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
// generated source. do not edit.
#![allow(non_upper_case_globals, unused_macros, unused_imports)]
use crate::low::macros::*;
// Copyright (c) 2024 The mlkem-native project authors
// Copyright (c) 2021-2022 Arm Limited
// Copyright (c) 2022 Matthias Kannwischer
// Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
// SPDX-License-Identifier: Apache-2.0 OR ISC OR MIT
// ----------------------------------------------------------------------------
// Keccak-f1600 permutation for SHA3, batch of two independent operations
// Input a[50], rc[24]; output a[50]
//
// The input/output argument is in effect two 25-element Keccak arrays
// a[0...24] and a[25..49], which could be considered as type a[25][2].
//
// Thinking of each such input/output array as a row-major flattening of a
// 5x5 matrix of 64-bit words, this performs the Keccak-f1600 permutation,
// all 24 rounds with the distinct round constants rc[i] for each one. For
// correct operation, the input pointer rc should point at the standard
// round constants as in the specification:
//
// https://keccak.team/keccak_specs_summary.html#roundConstants
//
// This operation is at the core of SHA3 and is fully specified here:
//
// https://keccak.team/files/Keccak-reference-3.0.pdf
//
// extern void sha3_keccak2_f1600(uint64_t a[static 50],
// const uint64_t rc[static 24]);
//
// Standard ARM ABI: X0 = a, X1 = rc
// ----------------------------------------------------------------------------
/// Keccak-f1600 permutation for SHA3, batch of two independent operations
///
/// Input a[50], rc[24]; output a[50]
///
/// The input/output argument is in effect two 25-element Keccak arrays
/// a[0...24] and a[25..49], which could be considered as type a[25][2].
///
/// Thinking of each such input/output array as a row-major flattening of a
/// 5x5 matrix of 64-bit words, this performs the Keccak-f1600 permutation,
/// all 24 rounds with the distinct round constants rc[i] for each one. For
/// correct operation, the input pointer rc should point at the standard
/// round constants as in the specification:
///
/// https://keccak.team/keccak_specs_summary.html#roundConstants
///
/// This operation is at the core of SHA3 and is fully specified here:
///
/// https://keccak.team/files/Keccak-reference-3.0.pdf
#[target_feature(enable = "sha3")]
pub(crate) fn sha3_keccak2_f1600(a: &mut [[u64; 25]; 2], rc: &[u64; 24]) {
// SAFETY: inline assembly. see [crate::low::inline_assembly_safety] for safety info.
unsafe {
core::arch::asm!(
// This is similar to the code in the mlkem-native repository here:
//
// mlkem/fips202/native/aarch64/src/keccak_f1600_x2_v84a_asm.S
//
// The main difference is the avoidance of ld2/st2 in favour of explicit
// transposition operations and conventional loads and stores.
Q!(" sub " "sp, sp, # (64 + 0)"),
Q!(" stp " "d8, d9, [sp, # (0 + 0)]"),
Q!(" stp " "d10, d11, [sp, # (16 + 0)]"),
Q!(" stp " "d12, d13, [sp, # (32 + 0)]"),
Q!(" stp " "d14, d15, [sp, # (48 + 0)]"),
// Load the Keccak initial state into upper and lower parts of Q0..Q24
Q!(" add " "x2, x0, #0xc8"),
Q!(" ldp " "q24, q25, [x0]"),
Q!(" ldp " "q26, q27, [x2]"),
Q!(" trn1 " "v0.2d, v24.2d, v26.2d"),
Q!(" trn2 " "v1.2d, v24.2d, v26.2d"),
Q!(" trn1 " "v2.2d, v25.2d, v27.2d"),
Q!(" trn2 " "v3.2d, v25.2d, v27.2d"),
Q!(" ldp " "q24, q25, [x0, #0x20]"),
Q!(" ldp " "q26, q27, [x2, #0x20]"),
Q!(" trn1 " "v4.2d, v24.2d, v26.2d"),
Q!(" trn2 " "v5.2d, v24.2d, v26.2d"),
Q!(" trn1 " "v6.2d, v25.2d, v27.2d"),
Q!(" trn2 " "v7.2d, v25.2d, v27.2d"),
Q!(" ldp " "q24, q25, [x0, #0x40]"),
Q!(" ldp " "q26, q27, [x2, #0x40]"),
Q!(" trn1 " "v8.2d, v24.2d, v26.2d"),
Q!(" trn2 " "v9.2d, v24.2d, v26.2d"),
Q!(" trn1 " "v10.2d, v25.2d, v27.2d"),
Q!(" trn2 " "v11.2d, v25.2d, v27.2d"),
Q!(" ldp " "q24, q25, [x0, #0x60]"),
Q!(" ldp " "q26, q27, [x2, #0x60]"),
Q!(" trn1 " "v12.2d, v24.2d, v26.2d"),
Q!(" trn2 " "v13.2d, v24.2d, v26.2d"),
Q!(" trn1 " "v14.2d, v25.2d, v27.2d"),
Q!(" trn2 " "v15.2d, v25.2d, v27.2d"),
Q!(" ldp " "q24, q25, [x0, #0x80]"),
Q!(" ldp " "q26, q27, [x2, #0x80]"),
Q!(" trn1 " "v16.2d, v24.2d, v26.2d"),
Q!(" trn2 " "v17.2d, v24.2d, v26.2d"),
Q!(" trn1 " "v18.2d, v25.2d, v27.2d"),
Q!(" trn2 " "v19.2d, v25.2d, v27.2d"),
Q!(" ldp " "q24, q25, [x0, #0xa0]"),
Q!(" ldp " "q26, q27, [x2, #0xa0]"),
Q!(" trn1 " "v20.2d, v24.2d, v26.2d"),
Q!(" trn2 " "v21.2d, v24.2d, v26.2d"),
Q!(" trn1 " "v22.2d, v25.2d, v27.2d"),
Q!(" trn2 " "v23.2d, v25.2d, v27.2d"),
Q!(" ldr " "d24, [x0, #0xc0]"),
Q!(" ldr " "d25, [x2, #0xc0]"),
Q!(" trn1 " "v24.2d, v24.2d, v25.2d"),
// Now 24 rounds of the iteration
Q!(" mov " "x2, #24"),
Q!(Label!("Lsha3_keccak2_f1600_loop", 2) ":"),
Q!(" eor3 " "v30.16b, v0.16b, v5.16b, v10.16b"),
Q!(" eor3 " "v29.16b, v1.16b, v6.16b, v11.16b"),
Q!(" eor3 " "v28.16b, v2.16b, v7.16b, v12.16b"),
Q!(" eor3 " "v27.16b, v3.16b, v8.16b, v13.16b"),
Q!(" eor3 " "v26.16b, v4.16b, v9.16b, v14.16b"),
Q!(" eor3 " "v30.16b, v30.16b, v15.16b, v20.16b"),
Q!(" eor3 " "v29.16b, v29.16b, v16.16b, v21.16b"),
Q!(" eor3 " "v28.16b, v28.16b, v17.16b, v22.16b"),
Q!(" eor3 " "v27.16b, v27.16b, v18.16b, v23.16b"),
Q!(" eor3 " "v26.16b, v26.16b, v19.16b, v24.16b"),
Q!(" rax1 " "v25.2d, v30.2d, v28.2d"),
Q!(" rax1 " "v28.2d, v28.2d, v26.2d"),
Q!(" rax1 " "v26.2d, v26.2d, v29.2d"),
Q!(" rax1 " "v29.2d, v29.2d, v27.2d"),
Q!(" rax1 " "v27.2d, v27.2d, v30.2d"),
Q!(" eor " "v30.16b, v0.16b, v26.16b"),
Q!(" xar " "v0.2d, v2.2d, v29.2d, #0x2"),
Q!(" xar " "v2.2d, v12.2d, v29.2d, #0x15"),
Q!(" xar " "v12.2d, v13.2d, v28.2d, #0x27"),
Q!(" xar " "v13.2d, v19.2d, v27.2d, #0x38"),
Q!(" xar " "v19.2d, v23.2d, v28.2d, #0x8"),
Q!(" xar " "v23.2d, v15.2d, v26.2d, #0x17"),
Q!(" xar " "v15.2d, v1.2d, v25.2d, #0x3f"),
Q!(" xar " "v1.2d, v8.2d, v28.2d, #0x9"),
Q!(" xar " "v8.2d, v16.2d, v25.2d, #0x13"),
Q!(" xar " "v16.2d, v7.2d, v29.2d, #0x3a"),
Q!(" xar " "v7.2d, v10.2d, v26.2d, #0x3d"),
Q!(" xar " "v10.2d, v3.2d, v28.2d, #0x24"),
Q!(" xar " "v3.2d, v18.2d, v28.2d, #0x2b"),
Q!(" xar " "v18.2d, v17.2d, v29.2d, #0x31"),
Q!(" xar " "v17.2d, v11.2d, v25.2d, #0x36"),
Q!(" xar " "v11.2d, v9.2d, v27.2d, #0x2c"),
Q!(" xar " "v9.2d, v22.2d, v29.2d, #0x3"),
Q!(" xar " "v22.2d, v14.2d, v27.2d, #0x19"),
Q!(" xar " "v14.2d, v20.2d, v26.2d, #0x2e"),
Q!(" xar " "v20.2d, v4.2d, v27.2d, #0x25"),
Q!(" xar " "v4.2d, v24.2d, v27.2d, #0x32"),
Q!(" xar " "v24.2d, v21.2d, v25.2d, #0x3e"),
Q!(" xar " "v21.2d, v5.2d, v26.2d, #0x1c"),
Q!(" xar " "v27.2d, v6.2d, v25.2d, #0x14"),
Q!(" ld1r " "{{ v31.2d }}, [x1], #8"),
Q!(" bcax " "v5.16b, v10.16b, v7.16b, v11.16b"),
Q!(" bcax " "v6.16b, v11.16b, v8.16b, v7.16b"),
Q!(" bcax " "v7.16b, v7.16b, v9.16b, v8.16b"),
Q!(" bcax " "v8.16b, v8.16b, v10.16b, v9.16b"),
Q!(" bcax " "v9.16b, v9.16b, v11.16b, v10.16b"),
Q!(" bcax " "v10.16b, v15.16b, v12.16b, v16.16b"),
Q!(" bcax " "v11.16b, v16.16b, v13.16b, v12.16b"),
Q!(" bcax " "v12.16b, v12.16b, v14.16b, v13.16b"),
Q!(" bcax " "v13.16b, v13.16b, v15.16b, v14.16b"),
Q!(" bcax " "v14.16b, v14.16b, v16.16b, v15.16b"),
Q!(" bcax " "v15.16b, v20.16b, v17.16b, v21.16b"),
Q!(" bcax " "v16.16b, v21.16b, v18.16b, v17.16b"),
Q!(" bcax " "v17.16b, v17.16b, v19.16b, v18.16b"),
Q!(" bcax " "v18.16b, v18.16b, v20.16b, v19.16b"),
Q!(" bcax " "v19.16b, v19.16b, v21.16b, v20.16b"),
Q!(" bcax " "v20.16b, v0.16b, v22.16b, v1.16b"),
Q!(" bcax " "v21.16b, v1.16b, v23.16b, v22.16b"),
Q!(" bcax " "v22.16b, v22.16b, v24.16b, v23.16b"),
Q!(" bcax " "v23.16b, v23.16b, v0.16b, v24.16b"),
Q!(" bcax " "v24.16b, v24.16b, v1.16b, v0.16b"),
Q!(" bcax " "v0.16b, v30.16b, v2.16b, v27.16b"),
Q!(" bcax " "v1.16b, v27.16b, v3.16b, v2.16b"),
Q!(" bcax " "v2.16b, v2.16b, v4.16b, v3.16b"),
Q!(" bcax " "v3.16b, v3.16b, v30.16b, v4.16b"),
Q!(" bcax " "v4.16b, v4.16b, v27.16b, v30.16b"),
Q!(" eor " "v0.16b, v0.16b, v31.16b"),
Q!(" sub " "x2, x2, #0x1"),
Q!(" cbnz " "x2, " Label!("Lsha3_keccak2_f1600_loop", 2, Before)),
// Store back the state
Q!(" add " "x2, x0, #0xc8"),
Q!(" trn1 " "v25.2d, v0.2d, v1.2d"),
Q!(" trn1 " "v26.2d, v2.2d, v3.2d"),
Q!(" stp " "q25, q26, [x0]"),
Q!(" trn2 " "v25.2d, v0.2d, v1.2d"),
Q!(" trn2 " "v26.2d, v2.2d, v3.2d"),
Q!(" stp " "q25, q26, [x2]"),
Q!(" trn1 " "v25.2d, v4.2d, v5.2d"),
Q!(" trn1 " "v26.2d, v6.2d, v7.2d"),
Q!(" stp " "q25, q26, [x0, #0x20]"),
Q!(" trn2 " "v25.2d, v4.2d, v5.2d"),
Q!(" trn2 " "v26.2d, v6.2d, v7.2d"),
Q!(" stp " "q25, q26, [x2, #0x20]"),
Q!(" trn1 " "v25.2d, v8.2d, v9.2d"),
Q!(" trn1 " "v26.2d, v10.2d, v11.2d"),
Q!(" stp " "q25, q26, [x0, #0x40]"),
Q!(" trn2 " "v25.2d, v8.2d, v9.2d"),
Q!(" trn2 " "v26.2d, v10.2d, v11.2d"),
Q!(" stp " "q25, q26, [x2, #0x40]"),
Q!(" trn1 " "v25.2d, v12.2d, v13.2d"),
Q!(" trn1 " "v26.2d, v14.2d, v15.2d"),
Q!(" stp " "q25, q26, [x0, #0x60]"),
Q!(" trn2 " "v25.2d, v12.2d, v13.2d"),
Q!(" trn2 " "v26.2d, v14.2d, v15.2d"),
Q!(" stp " "q25, q26, [x2, #0x60]"),
Q!(" trn1 " "v25.2d, v16.2d, v17.2d"),
Q!(" trn1 " "v26.2d, v18.2d, v19.2d"),
Q!(" stp " "q25, q26, [x0, #0x80]"),
Q!(" trn2 " "v25.2d, v16.2d, v17.2d"),
Q!(" trn2 " "v26.2d, v18.2d, v19.2d"),
Q!(" stp " "q25, q26, [x2, #0x80]"),
Q!(" trn1 " "v25.2d, v20.2d, v21.2d"),
Q!(" trn1 " "v26.2d, v22.2d, v23.2d"),
Q!(" stp " "q25, q26, [x0, #0xa0]"),
Q!(" trn2 " "v25.2d, v20.2d, v21.2d"),
Q!(" trn2 " "v26.2d, v22.2d, v23.2d"),
Q!(" stp " "q25, q26, [x2, #0xa0]"),
Q!(" str " "d24, [x0, #0xc0]"),
Q!(" trn2 " "v24.2d, v24.2d, v24.2d"),
Q!(" str " "d24, [x2, #0xc0]"),
// Restore registers and return
Q!(" ldp " "d8, d9, [sp, # (0 + 0)]"),
Q!(" ldp " "d10, d11, [sp, # (16 + 0)]"),
Q!(" ldp " "d12, d13, [sp, # (32 + 0)]"),
Q!(" ldp " "d14, d15, [sp, # (48 + 0)]"),
Q!(" add " "sp, sp, # (64 + 0)"),
inout("x0") a.as_mut_ptr() => _,
inout("x1") rc.as_ptr() => _,
// clobbers
out("v0") _,
out("v1") _,
out("v10") _,
out("v11") _,
out("v12") _,
out("v13") _,
out("v14") _,
out("v15") _,
out("v16") _,
out("v17") _,
out("v18") _,
out("v19") _,
out("v2") _,
out("v20") _,
out("v21") _,
out("v22") _,
out("v23") _,
out("v24") _,
out("v25") _,
out("v26") _,
out("v27") _,
out("v28") _,
out("v29") _,
out("v3") _,
out("v30") _,
out("v31") _,
out("v4") _,
out("v5") _,
out("v6") _,
out("v7") _,
out("v8") _,
out("v9") _,
out("x2") _,
)
};
}