rscrypto 0.1.0

Rust crypto with zero default deps: BLAKE3, Ed25519/X25519, hashes, MACs, KDFs, AEADs, and checksums with SIMD/ASM acceleration.
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

//! Blake2b POWER VSX-accelerated compression for powerpc64.
//!
//! This backend keeps the working state in canonical `u64` words and uses
//! `core::simd::u64x2` row pairs plus lane swizzles for diagonalization. That
//! avoids any dependence on POWER byte-permute register layout, so the kernel
//! is correct on both big- and little-endian POWER while still mapping cleanly
//! to VSX-sized 128-bit vectors.

#![allow(clippy::indexing_slicing)]

use core::simd::u64x2;

use super::kernels::{SIGMA, init_v, load_msg};

type RowPair = [u64x2; 2];

#[inline(always)]
fn rotr64<const N: u32>(v: u64x2) -> u64x2 {
  const { assert!(N > 0 && N < 64) }
  let s0 = u64x2::splat(N as u64);
  let s1 = u64x2::splat((64 - N) as u64);
  (v >> s0) | (v << s1)
}

#[inline(always)]
fn load_msg_pair(m: &[u64; 16], i0: u8, i1: u8) -> u64x2 {
  u64x2::from_array([m[i0 as usize], m[i1 as usize]])
}

#[inline(always)]
fn pair_a1_b0(a: u64x2, b: u64x2) -> u64x2 {
  core::simd::simd_swizzle!(a, b, [1, 2])
}

#[inline(always)]
fn pair_b1_a0(a: u64x2, b: u64x2) -> u64x2 {
  core::simd::simd_swizzle!(a, b, [3, 0])
}

#[inline(always)]
fn g2(a: &mut RowPair, b: &mut RowPair, c: &mut RowPair, d: &mut RowPair, mx: RowPair, my: RowPair) {
  a[0] += b[0];
  a[0] += mx[0];
  a[1] += b[1];
  a[1] += mx[1];
  d[0] ^= a[0];
  d[1] ^= a[1];
  d[0] = rotr64::<32>(d[0]);
  d[1] = rotr64::<32>(d[1]);
  c[0] += d[0];
  c[1] += d[1];
  b[0] ^= c[0];
  b[1] ^= c[1];
  b[0] = rotr64::<24>(b[0]);
  b[1] = rotr64::<24>(b[1]);
  a[0] += b[0];
  a[0] += my[0];
  a[1] += b[1];
  a[1] += my[1];
  d[0] ^= a[0];
  d[1] ^= a[1];
  d[0] = rotr64::<16>(d[0]);
  d[1] = rotr64::<16>(d[1]);
  c[0] += d[0];
  c[1] += d[1];
  b[0] ^= c[0];
  b[1] ^= c[1];
  b[0] = rotr64::<63>(b[0]);
  b[1] = rotr64::<63>(b[1]);
}

#[inline(always)]
fn diagonalize(b: &mut RowPair, c: &mut RowPair, d: &mut RowPair) {
  let tb0 = b[0];
  let tb1 = b[1];
  b[0] = pair_a1_b0(tb0, tb1);
  b[1] = pair_b1_a0(tb0, tb1);
  c.swap(0, 1);
  let td0 = d[0];
  let td1 = d[1];
  d[0] = pair_b1_a0(td0, td1);
  d[1] = pair_a1_b0(td0, td1);
}

#[inline(always)]
fn undiagonalize(b: &mut RowPair, c: &mut RowPair, d: &mut RowPair) {
  let tb0 = b[0];
  let tb1 = b[1];
  b[0] = pair_b1_a0(tb0, tb1);
  b[1] = pair_a1_b0(tb0, tb1);
  c.swap(0, 1);
  let td0 = d[0];
  let td1 = d[1];
  d[0] = pair_a1_b0(td0, td1);
  d[1] = pair_b1_a0(td0, td1);
}

/// Blake2b POWER VSX-accelerated compress.
///
/// # Safety
///
/// Caller must ensure POWER8+ with VSX is available.
#[target_feature(enable = "vsx")]
pub(super) unsafe fn compress_vsx(h: &mut [u64; 8], block: &[u8; 128], t: u128, last: bool) {
  let m = load_msg(block);
  let v = init_v(h, t, last);

  let mut a = [u64x2::from_array([v[0], v[1]]), u64x2::from_array([v[2], v[3]])];
  let mut b = [u64x2::from_array([v[4], v[5]]), u64x2::from_array([v[6], v[7]])];
  let mut c = [u64x2::from_array([v[8], v[9]]), u64x2::from_array([v[10], v[11]])];
  let mut d = [u64x2::from_array([v[12], v[13]]), u64x2::from_array([v[14], v[15]])];

  for round in 0..12u8 {
    let s = &SIGMA[(round % 10) as usize];
    let mx = [load_msg_pair(&m, s[0], s[2]), load_msg_pair(&m, s[4], s[6])];
    let my = [load_msg_pair(&m, s[1], s[3]), load_msg_pair(&m, s[5], s[7])];
    g2(&mut a, &mut b, &mut c, &mut d, mx, my);
    diagonalize(&mut b, &mut c, &mut d);

    let mx = [load_msg_pair(&m, s[8], s[10]), load_msg_pair(&m, s[12], s[14])];
    let my = [load_msg_pair(&m, s[9], s[11]), load_msg_pair(&m, s[13], s[15])];
    g2(&mut a, &mut b, &mut c, &mut d, mx, my);
    undiagonalize(&mut b, &mut c, &mut d);
  }

  let h0 = u64x2::from_array([h[0], h[1]]);
  let h1 = u64x2::from_array([h[2], h[3]]);
  let h2 = u64x2::from_array([h[4], h[5]]);
  let h3 = u64x2::from_array([h[6], h[7]]);
  let r0 = h0 ^ a[0] ^ c[0];
  let r1 = h1 ^ a[1] ^ c[1];
  let r2 = h2 ^ b[0] ^ d[0];
  let r3 = h3 ^ b[1] ^ d[1];
  h[..2].copy_from_slice(&r0.to_array());
  h[2..4].copy_from_slice(&r1.to_array());
  h[4..6].copy_from_slice(&r2.to_array());
  h[6..].copy_from_slice(&r3.to_array());
}

#[cfg(test)]
mod tests {
  use core::simd::u64x2;

  use super::{RowPair, diagonalize, pair_a1_b0, pair_b1_a0, undiagonalize};

  #[test]
  fn lane_selectors_match_expected_pairs() {
    if !crate::platform::caps().has(crate::platform::caps::power::VSX) {
      return;
    }

    let a = u64x2::from_array([10, 11]);
    let b = u64x2::from_array([20, 21]);
    assert_eq!(pair_a1_b0(a, b).to_array(), [11, 20]);
    assert_eq!(pair_b1_a0(a, b).to_array(), [21, 10]);
  }

  #[test]
  fn diagonalize_round_trip_restores_rows() {
    if !crate::platform::caps().has(crate::platform::caps::power::VSX) {
      return;
    }

    let mut b: RowPair = [u64x2::from_array([4, 5]), u64x2::from_array([6, 7])];
    let mut c: RowPair = [u64x2::from_array([8, 9]), u64x2::from_array([10, 11])];
    let mut d: RowPair = [u64x2::from_array([12, 13]), u64x2::from_array([14, 15])];
    let original = (b, c, d);

    diagonalize(&mut b, &mut c, &mut d);
    assert_eq!(b[0].to_array(), [5, 6]);
    assert_eq!(b[1].to_array(), [7, 4]);
    assert_eq!(c[0].to_array(), [10, 11]);
    assert_eq!(c[1].to_array(), [8, 9]);
    assert_eq!(d[0].to_array(), [15, 12]);
    assert_eq!(d[1].to_array(), [13, 14]);

    undiagonalize(&mut b, &mut c, &mut d);
    assert_eq!((b, c, d), original);
  }
}