sha2/sha512/
soft.rs

1#![allow(clippy::many_single_char_names)]
2use crate::consts::{BLOCK_LEN, K64X2};
3use core::convert::TryInto;
4
5fn add(a: [u64; 2], b: [u64; 2]) -> [u64; 2] {
6    [a[0].wrapping_add(b[0]), a[1].wrapping_add(b[1])]
7}
8
9/// Not an intrinsic, but works like an unaligned load.
10fn sha512load(v0: [u64; 2], v1: [u64; 2]) -> [u64; 2] {
11    [v1[1], v0[0]]
12}
13
14/// Performs 2 rounds of the SHA-512 message schedule update.
15pub fn sha512_schedule_x2(v0: [u64; 2], v1: [u64; 2], v4to5: [u64; 2], v7: [u64; 2]) -> [u64; 2] {
16    // sigma 0
17    fn sigma0(x: u64) -> u64 {
18        ((x << 63) | (x >> 1)) ^ ((x << 56) | (x >> 8)) ^ (x >> 7)
19    }
20
21    // sigma 1
22    fn sigma1(x: u64) -> u64 {
23        ((x << 45) | (x >> 19)) ^ ((x << 3) | (x >> 61)) ^ (x >> 6)
24    }
25
26    let [w1, w0] = v0;
27    let [_, w2] = v1;
28    let [w10, w9] = v4to5;
29    let [w15, w14] = v7;
30
31    let w16 = sigma1(w14)
32        .wrapping_add(w9)
33        .wrapping_add(sigma0(w1))
34        .wrapping_add(w0);
35    let w17 = sigma1(w15)
36        .wrapping_add(w10)
37        .wrapping_add(sigma0(w2))
38        .wrapping_add(w1);
39
40    [w17, w16]
41}
42
43/// Performs one round of the SHA-512 message block digest.
44pub fn sha512_digest_round(
45    ae: [u64; 2],
46    bf: [u64; 2],
47    cg: [u64; 2],
48    dh: [u64; 2],
49    wk0: u64,
50) -> [u64; 2] {
51    macro_rules! big_sigma0 {
52        ($a:expr) => {
53            ($a.rotate_right(28) ^ $a.rotate_right(34) ^ $a.rotate_right(39))
54        };
55    }
56    macro_rules! big_sigma1 {
57        ($a:expr) => {
58            ($a.rotate_right(14) ^ $a.rotate_right(18) ^ $a.rotate_right(41))
59        };
60    }
61    macro_rules! bool3ary_202 {
62        ($a:expr, $b:expr, $c:expr) => {
63            $c ^ ($a & ($b ^ $c))
64        };
65    } // Choose, MD5F, SHA1C
66    macro_rules! bool3ary_232 {
67        ($a:expr, $b:expr, $c:expr) => {
68            ($a & $b) ^ ($a & $c) ^ ($b & $c)
69        };
70    } // Majority, SHA1M
71
72    let [a0, e0] = ae;
73    let [b0, f0] = bf;
74    let [c0, g0] = cg;
75    let [d0, h0] = dh;
76
77    // a round
78    let x0 = big_sigma1!(e0)
79        .wrapping_add(bool3ary_202!(e0, f0, g0))
80        .wrapping_add(wk0)
81        .wrapping_add(h0);
82    let y0 = big_sigma0!(a0).wrapping_add(bool3ary_232!(a0, b0, c0));
83    let (a1, _, _, _, e1, _, _, _) = (
84        x0.wrapping_add(y0),
85        a0,
86        b0,
87        c0,
88        x0.wrapping_add(d0),
89        e0,
90        f0,
91        g0,
92    );
93
94    [a1, e1]
95}
96
97/// Process a block with the SHA-512 algorithm.
98pub fn sha512_digest_block_u64(state: &mut [u64; 8], block: &[u64; 16]) {
99    let k = &K64X2;
100
101    macro_rules! schedule {
102        ($v0:expr, $v1:expr, $v4:expr, $v5:expr, $v7:expr) => {
103            sha512_schedule_x2($v0, $v1, sha512load($v4, $v5), $v7)
104        };
105    }
106
107    macro_rules! rounds4 {
108        ($ae:ident, $bf:ident, $cg:ident, $dh:ident, $wk0:expr, $wk1:expr) => {{
109            let [u, t] = $wk0;
110            let [w, v] = $wk1;
111
112            $dh = sha512_digest_round($ae, $bf, $cg, $dh, t);
113            $cg = sha512_digest_round($dh, $ae, $bf, $cg, u);
114            $bf = sha512_digest_round($cg, $dh, $ae, $bf, v);
115            $ae = sha512_digest_round($bf, $cg, $dh, $ae, w);
116        }};
117    }
118
119    let mut ae = [state[0], state[4]];
120    let mut bf = [state[1], state[5]];
121    let mut cg = [state[2], state[6]];
122    let mut dh = [state[3], state[7]];
123
124    // Rounds 0..20
125    let (mut w1, mut w0) = ([block[3], block[2]], [block[1], block[0]]);
126    rounds4!(ae, bf, cg, dh, add(k[0], w0), add(k[1], w1));
127    let (mut w3, mut w2) = ([block[7], block[6]], [block[5], block[4]]);
128    rounds4!(ae, bf, cg, dh, add(k[2], w2), add(k[3], w3));
129    let (mut w5, mut w4) = ([block[11], block[10]], [block[9], block[8]]);
130    rounds4!(ae, bf, cg, dh, add(k[4], w4), add(k[5], w5));
131    let (mut w7, mut w6) = ([block[15], block[14]], [block[13], block[12]]);
132    rounds4!(ae, bf, cg, dh, add(k[6], w6), add(k[7], w7));
133    let mut w8 = schedule!(w0, w1, w4, w5, w7);
134    let mut w9 = schedule!(w1, w2, w5, w6, w8);
135    rounds4!(ae, bf, cg, dh, add(k[8], w8), add(k[9], w9));
136
137    // Rounds 20..40
138    w0 = schedule!(w2, w3, w6, w7, w9);
139    w1 = schedule!(w3, w4, w7, w8, w0);
140    rounds4!(ae, bf, cg, dh, add(k[10], w0), add(k[11], w1));
141    w2 = schedule!(w4, w5, w8, w9, w1);
142    w3 = schedule!(w5, w6, w9, w0, w2);
143    rounds4!(ae, bf, cg, dh, add(k[12], w2), add(k[13], w3));
144    w4 = schedule!(w6, w7, w0, w1, w3);
145    w5 = schedule!(w7, w8, w1, w2, w4);
146    rounds4!(ae, bf, cg, dh, add(k[14], w4), add(k[15], w5));
147    w6 = schedule!(w8, w9, w2, w3, w5);
148    w7 = schedule!(w9, w0, w3, w4, w6);
149    rounds4!(ae, bf, cg, dh, add(k[16], w6), add(k[17], w7));
150    w8 = schedule!(w0, w1, w4, w5, w7);
151    w9 = schedule!(w1, w2, w5, w6, w8);
152    rounds4!(ae, bf, cg, dh, add(k[18], w8), add(k[19], w9));
153
154    // Rounds 40..60
155    w0 = schedule!(w2, w3, w6, w7, w9);
156    w1 = schedule!(w3, w4, w7, w8, w0);
157    rounds4!(ae, bf, cg, dh, add(k[20], w0), add(k[21], w1));
158    w2 = schedule!(w4, w5, w8, w9, w1);
159    w3 = schedule!(w5, w6, w9, w0, w2);
160    rounds4!(ae, bf, cg, dh, add(k[22], w2), add(k[23], w3));
161    w4 = schedule!(w6, w7, w0, w1, w3);
162    w5 = schedule!(w7, w8, w1, w2, w4);
163    rounds4!(ae, bf, cg, dh, add(k[24], w4), add(k[25], w5));
164    w6 = schedule!(w8, w9, w2, w3, w5);
165    w7 = schedule!(w9, w0, w3, w4, w6);
166    rounds4!(ae, bf, cg, dh, add(k[26], w6), add(k[27], w7));
167    w8 = schedule!(w0, w1, w4, w5, w7);
168    w9 = schedule!(w1, w2, w5, w6, w8);
169    rounds4!(ae, bf, cg, dh, add(k[28], w8), add(k[29], w9));
170
171    // Rounds 60..80
172    w0 = schedule!(w2, w3, w6, w7, w9);
173    w1 = schedule!(w3, w4, w7, w8, w0);
174    rounds4!(ae, bf, cg, dh, add(k[30], w0), add(k[31], w1));
175    w2 = schedule!(w4, w5, w8, w9, w1);
176    w3 = schedule!(w5, w6, w9, w0, w2);
177    rounds4!(ae, bf, cg, dh, add(k[32], w2), add(k[33], w3));
178    w4 = schedule!(w6, w7, w0, w1, w3);
179    w5 = schedule!(w7, w8, w1, w2, w4);
180    rounds4!(ae, bf, cg, dh, add(k[34], w4), add(k[35], w5));
181    w6 = schedule!(w8, w9, w2, w3, w5);
182    w7 = schedule!(w9, w0, w3, w4, w6);
183    rounds4!(ae, bf, cg, dh, add(k[36], w6), add(k[37], w7));
184    w8 = schedule!(w0, w1, w4, w5, w7);
185    w9 = schedule!(w1, w2, w5, w6, w8);
186    rounds4!(ae, bf, cg, dh, add(k[38], w8), add(k[39], w9));
187
188    let [a, e] = ae;
189    let [b, f] = bf;
190    let [c, g] = cg;
191    let [d, h] = dh;
192
193    state[0] = state[0].wrapping_add(a);
194    state[1] = state[1].wrapping_add(b);
195    state[2] = state[2].wrapping_add(c);
196    state[3] = state[3].wrapping_add(d);
197    state[4] = state[4].wrapping_add(e);
198    state[5] = state[5].wrapping_add(f);
199    state[6] = state[6].wrapping_add(g);
200    state[7] = state[7].wrapping_add(h);
201}
202
203pub fn compress(state: &mut [u64; 8], blocks: &[[u8; 128]]) {
204    let mut block_u32 = [0u64; BLOCK_LEN];
205    // since LLVM can't properly use aliasing yet it will make
206    // unnecessary state stores without this copy
207    let mut state_cpy = *state;
208    for block in blocks {
209        for (o, chunk) in block_u32.iter_mut().zip(block.chunks_exact(8)) {
210            *o = u64::from_be_bytes(chunk.try_into().unwrap());
211        }
212        sha512_digest_block_u64(&mut state_cpy, &block_u32);
213    }
214    *state = state_cpy;
215}
sha2/sha512/soft.rs

sha2/sha512/
soft.rs
