextq_u64(ef, gh, 1), vextq_u64(cd, ef, 1));
        gh = vsha512h2q_u64(intermed, cd, ab);
        cd = vaddq_u64(cd, intermed);

        // Rounds 2 and 3
        initial_sum = vaddq_u64(s1, rk(2));
        sum = vaddq_u64(vextq_u64(initial_sum, initial_sum, 1), ef);
        intermed = vsha512hq_u64(sum, vextq_u64(cd, ef, 1), vextq_u64(ab, cd, 1));
        ef = vsha512h2q_u64(intermed, ab, gh);
        ab = vaddq_u64(ab, intermed);

        // Rounds 4 and 5
        initial_sum = vaddq_u64(s2, rk(4));
        sum = vaddq_u64(vextq_u64(initial_sum, initial_sum, 1), cd);
        intermed = vsha512hq_u64(sum, vextq_u64(ab, cd, 1), vextq_u64(gh, ab, 1));
        cd = vsha512h2q_u64(intermed, gh, ef);
        gh = vaddq_u64(gh, intermed);

        // Rounds 6 and 7
        initial_sum = vaddq_u64(s3, rk(6));
        sum = vaddq_u64(vextq_u64(initial_sum, initial_sum, 1), ab);
        intermed = vsha512hq_u64(sum, vextq_u64(gh, ab, 1), vextq_u64(ef, gh, 1));
        ab = vsha512h2q_u64(intermed, ef, cd);
        ef = vaddq_u64(ef, intermed);

        // Rounds 8 and 9
        initial_sum = vaddq_u64(s4, rk(8));
        sum = vaddq_u64(vextq_u64(initial_sum, initial_sum, 1), gh);
        intermed = vsha512hq_u64(sum, vextq_u64(ef, gh, 1), vextq_u64(cd, ef, 1));
        gh = vsha512h2q_u64(intermed, cd, ab);
        cd = vaddq_u64(cd, intermed);

        // Rounds 10 and 11
        initial_sum = vaddq_u64(s5, rk(10));
        sum = vaddq_u64(vextq_u64(initial_sum, initial_sum, 1), ef);
        intermed = vsha512hq_u64(sum, vextq_u64(cd, ef, 1), vextq_u64(ab, cd, 1));
        ef = vsha512h2q_u64(intermed, ab, gh);
        ab = vaddq_u64(ab, intermed);

        // Rounds 12 and 13
        initial_sum = vaddq_u64(s6, rk(12));
        sum = vaddq_u64(vextq_u64(initial_sum, initial_sum, 1), cd);
        intermed = vsha512hq_u64(sum, vextq_u64(ab, cd, 1), vextq_u64(gh, ab, 1));
        cd = vsha512h2q_u64(intermed, gh, ef);
        gh = vaddq_u64(gh, intermed);

        // Rounds 14 and 15
        initial_sum = vaddq_u64(s7, rk(14));
        sum = vaddq_u64(vextq_u64(initial_sum, initial_sum, 1), ab);
        intermed = vsha512hq_u64(sum, vextq_u64(gh, ab, 1), vextq_u64(ef, gh, 1));
        ab = vsha512h2q_u64(intermed, ef, cd);
        ef = vaddq_u64(ef, intermed);

        for t in (16..80).step_by(16) {
            // Rounds t and t + 1
            s0 = vsha512su1q_u64(vsha512su0q_u64(s0, s1), s7, vextq_u64(s4, s5, 1));
            initial_sum = vaddq_u64(s0, rk(t));
            sum = vaddq_u64(vextq_u64(initial_sum, initial_sum, 1), gh);
            intermed = vsha512hq_u64(sum, vextq_u64(ef, gh, 1), vextq_u64(cd, ef, 1));
            gh = vsha512h2q_u64(intermed, cd, ab);
            cd = vaddq_u64(cd, intermed);

            // Rounds t + 2 and t + 3
            s1 = vsha512su1q_u64(vsha512su0q_u64(s1, s2), s0, vextq_u64(s5, s6, 1));
            initial_sum = vaddq_u64(s1, rk(t + 2));
            sum = vaddq_u64(vextq_u64(initial_sum, initial_sum, 1), ef);
            intermed = vsha512hq_u64(sum, vextq_u64(cd, ef, 1), vextq_u64(ab, cd, 1));
            ef = vsha512h2q_u64(intermed, ab, gh);
            ab = vaddq_u64(ab, intermed);

            // Rounds t + 4 and t + 5
            s2 = vsha512su1q_u64(vsha512su0q_u64(s2, s3), s1, vextq_u64(s6, s7, 1));
            initial_sum = vaddq_u64(s2, rk(t + 4));
            sum = vaddq_u64(vextq_u64(initial_sum, initial_sum, 1), cd);
            intermed = vsha512hq_u64(sum, vextq_u64(ab, cd, 1), vextq_u64(gh, ab, 1));
            cd = vsha512h2q_u64(intermed, gh, ef);
            gh = vaddq_u64(gh, intermed);

            // Rounds t + 6 and t + 7
            s3 = vsha512su1q_u64(vsha512su0q_u64(s3, s4), s2, vextq_u64(s7, s0, 1));
            initial_sum = vaddq_u64(s3, rk(t + 6));
            sum = vaddq_u64(vextq_u64(initial_sum, initial_sum, 1), ab);
            intermed = vsha512hq_u64(sum, vextq_u64(gh, ab, 1), vextq_u64(ef, gh, 1));
            ab = vsha512h2q_u64(intermed, ef, cd);
            ef = vaddq_u64(ef, intermed);

            // Rounds t + 8 and t + 9
            s4 = vsha512su1q_u64(vsha512su0q_u64(s4, s5), s3, vextq_u64(s0, s1, 1));
            initial_sum = vaddq_u64(s4, rk(t + 8));
            sum = vaddq_u64(vextq_u64(initial_sum, initial_sum, 1), gh);
            intermed = vsha512hq_u64(sum, vextq_u64(ef, gh, 1), vextq_u64(cd, ef, 1));
            gh = vsha512h2q_u64(intermed, cd, ab);
            cd = vaddq_u64(cd, intermed);

            // Rounds t + 10 and t + 11
            s5 = vsha512su1q_u64(vsha512su0q_u64(s5, s6), s4, vextq_u64(s1, s2, 1));
            initial_sum = vaddq_u64(s5, rk(t + 10));
            sum = vaddq_u64(vextq_u64(initial_sum, initial_sum, 1), ef);
            intermed = vsha512hq_u64(sum, vextq_u64(cd, ef, 1), vextq_u64(ab, cd, 1));
            ef = vsha512h2q_u64(intermed, ab, gh);
            ab = vaddq_u64(ab, intermed);

            // Rounds t + 12 and t + 13
            s6 = vsha512su1q_u64(vsha512su0q_u64(s6, s7), s5, vextq_u64(s2, s3, 1));
            initial_sum = vaddq_u64(s6, rk(t + 12));
            sum = vaddq_u64(vextq_u64(initial_sum, initial_sum, 1), cd);
            intermed = vsha512hq_u64(sum, vextq_u64(ab, cd, 1), vextq_u64(gh, ab, 1));
            cd = vsha512h2q_u64(intermed, gh, ef);
            gh = vaddq_u64(gh, intermed);

            // Rounds t + 14 and t + 15
            s7 = vsha512su1q_u64(vsha512su0q_u64(s7, s0), s6, vextq_u64(s3, s4, 1));
            initial_sum = vaddq_u64(s7, rk(t + 14));
            sum = vaddq_u64(vextq_u64(initial_sum, initial_sum, 1), ab);
            intermed = vsha512hq_u64(sum, vextq_u64(gh, ab, 1), vextq_u64(ef, gh, 1));
            ab = vsha512h2q_u64(intermed, ef, cd);
            ef = vaddq_u64(ef, intermed);
        }

        // Add the block-specific state to the original state.
        ab = vaddq_u64(ab, ab_orig);
        cd = vaddq_u64(cd, cd_orig);
        ef = vaddq_u64(ef, ef_orig);
        gh = vaddq_u64(gh, gh_orig);
    }

    // Store vectors into state.
    store_state(state, [ab, cd, ef, gh]);
}

#[inline]
#[target_feature(enable = "neon")]
fn rk(i: usize) -> uint64x2_t {
    let chunk = &crate::consts::K64[i..][..2];
    unsafe { vld1q_u64(chunk.as_ptr()) }
}

#[inline]
#[target_feature(enable = "neon")]
fn load_block(block: &[u8; 128]) -> [uint64x2_t; 8] {
    core::array::from_fn(|i| {
        let chunk = &block[16 * i..][..16];
        let b = unsafe { vld1q_u8(chunk.as_ptr()) };
        vreinterpretq_u64_u8(vrev64q_u8(b))
    })
}

#[inline]
#[target_feature(enable = "neon")]
fn load_state(state: &[u64; 8]) -> [uint64x2_t; 4] {
    core::array::from_fn(|i| {
        let chunk = &state[2 * i..][..2];
        unsafe { vld1q_u64(chunk.as_ptr()) }
    })
}

#[inline]
#[target_feature(enable = "neon")]
fn store_state(state_dst: &mut [u64; 8], state: [uint64x2_t; 4]) {
    for i in 0..4 {
        let chunk = &mut state_dst[2 * i..][..2];
        unsafe { vst1q_u64(chunk.as_mut_ptr(), state[i]) };
    }
}