Bitcoin Core  0.20.99
P2P Digital Currency
sha256_shani.cpp
Go to the documentation of this file.
1 // Copyright (c) 2018-2019 The Bitcoin Core developers
2 // Distributed under the MIT software license, see the accompanying
3 // file COPYING or http://www.opensource.org/licenses/mit-license.php.
4 //
5 // Based on https://github.com/noloader/SHA-Intrinsics/blob/master/sha256-x86.c,
6 // Written and placed in public domain by Jeffrey Walton.
7 // Based on code from Intel, and by Sean Gulley for the miTLS project.
8 
9 #ifdef ENABLE_SHANI
10 
11 #include <stdint.h>
12 #include <immintrin.h>
13 
14 namespace {
15 
16 alignas(__m128i) const uint8_t MASK[16] = {0x03, 0x02, 0x01, 0x00, 0x07, 0x06, 0x05, 0x04, 0x0b, 0x0a, 0x09, 0x08, 0x0f, 0x0e, 0x0d, 0x0c};
17 alignas(__m128i) const uint8_t INIT0[16] = {0x8c, 0x68, 0x05, 0x9b, 0x7f, 0x52, 0x0e, 0x51, 0x85, 0xae, 0x67, 0xbb, 0x67, 0xe6, 0x09, 0x6a};
18 alignas(__m128i) const uint8_t INIT1[16] = {0x19, 0xcd, 0xe0, 0x5b, 0xab, 0xd9, 0x83, 0x1f, 0x3a, 0xf5, 0x4f, 0xa5, 0x72, 0xf3, 0x6e, 0x3c};
19 
20 void inline __attribute__((always_inline)) QuadRound(__m128i& state0, __m128i& state1, uint64_t k1, uint64_t k0)
21 {
22  const __m128i msg = _mm_set_epi64x(k1, k0);
23  state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
24  state0 = _mm_sha256rnds2_epu32(state0, state1, _mm_shuffle_epi32(msg, 0x0e));
25 }
26 
27 void inline __attribute__((always_inline)) QuadRound(__m128i& state0, __m128i& state1, __m128i m, uint64_t k1, uint64_t k0)
28 {
29  const __m128i msg = _mm_add_epi32(m, _mm_set_epi64x(k1, k0));
30  state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
31  state0 = _mm_sha256rnds2_epu32(state0, state1, _mm_shuffle_epi32(msg, 0x0e));
32 }
33 
34 void inline __attribute__((always_inline)) ShiftMessageA(__m128i& m0, __m128i m1)
35 {
36  m0 = _mm_sha256msg1_epu32(m0, m1);
37 }
38 
39 void inline __attribute__((always_inline)) ShiftMessageC(__m128i& m0, __m128i m1, __m128i& m2)
40 {
41  m2 = _mm_sha256msg2_epu32(_mm_add_epi32(m2, _mm_alignr_epi8(m1, m0, 4)), m1);
42 }
43 
44 void inline __attribute__((always_inline)) ShiftMessageB(__m128i& m0, __m128i m1, __m128i& m2)
45 {
46  ShiftMessageC(m0, m1, m2);
47  ShiftMessageA(m0, m1);
48 }
49 
50 void inline __attribute__((always_inline)) Shuffle(__m128i& s0, __m128i& s1)
51 {
52  const __m128i t1 = _mm_shuffle_epi32(s0, 0xB1);
53  const __m128i t2 = _mm_shuffle_epi32(s1, 0x1B);
54  s0 = _mm_alignr_epi8(t1, t2, 0x08);
55  s1 = _mm_blend_epi16(t2, t1, 0xF0);
56 }
57 
58 void inline __attribute__((always_inline)) Unshuffle(__m128i& s0, __m128i& s1)
59 {
60  const __m128i t1 = _mm_shuffle_epi32(s0, 0x1B);
61  const __m128i t2 = _mm_shuffle_epi32(s1, 0xB1);
62  s0 = _mm_blend_epi16(t1, t2, 0xF0);
63  s1 = _mm_alignr_epi8(t2, t1, 0x08);
64 }
65 
66 __m128i inline __attribute__((always_inline)) Load(const unsigned char* in)
67 {
68  return _mm_shuffle_epi8(_mm_loadu_si128((const __m128i*)in), _mm_load_si128((const __m128i*)MASK));
69 }
70 
71 void inline __attribute__((always_inline)) Save(unsigned char* out, __m128i s)
72 {
73  _mm_storeu_si128((__m128i*)out, _mm_shuffle_epi8(s, _mm_load_si128((const __m128i*)MASK)));
74 }
75 }
76 
77 namespace sha256_shani {
78 void Transform(uint32_t* s, const unsigned char* chunk, size_t blocks)
79 {
80  __m128i m0, m1, m2, m3, s0, s1, so0, so1;
81 
82  /* Load state */
83  s0 = _mm_loadu_si128((const __m128i*)s);
84  s1 = _mm_loadu_si128((const __m128i*)(s + 4));
85  Shuffle(s0, s1);
86 
87  while (blocks--) {
88  /* Remember old state */
89  so0 = s0;
90  so1 = s1;
91 
92  /* Load data and transform */
93  m0 = Load(chunk);
94  QuadRound(s0, s1, m0, 0xe9b5dba5b5c0fbcfull, 0x71374491428a2f98ull);
95  m1 = Load(chunk + 16);
96  QuadRound(s0, s1, m1, 0xab1c5ed5923f82a4ull, 0x59f111f13956c25bull);
97  ShiftMessageA(m0, m1);
98  m2 = Load(chunk + 32);
99  QuadRound(s0, s1, m2, 0x550c7dc3243185beull, 0x12835b01d807aa98ull);
100  ShiftMessageA(m1, m2);
101  m3 = Load(chunk + 48);
102  QuadRound(s0, s1, m3, 0xc19bf1749bdc06a7ull, 0x80deb1fe72be5d74ull);
103  ShiftMessageB(m2, m3, m0);
104  QuadRound(s0, s1, m0, 0x240ca1cc0fc19dc6ull, 0xefbe4786E49b69c1ull);
105  ShiftMessageB(m3, m0, m1);
106  QuadRound(s0, s1, m1, 0x76f988da5cb0a9dcull, 0x4a7484aa2de92c6full);
107  ShiftMessageB(m0, m1, m2);
108  QuadRound(s0, s1, m2, 0xbf597fc7b00327c8ull, 0xa831c66d983e5152ull);
109  ShiftMessageB(m1, m2, m3);
110  QuadRound(s0, s1, m3, 0x1429296706ca6351ull, 0xd5a79147c6e00bf3ull);
111  ShiftMessageB(m2, m3, m0);
112  QuadRound(s0, s1, m0, 0x53380d134d2c6dfcull, 0x2e1b213827b70a85ull);
113  ShiftMessageB(m3, m0, m1);
114  QuadRound(s0, s1, m1, 0x92722c8581c2c92eull, 0x766a0abb650a7354ull);
115  ShiftMessageB(m0, m1, m2);
116  QuadRound(s0, s1, m2, 0xc76c51A3c24b8b70ull, 0xa81a664ba2bfe8a1ull);
117  ShiftMessageB(m1, m2, m3);
118  QuadRound(s0, s1, m3, 0x106aa070f40e3585ull, 0xd6990624d192e819ull);
119  ShiftMessageB(m2, m3, m0);
120  QuadRound(s0, s1, m0, 0x34b0bcb52748774cull, 0x1e376c0819a4c116ull);
121  ShiftMessageB(m3, m0, m1);
122  QuadRound(s0, s1, m1, 0x682e6ff35b9cca4full, 0x4ed8aa4a391c0cb3ull);
123  ShiftMessageC(m0, m1, m2);
124  QuadRound(s0, s1, m2, 0x8cc7020884c87814ull, 0x78a5636f748f82eeull);
125  ShiftMessageC(m1, m2, m3);
126  QuadRound(s0, s1, m3, 0xc67178f2bef9A3f7ull, 0xa4506ceb90befffaull);
127 
128  /* Combine with old state */
129  s0 = _mm_add_epi32(s0, so0);
130  s1 = _mm_add_epi32(s1, so1);
131 
132  /* Advance */
133  chunk += 64;
134  }
135 
136  Unshuffle(s0, s1);
137  _mm_storeu_si128((__m128i*)s, s0);
138  _mm_storeu_si128((__m128i*)(s + 4), s1);
139 }
140 }
141 
142 namespace sha256d64_shani {
143 
144 void Transform_2way(unsigned char* out, const unsigned char* in)
145 {
146  __m128i am0, am1, am2, am3, as0, as1, aso0, aso1;
147  __m128i bm0, bm1, bm2, bm3, bs0, bs1, bso0, bso1;
148 
149  /* Transform 1 */
150  bs0 = as0 = _mm_load_si128((const __m128i*)INIT0);
151  bs1 = as1 = _mm_load_si128((const __m128i*)INIT1);
152  am0 = Load(in);
153  bm0 = Load(in + 64);
154  QuadRound(as0, as1, am0, 0xe9b5dba5b5c0fbcfull, 0x71374491428a2f98ull);
155  QuadRound(bs0, bs1, bm0, 0xe9b5dba5b5c0fbcfull, 0x71374491428a2f98ull);
156  am1 = Load(in + 16);
157  bm1 = Load(in + 80);
158  QuadRound(as0, as1, am1, 0xab1c5ed5923f82a4ull, 0x59f111f13956c25bull);
159  QuadRound(bs0, bs1, bm1, 0xab1c5ed5923f82a4ull, 0x59f111f13956c25bull);
160  ShiftMessageA(am0, am1);
161  ShiftMessageA(bm0, bm1);
162  am2 = Load(in + 32);
163  bm2 = Load(in + 96);
164  QuadRound(as0, as1, am2, 0x550c7dc3243185beull, 0x12835b01d807aa98ull);
165  QuadRound(bs0, bs1, bm2, 0x550c7dc3243185beull, 0x12835b01d807aa98ull);
166  ShiftMessageA(am1, am2);
167  ShiftMessageA(bm1, bm2);
168  am3 = Load(in + 48);
169  bm3 = Load(in + 112);
170  QuadRound(as0, as1, am3, 0xc19bf1749bdc06a7ull, 0x80deb1fe72be5d74ull);
171  QuadRound(bs0, bs1, bm3, 0xc19bf1749bdc06a7ull, 0x80deb1fe72be5d74ull);
172  ShiftMessageB(am2, am3, am0);
173  ShiftMessageB(bm2, bm3, bm0);
174  QuadRound(as0, as1, am0, 0x240ca1cc0fc19dc6ull, 0xefbe4786E49b69c1ull);
175  QuadRound(bs0, bs1, bm0, 0x240ca1cc0fc19dc6ull, 0xefbe4786E49b69c1ull);
176  ShiftMessageB(am3, am0, am1);
177  ShiftMessageB(bm3, bm0, bm1);
178  QuadRound(as0, as1, am1, 0x76f988da5cb0a9dcull, 0x4a7484aa2de92c6full);
179  QuadRound(bs0, bs1, bm1, 0x76f988da5cb0a9dcull, 0x4a7484aa2de92c6full);
180  ShiftMessageB(am0, am1, am2);
181  ShiftMessageB(bm0, bm1, bm2);
182  QuadRound(as0, as1, am2, 0xbf597fc7b00327c8ull, 0xa831c66d983e5152ull);
183  QuadRound(bs0, bs1, bm2, 0xbf597fc7b00327c8ull, 0xa831c66d983e5152ull);
184  ShiftMessageB(am1, am2, am3);
185  ShiftMessageB(bm1, bm2, bm3);
186  QuadRound(as0, as1, am3, 0x1429296706ca6351ull, 0xd5a79147c6e00bf3ull);
187  QuadRound(bs0, bs1, bm3, 0x1429296706ca6351ull, 0xd5a79147c6e00bf3ull);
188  ShiftMessageB(am2, am3, am0);
189  ShiftMessageB(bm2, bm3, bm0);
190  QuadRound(as0, as1, am0, 0x53380d134d2c6dfcull, 0x2e1b213827b70a85ull);
191  QuadRound(bs0, bs1, bm0, 0x53380d134d2c6dfcull, 0x2e1b213827b70a85ull);
192  ShiftMessageB(am3, am0, am1);
193  ShiftMessageB(bm3, bm0, bm1);
194  QuadRound(as0, as1, am1, 0x92722c8581c2c92eull, 0x766a0abb650a7354ull);
195  QuadRound(bs0, bs1, bm1, 0x92722c8581c2c92eull, 0x766a0abb650a7354ull);
196  ShiftMessageB(am0, am1, am2);
197  ShiftMessageB(bm0, bm1, bm2);
198  QuadRound(as0, as1, am2, 0xc76c51A3c24b8b70ull, 0xa81a664ba2bfe8a1ull);
199  QuadRound(bs0, bs1, bm2, 0xc76c51A3c24b8b70ull, 0xa81a664ba2bfe8a1ull);
200  ShiftMessageB(am1, am2, am3);
201  ShiftMessageB(bm1, bm2, bm3);
202  QuadRound(as0, as1, am3, 0x106aa070f40e3585ull, 0xd6990624d192e819ull);
203  QuadRound(bs0, bs1, bm3, 0x106aa070f40e3585ull, 0xd6990624d192e819ull);
204  ShiftMessageB(am2, am3, am0);
205  ShiftMessageB(bm2, bm3, bm0);
206  QuadRound(as0, as1, am0, 0x34b0bcb52748774cull, 0x1e376c0819a4c116ull);
207  QuadRound(bs0, bs1, bm0, 0x34b0bcb52748774cull, 0x1e376c0819a4c116ull);
208  ShiftMessageB(am3, am0, am1);
209  ShiftMessageB(bm3, bm0, bm1);
210  QuadRound(as0, as1, am1, 0x682e6ff35b9cca4full, 0x4ed8aa4a391c0cb3ull);
211  QuadRound(bs0, bs1, bm1, 0x682e6ff35b9cca4full, 0x4ed8aa4a391c0cb3ull);
212  ShiftMessageC(am0, am1, am2);
213  ShiftMessageC(bm0, bm1, bm2);
214  QuadRound(as0, as1, am2, 0x8cc7020884c87814ull, 0x78a5636f748f82eeull);
215  QuadRound(bs0, bs1, bm2, 0x8cc7020884c87814ull, 0x78a5636f748f82eeull);
216  ShiftMessageC(am1, am2, am3);
217  ShiftMessageC(bm1, bm2, bm3);
218  QuadRound(as0, as1, am3, 0xc67178f2bef9A3f7ull, 0xa4506ceb90befffaull);
219  QuadRound(bs0, bs1, bm3, 0xc67178f2bef9A3f7ull, 0xa4506ceb90befffaull);
220  as0 = _mm_add_epi32(as0, _mm_load_si128((const __m128i*)INIT0));
221  bs0 = _mm_add_epi32(bs0, _mm_load_si128((const __m128i*)INIT0));
222  as1 = _mm_add_epi32(as1, _mm_load_si128((const __m128i*)INIT1));
223  bs1 = _mm_add_epi32(bs1, _mm_load_si128((const __m128i*)INIT1));
224 
225  /* Transform 2 */
226  aso0 = as0;
227  bso0 = bs0;
228  aso1 = as1;
229  bso1 = bs1;
230  QuadRound(as0, as1, 0xe9b5dba5b5c0fbcfull, 0x71374491c28a2f98ull);
231  QuadRound(bs0, bs1, 0xe9b5dba5b5c0fbcfull, 0x71374491c28a2f98ull);
232  QuadRound(as0, as1, 0xab1c5ed5923f82a4ull, 0x59f111f13956c25bull);
233  QuadRound(bs0, bs1, 0xab1c5ed5923f82a4ull, 0x59f111f13956c25bull);
234  QuadRound(as0, as1, 0x550c7dc3243185beull, 0x12835b01d807aa98ull);
235  QuadRound(bs0, bs1, 0x550c7dc3243185beull, 0x12835b01d807aa98ull);
236  QuadRound(as0, as1, 0xc19bf3749bdc06a7ull, 0x80deb1fe72be5d74ull);
237  QuadRound(bs0, bs1, 0xc19bf3749bdc06a7ull, 0x80deb1fe72be5d74ull);
238  QuadRound(as0, as1, 0x240cf2540fe1edc6ull, 0xf0fe4786649b69c1ull);
239  QuadRound(bs0, bs1, 0x240cf2540fe1edc6ull, 0xf0fe4786649b69c1ull);
240  QuadRound(as0, as1, 0x16f988fa61b9411eull, 0x6cc984be4fe9346full);
241  QuadRound(bs0, bs1, 0x16f988fa61b9411eull, 0x6cc984be4fe9346full);
242  QuadRound(as0, as1, 0xb9d99ec7b019fc65ull, 0xa88e5a6df2c65152ull);
243  QuadRound(bs0, bs1, 0xb9d99ec7b019fc65ull, 0xa88e5a6df2c65152ull);
244  QuadRound(as0, as1, 0xc7353eb0fdb1232bull, 0xe70eeaa09a1231c3ull);
245  QuadRound(bs0, bs1, 0xc7353eb0fdb1232bull, 0xe70eeaa09a1231c3ull);
246  QuadRound(as0, as1, 0xdc1eeefd5a0f118full, 0xcb976d5f3069bad5ull);
247  QuadRound(bs0, bs1, 0xdc1eeefd5a0f118full, 0xcb976d5f3069bad5ull);
248  QuadRound(as0, as1, 0xe15d5b1658f4ca9dull, 0xde0b7a040a35b689ull);
249  QuadRound(bs0, bs1, 0xe15d5b1658f4ca9dull, 0xde0b7a040a35b689ull);
250  QuadRound(as0, as1, 0x6fab9537a507ea32ull, 0x37088980007f3e86ull);
251  QuadRound(bs0, bs1, 0x6fab9537a507ea32ull, 0x37088980007f3e86ull);
252  QuadRound(as0, as1, 0xc0bbbe37cdaa3b6dull, 0x0d8cd6f117406110ull);
253  QuadRound(bs0, bs1, 0xc0bbbe37cdaa3b6dull, 0x0d8cd6f117406110ull);
254  QuadRound(as0, as1, 0x6fd15ca70b02e931ull, 0xdb48a36383613bdaull);
255  QuadRound(bs0, bs1, 0x6fd15ca70b02e931ull, 0xdb48a36383613bdaull);
256  QuadRound(as0, as1, 0x6d4378906ed41a95ull, 0x31338431521afacaull);
257  QuadRound(bs0, bs1, 0x6d4378906ed41a95ull, 0x31338431521afacaull);
258  QuadRound(as0, as1, 0x532fb63cb5c9a0e6ull, 0x9eccabbdc39c91f2ull);
259  QuadRound(bs0, bs1, 0x532fb63cb5c9a0e6ull, 0x9eccabbdc39c91f2ull);
260  QuadRound(as0, as1, 0x4c191d76a4954b68ull, 0x07237ea3d2c741c6ull);
261  QuadRound(bs0, bs1, 0x4c191d76a4954b68ull, 0x07237ea3d2c741c6ull);
262  as0 = _mm_add_epi32(as0, aso0);
263  bs0 = _mm_add_epi32(bs0, bso0);
264  as1 = _mm_add_epi32(as1, aso1);
265  bs1 = _mm_add_epi32(bs1, bso1);
266 
267  /* Extract hash */
268  Unshuffle(as0, as1);
269  Unshuffle(bs0, bs1);
270  am0 = as0;
271  bm0 = bs0;
272  am1 = as1;
273  bm1 = bs1;
274 
275  /* Transform 3 */
276  bs0 = as0 = _mm_load_si128((const __m128i*)INIT0);
277  bs1 = as1 = _mm_load_si128((const __m128i*)INIT1);
278  QuadRound(as0, as1, am0, 0xe9b5dba5B5c0fbcfull, 0x71374491428a2f98ull);
279  QuadRound(bs0, bs1, bm0, 0xe9b5dba5B5c0fbcfull, 0x71374491428a2f98ull);
280  QuadRound(as0, as1, am1, 0xab1c5ed5923f82a4ull, 0x59f111f13956c25bull);
281  QuadRound(bs0, bs1, bm1, 0xab1c5ed5923f82a4ull, 0x59f111f13956c25bull);
282  ShiftMessageA(am0, am1);
283  ShiftMessageA(bm0, bm1);
284  bm2 = am2 = _mm_set_epi64x(0x0ull, 0x80000000ull);
285  QuadRound(as0, as1, 0x550c7dc3243185beull, 0x12835b015807aa98ull);
286  QuadRound(bs0, bs1, 0x550c7dc3243185beull, 0x12835b015807aa98ull);
287  ShiftMessageA(am1, am2);
288  ShiftMessageA(bm1, bm2);
289  bm3 = am3 = _mm_set_epi64x(0x10000000000ull, 0x0ull);
290  QuadRound(as0, as1, 0xc19bf2749bdc06a7ull, 0x80deb1fe72be5d74ull);
291  QuadRound(bs0, bs1, 0xc19bf2749bdc06a7ull, 0x80deb1fe72be5d74ull);
292  ShiftMessageB(am2, am3, am0);
293  ShiftMessageB(bm2, bm3, bm0);
294  QuadRound(as0, as1, am0, 0x240ca1cc0fc19dc6ull, 0xefbe4786e49b69c1ull);
295  QuadRound(bs0, bs1, bm0, 0x240ca1cc0fc19dc6ull, 0xefbe4786e49b69c1ull);
296  ShiftMessageB(am3, am0, am1);
297  ShiftMessageB(bm3, bm0, bm1);
298  QuadRound(as0, as1, am1, 0x76f988da5cb0a9dcull, 0x4a7484aa2de92c6full);
299  QuadRound(bs0, bs1, bm1, 0x76f988da5cb0a9dcull, 0x4a7484aa2de92c6full);
300  ShiftMessageB(am0, am1, am2);
301  ShiftMessageB(bm0, bm1, bm2);
302  QuadRound(as0, as1, am2, 0xbf597fc7b00327c8ull, 0xa831c66d983e5152ull);
303  QuadRound(bs0, bs1, bm2, 0xbf597fc7b00327c8ull, 0xa831c66d983e5152ull);
304  ShiftMessageB(am1, am2, am3);
305  ShiftMessageB(bm1, bm2, bm3);
306  QuadRound(as0, as1, am3, 0x1429296706ca6351ull, 0xd5a79147c6e00bf3ull);
307  QuadRound(bs0, bs1, bm3, 0x1429296706ca6351ull, 0xd5a79147c6e00bf3ull);
308  ShiftMessageB(am2, am3, am0);
309  ShiftMessageB(bm2, bm3, bm0);
310  QuadRound(as0, as1, am0, 0x53380d134d2c6dfcull, 0x2e1b213827b70a85ull);
311  QuadRound(bs0, bs1, bm0, 0x53380d134d2c6dfcull, 0x2e1b213827b70a85ull);
312  ShiftMessageB(am3, am0, am1);
313  ShiftMessageB(bm3, bm0, bm1);
314  QuadRound(as0, as1, am1, 0x92722c8581c2c92eull, 0x766a0abb650a7354ull);
315  QuadRound(bs0, bs1, bm1, 0x92722c8581c2c92eull, 0x766a0abb650a7354ull);
316  ShiftMessageB(am0, am1, am2);
317  ShiftMessageB(bm0, bm1, bm2);
318  QuadRound(as0, as1, am2, 0xc76c51a3c24b8b70ull, 0xa81a664ba2bfe8A1ull);
319  QuadRound(bs0, bs1, bm2, 0xc76c51a3c24b8b70ull, 0xa81a664ba2bfe8A1ull);
320  ShiftMessageB(am1, am2, am3);
321  ShiftMessageB(bm1, bm2, bm3);
322  QuadRound(as0, as1, am3, 0x106aa070f40e3585ull, 0xd6990624d192e819ull);
323  QuadRound(bs0, bs1, bm3, 0x106aa070f40e3585ull, 0xd6990624d192e819ull);
324  ShiftMessageB(am2, am3, am0);
325  ShiftMessageB(bm2, bm3, bm0);
326  QuadRound(as0, as1, am0, 0x34b0bcb52748774cull, 0x1e376c0819a4c116ull);
327  QuadRound(bs0, bs1, bm0, 0x34b0bcb52748774cull, 0x1e376c0819a4c116ull);
328  ShiftMessageB(am3, am0, am1);
329  ShiftMessageB(bm3, bm0, bm1);
330  QuadRound(as0, as1, am1, 0x682e6ff35b9cca4full, 0x4ed8aa4a391c0cb3ull);
331  QuadRound(bs0, bs1, bm1, 0x682e6ff35b9cca4full, 0x4ed8aa4a391c0cb3ull);
332  ShiftMessageC(am0, am1, am2);
333  ShiftMessageC(bm0, bm1, bm2);
334  QuadRound(as0, as1, am2, 0x8cc7020884c87814ull, 0x78a5636f748f82eeull);
335  QuadRound(bs0, bs1, bm2, 0x8cc7020884c87814ull, 0x78a5636f748f82eeull);
336  ShiftMessageC(am1, am2, am3);
337  ShiftMessageC(bm1, bm2, bm3);
338  QuadRound(as0, as1, am3, 0xc67178f2bef9a3f7ull, 0xa4506ceb90befffaull);
339  QuadRound(bs0, bs1, bm3, 0xc67178f2bef9a3f7ull, 0xa4506ceb90befffaull);
340  as0 = _mm_add_epi32(as0, _mm_load_si128((const __m128i*)INIT0));
341  bs0 = _mm_add_epi32(bs0, _mm_load_si128((const __m128i*)INIT0));
342  as1 = _mm_add_epi32(as1, _mm_load_si128((const __m128i*)INIT1));
343  bs1 = _mm_add_epi32(bs1, _mm_load_si128((const __m128i*)INIT1));
344 
345  /* Extract hash into out */
346  Unshuffle(as0, as1);
347  Unshuffle(bs0, bs1);
348  Save(out, as0);
349  Save(out + 16, as1);
350  Save(out + 32, bs0);
351  Save(out + 48, bs1);
352 }
353 
354 }
355 
356 #endif
void Transform_2way(unsigned char *out, const unsigned char *in)
void Transform(uint32_t *s, const unsigned char *chunk, size_t blocks)
static const unsigned char k1[32]
void Shuffle(I first, I last, R &&rng)
More efficient than using std::shuffle on a FastRandomContext.
Definition: random.h:231