rscrypto 0.1.1

Pure Rust cryptography, hardware-accelerated: BLAKE3, SHA-2/3, AES-GCM, ChaCha20-Poly1305, Ed25519, X25519, HMAC, HKDF, Argon2, CRC. no_std, WASM, ten CPU architectures.
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
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329

use core::arch::x86_64::*;

use super::{BLOCK_SIZE, C0, C1, KEY_SIZE, NONCE_SIZE, TAG_SIZE};

#[inline]
/// # Safety
///
/// `bytes` must refer to a valid 16-byte block.
unsafe fn load(bytes: &[u8; BLOCK_SIZE]) -> __m128i {
  _mm_loadu_si128(bytes.as_ptr().cast())
}

#[inline]
/// # Safety
///
/// `out` must refer to a valid writable 16-byte block.
unsafe fn store(v: __m128i, out: &mut [u8; BLOCK_SIZE]) {
  _mm_storeu_si128(out.as_mut_ptr().cast(), v);
}

// ── Register-based helpers ──────────────────────────────────────────────
//
// All 6 AES rounds in an AEGIS-256 update read from the OLD state and are
// mutually independent. Keeping state in 6 local `__m128i` values (rather
// than indexing an array through `&mut State`) lets the register allocator
// pin them to XMM registers across loop iterations, eliminating store-to-
// load round-trips that serialize the OOO pipeline.

#[inline]
/// # Safety
///
/// Caller must ensure AES-NI is available and all register arguments come
/// from a valid AEGIS-256 state.
unsafe fn update_regs(
  s0: &mut __m128i,
  s1: &mut __m128i,
  s2: &mut __m128i,
  s3: &mut __m128i,
  s4: &mut __m128i,
  s5: &mut __m128i,
  m: __m128i,
) {
  let tmp = *s5;
  *s5 = _mm_aesenc_si128(*s4, *s5);
  *s4 = _mm_aesenc_si128(*s3, *s4);
  *s3 = _mm_aesenc_si128(*s2, *s3);
  *s2 = _mm_aesenc_si128(*s1, *s2);
  *s1 = _mm_aesenc_si128(*s0, *s1);
  *s0 = _mm_xor_si128(_mm_aesenc_si128(tmp, *s0), m);
}

#[inline]
/// # Safety
///
/// The provided registers must come from a valid AEGIS-256 state on an
/// AES-NI capable CPU.
unsafe fn keystream_regs(s1: __m128i, s2: __m128i, s3: __m128i, s4: __m128i, s5: __m128i) -> __m128i {
  _mm_xor_si128(_mm_xor_si128(s1, s4), _mm_xor_si128(s5, _mm_and_si128(s2, s3)))
}

// ── Fused encrypt/decrypt ───────────────────────────────────────────────
//
// Single `#[target_feature]` entry points that keep state in XMM registers
// from init through finalize, eliminating ~15 cycles of stack spills that
// occur when init/aad/encrypt/finalize are separate function calls.

#[target_feature(enable = "aes,avx")]
/// # Safety
///
/// Caller must ensure the CPU supports `aes` and `avx`.
pub(super) unsafe fn encrypt_fused(
  key: &[u8; KEY_SIZE],
  nonce: &[u8; NONCE_SIZE],
  aad: &[u8],
  buffer: &mut [u8],
) -> [u8; TAG_SIZE] {
  // ── init ──
  let (kh0, kh1) = super::split_halves(key);
  let (nh0, nh1) = super::split_halves(nonce);
  let k0 = load(kh0);
  let k1 = load(kh1);
  let n0 = load(nh0);
  let n1 = load(nh1);
  let c0 = load(&C0);
  let c1 = load(&C1);
  let k0_xor_n0 = _mm_xor_si128(k0, n0);
  let k1_xor_n1 = _mm_xor_si128(k1, n1);
  let (mut s0, mut s1, mut s2, mut s3, mut s4, mut s5) = (
    k0_xor_n0,
    k1_xor_n1,
    c1,
    c0,
    _mm_xor_si128(k0, c0),
    _mm_xor_si128(k1, c1),
  );
  for _ in 0..4 {
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, k0);
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, k1);
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, k0_xor_n0);
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, k1_xor_n1);
  }

  // ── aad ──
  let mut offset = 0usize;
  while offset.strict_add(BLOCK_SIZE) <= aad.len() {
    let block = _mm_loadu_si128(aad.as_ptr().add(offset).cast());
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, block);
    offset = offset.strict_add(BLOCK_SIZE);
  }
  if offset < aad.len() {
    let mut pad = [0u8; BLOCK_SIZE];
    pad[..aad.len().strict_sub(offset)].copy_from_slice(&aad[offset..]);
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, load(&pad));
  }

  // ── encrypt ──
  //
  // Counted loop with raw pointer advancement: eliminates per-iteration
  // overflow-check branches (strict_add) from the hot path. The loop count
  // is pre-validated via integer division, and all intra-block offsets are
  // compile-time constants applied through ptr::add inside this unsafe block.
  let msg_len = buffer.len();
  let ptr = buffer.as_mut_ptr();
  let n_quads = msg_len / 64;
  let mut p = ptr;
  for _ in 0..n_quads {
    _mm_prefetch(p.add(256).cast::<i8>(), _MM_HINT_T0);
    let z_a = keystream_regs(s1, s2, s3, s4, s5);
    let xi_a = _mm_loadu_si128(p.cast());
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, xi_a);
    _mm_storeu_si128(p.cast(), _mm_xor_si128(xi_a, z_a));
    let z_b = keystream_regs(s1, s2, s3, s4, s5);
    let xi_b = _mm_loadu_si128(p.add(16).cast());
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, xi_b);
    _mm_storeu_si128(p.add(16).cast(), _mm_xor_si128(xi_b, z_b));
    let z_c = keystream_regs(s1, s2, s3, s4, s5);
    let xi_c = _mm_loadu_si128(p.add(32).cast());
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, xi_c);
    _mm_storeu_si128(p.add(32).cast(), _mm_xor_si128(xi_c, z_c));
    let z_d = keystream_regs(s1, s2, s3, s4, s5);
    let xi_d = _mm_loadu_si128(p.add(48).cast());
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, xi_d);
    _mm_storeu_si128(p.add(48).cast(), _mm_xor_si128(xi_d, z_d));
    p = p.add(64);
  }
  let mut remaining = msg_len.strict_sub(n_quads.strict_mul(64));
  if remaining >= 32 {
    let z_a = keystream_regs(s1, s2, s3, s4, s5);
    let xi_a = _mm_loadu_si128(p.cast());
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, xi_a);
    _mm_storeu_si128(p.cast(), _mm_xor_si128(xi_a, z_a));
    let z_b = keystream_regs(s1, s2, s3, s4, s5);
    let xi_b = _mm_loadu_si128(p.add(16).cast());
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, xi_b);
    _mm_storeu_si128(p.add(16).cast(), _mm_xor_si128(xi_b, z_b));
    p = p.add(32);
    remaining = remaining.strict_sub(32);
  }
  if remaining >= 16 {
    let z = keystream_regs(s1, s2, s3, s4, s5);
    let xi = _mm_loadu_si128(p.cast());
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, xi);
    _mm_storeu_si128(p.cast(), _mm_xor_si128(xi, z));
    remaining = remaining.strict_sub(16);
  }
  if remaining > 0 {
    let z = keystream_regs(s1, s2, s3, s4, s5);
    let tail_off = msg_len.strict_sub(remaining);
    let mut pad = [0u8; BLOCK_SIZE];
    pad[..remaining].copy_from_slice(&buffer[tail_off..]);
    let xi = load(&pad);
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, xi);
    let mut ct_bytes = [0u8; BLOCK_SIZE];
    store(_mm_xor_si128(xi, z), &mut ct_bytes);
    buffer[tail_off..].copy_from_slice(&ct_bytes[..remaining]);
  }

  // ── finalize ──
  let ad_bits = (aad.len() as u64).strict_mul(8);
  let msg_bits = (msg_len as u64).strict_mul(8);
  let len_block = _mm_set_epi64x(msg_bits as i64, ad_bits as i64);
  let t = _mm_xor_si128(s3, len_block);
  for _ in 0..7 {
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, t);
  }
  let tag_vec = _mm_xor_si128(
    _mm_xor_si128(_mm_xor_si128(s0, s1), _mm_xor_si128(s2, s3)),
    _mm_xor_si128(s4, s5),
  );
  let mut tag = [0u8; TAG_SIZE];
  store(tag_vec, &mut tag);
  tag
}

#[target_feature(enable = "aes,avx")]
/// # Safety
///
/// Caller must ensure the CPU supports `aes` and `avx`.
pub(super) unsafe fn decrypt_fused(
  key: &[u8; KEY_SIZE],
  nonce: &[u8; NONCE_SIZE],
  aad: &[u8],
  buffer: &mut [u8],
) -> [u8; TAG_SIZE] {
  // ── init ──
  let (kh0, kh1) = super::split_halves(key);
  let (nh0, nh1) = super::split_halves(nonce);
  let k0 = load(kh0);
  let k1 = load(kh1);
  let n0 = load(nh0);
  let n1 = load(nh1);
  let c0 = load(&C0);
  let c1 = load(&C1);
  let k0_xor_n0 = _mm_xor_si128(k0, n0);
  let k1_xor_n1 = _mm_xor_si128(k1, n1);
  let (mut s0, mut s1, mut s2, mut s3, mut s4, mut s5) = (
    k0_xor_n0,
    k1_xor_n1,
    c1,
    c0,
    _mm_xor_si128(k0, c0),
    _mm_xor_si128(k1, c1),
  );
  for _ in 0..4 {
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, k0);
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, k1);
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, k0_xor_n0);
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, k1_xor_n1);
  }

  // ── aad ──
  let mut offset = 0usize;
  while offset.strict_add(BLOCK_SIZE) <= aad.len() {
    let block = _mm_loadu_si128(aad.as_ptr().add(offset).cast());
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, block);
    offset = offset.strict_add(BLOCK_SIZE);
  }
  if offset < aad.len() {
    let mut pad = [0u8; BLOCK_SIZE];
    pad[..aad.len().strict_sub(offset)].copy_from_slice(&aad[offset..]);
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, load(&pad));
  }

  // ── decrypt ──
  //
  // Same counted-loop strategy as encrypt: raw pointer advancement with
  // compile-time-constant offsets, zero overflow checks in the hot path.
  let ct_len = buffer.len();
  let ptr = buffer.as_mut_ptr();
  let n_quads = ct_len / 64;
  let mut p = ptr;
  for _ in 0..n_quads {
    _mm_prefetch(p.add(256).cast::<i8>(), _MM_HINT_T0);
    let z_a = keystream_regs(s1, s2, s3, s4, s5);
    let ci_a = _mm_loadu_si128(p.cast());
    let xi_a = _mm_xor_si128(ci_a, z_a);
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, xi_a);
    _mm_storeu_si128(p.cast(), xi_a);
    let z_b = keystream_regs(s1, s2, s3, s4, s5);
    let ci_b = _mm_loadu_si128(p.add(16).cast());
    let xi_b = _mm_xor_si128(ci_b, z_b);
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, xi_b);
    _mm_storeu_si128(p.add(16).cast(), xi_b);
    let z_c = keystream_regs(s1, s2, s3, s4, s5);
    let ci_c = _mm_loadu_si128(p.add(32).cast());
    let xi_c = _mm_xor_si128(ci_c, z_c);
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, xi_c);
    _mm_storeu_si128(p.add(32).cast(), xi_c);
    let z_d = keystream_regs(s1, s2, s3, s4, s5);
    let ci_d = _mm_loadu_si128(p.add(48).cast());
    let xi_d = _mm_xor_si128(ci_d, z_d);
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, xi_d);
    _mm_storeu_si128(p.add(48).cast(), xi_d);
    p = p.add(64);
  }
  let mut remaining = ct_len.strict_sub(n_quads.strict_mul(64));
  if remaining >= 32 {
    let z_a = keystream_regs(s1, s2, s3, s4, s5);
    let ci_a = _mm_loadu_si128(p.cast());
    let xi_a = _mm_xor_si128(ci_a, z_a);
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, xi_a);
    _mm_storeu_si128(p.cast(), xi_a);
    let z_b = keystream_regs(s1, s2, s3, s4, s5);
    let ci_b = _mm_loadu_si128(p.add(16).cast());
    let xi_b = _mm_xor_si128(ci_b, z_b);
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, xi_b);
    _mm_storeu_si128(p.add(16).cast(), xi_b);
    p = p.add(32);
    remaining = remaining.strict_sub(32);
  }
  if remaining >= 16 {
    let z = keystream_regs(s1, s2, s3, s4, s5);
    let ci = _mm_loadu_si128(p.cast());
    let xi = _mm_xor_si128(ci, z);
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, xi);
    _mm_storeu_si128(p.cast(), xi);
    remaining = remaining.strict_sub(16);
  }
  if remaining > 0 {
    let z = keystream_regs(s1, s2, s3, s4, s5);
    let tail_off = ct_len.strict_sub(remaining);
    let mut pad = [0u8; BLOCK_SIZE];
    pad[..remaining].copy_from_slice(&buffer[tail_off..]);
    let mut z_bytes = [0u8; BLOCK_SIZE];
    store(z, &mut z_bytes);
    let mut pt_pad = [0u8; BLOCK_SIZE];
    for i in 0..remaining {
      pt_pad[i] = pad[i] ^ z_bytes[i];
    }
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, load(&pt_pad));
    buffer[tail_off..].copy_from_slice(&pt_pad[..remaining]);
  }

  // ── finalize ──
  let ad_bits = (aad.len() as u64).strict_mul(8);
  let ct_bits = (ct_len as u64).strict_mul(8);
  let len_block = _mm_set_epi64x(ct_bits as i64, ad_bits as i64);
  let t = _mm_xor_si128(s3, len_block);
  for _ in 0..7 {
    update_regs(&mut s0, &mut s1, &mut s2, &mut s3, &mut s4, &mut s5, t);
  }
  let tag_vec = _mm_xor_si128(
    _mm_xor_si128(_mm_xor_si128(s0, s1), _mm_xor_si128(s2, s3)),
    _mm_xor_si128(s4, s5),
  );
  let mut tag = [0u8; TAG_SIZE];
  store(tag_vec, &mut tag);
  tag
}