Bitcoin Core  27.99.0
P2P Digital Currency
sha256_x86_shani.cpp
Go to the documentation of this file.
1 // Copyright (c) 2018-2022 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_X86_SHANI
10 
11 #include <stdint.h>
12 #include <immintrin.h>
13 
14 #include <attributes.h>
15 
16 namespace {
17 
18 alignas(__m128i) const uint8_t MASK[16] = {0x03, 0x02, 0x01, 0x00, 0x07, 0x06, 0x05, 0x04, 0x0b, 0x0a, 0x09, 0x08, 0x0f, 0x0e, 0x0d, 0x0c};
19 alignas(__m128i) const uint8_t INIT0[16] = {0x8c, 0x68, 0x05, 0x9b, 0x7f, 0x52, 0x0e, 0x51, 0x85, 0xae, 0x67, 0xbb, 0x67, 0xe6, 0x09, 0x6a};
20 alignas(__m128i) const uint8_t INIT1[16] = {0x19, 0xcd, 0xe0, 0x5b, 0xab, 0xd9, 0x83, 0x1f, 0x3a, 0xf5, 0x4f, 0xa5, 0x72, 0xf3, 0x6e, 0x3c};
21 
22 void ALWAYS_INLINE QuadRound(__m128i& state0, __m128i& state1, uint64_t k1, uint64_t k0)
23 {
24  const __m128i msg = _mm_set_epi64x(k1, k0);
25  state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
26  state0 = _mm_sha256rnds2_epu32(state0, state1, _mm_shuffle_epi32(msg, 0x0e));
27 }
28 
29 void ALWAYS_INLINE QuadRound(__m128i& state0, __m128i& state1, __m128i m, uint64_t k1, uint64_t k0)
30 {
31  const __m128i msg = _mm_add_epi32(m, _mm_set_epi64x(k1, k0));
32  state1 = _mm_sha256rnds2_epu32(state1, state0, msg);
33  state0 = _mm_sha256rnds2_epu32(state0, state1, _mm_shuffle_epi32(msg, 0x0e));
34 }
35 
36 void ALWAYS_INLINE ShiftMessageA(__m128i& m0, __m128i m1)
37 {
38  m0 = _mm_sha256msg1_epu32(m0, m1);
39 }
40 
41 void ALWAYS_INLINE ShiftMessageC(__m128i& m0, __m128i m1, __m128i& m2)
42 {
43  m2 = _mm_sha256msg2_epu32(_mm_add_epi32(m2, _mm_alignr_epi8(m1, m0, 4)), m1);
44 }
45 
46 void ALWAYS_INLINE ShiftMessageB(__m128i& m0, __m128i m1, __m128i& m2)
47 {
48  ShiftMessageC(m0, m1, m2);
49  ShiftMessageA(m0, m1);
50 }
51 
52 void ALWAYS_INLINE Shuffle(__m128i& s0, __m128i& s1)
53 {
54  const __m128i t1 = _mm_shuffle_epi32(s0, 0xB1);
55  const __m128i t2 = _mm_shuffle_epi32(s1, 0x1B);
56  s0 = _mm_alignr_epi8(t1, t2, 0x08);
57  s1 = _mm_blend_epi16(t2, t1, 0xF0);
58 }
59 
60 void ALWAYS_INLINE Unshuffle(__m128i& s0, __m128i& s1)
61 {
62  const __m128i t1 = _mm_shuffle_epi32(s0, 0x1B);
63  const __m128i t2 = _mm_shuffle_epi32(s1, 0xB1);
64  s0 = _mm_blend_epi16(t1, t2, 0xF0);
65  s1 = _mm_alignr_epi8(t2, t1, 0x08);
66 }
67 
68 __m128i ALWAYS_INLINE Load(const unsigned char* in)
69 {
70  return _mm_shuffle_epi8(_mm_loadu_si128((const __m128i*)in), _mm_load_si128((const __m128i*)MASK));
71 }
72 
73 void ALWAYS_INLINE Save(unsigned char* out, __m128i s)
74 {
75  _mm_storeu_si128((__m128i*)out, _mm_shuffle_epi8(s, _mm_load_si128((const __m128i*)MASK)));
76 }
77 }
78 
79 namespace sha256_x86_shani {
80 void Transform(uint32_t* s, const unsigned char* chunk, size_t blocks)
81 {
82  __m128i m0, m1, m2, m3, s0, s1, so0, so1;
83 
84  /* Load state */
85  s0 = _mm_loadu_si128((const __m128i*)s);
86  s1 = _mm_loadu_si128((const __m128i*)(s + 4));
87  Shuffle(s0, s1);
88 
89  while (blocks--) {
90  /* Remember old state */
91  so0 = s0;
92  so1 = s1;
93 
94  /* Load data and transform */
95  m0 = Load(chunk);
96  QuadRound(s0, s1, m0, 0xe9b5dba5b5c0fbcfull, 0x71374491428a2f98ull);
97  m1 = Load(chunk + 16);
98  QuadRound(s0, s1, m1, 0xab1c5ed5923f82a4ull, 0x59f111f13956c25bull);
99  ShiftMessageA(m0, m1);
100  m2 = Load(chunk + 32);
101  QuadRound(s0, s1, m2, 0x550c7dc3243185beull, 0x12835b01d807aa98ull);
102  ShiftMessageA(m1, m2);
103  m3 = Load(chunk + 48);
104  QuadRound(s0, s1, m3, 0xc19bf1749bdc06a7ull, 0x80deb1fe72be5d74ull);
105  ShiftMessageB(m2, m3, m0);
106  QuadRound(s0, s1, m0, 0x240ca1cc0fc19dc6ull, 0xefbe4786E49b69c1ull);
107  ShiftMessageB(m3, m0, m1);
108  QuadRound(s0, s1, m1, 0x76f988da5cb0a9dcull, 0x4a7484aa2de92c6full);
109  ShiftMessageB(m0, m1, m2);
110  QuadRound(s0, s1, m2, 0xbf597fc7b00327c8ull, 0xa831c66d983e5152ull);
111  ShiftMessageB(m1, m2, m3);
112  QuadRound(s0, s1, m3, 0x1429296706ca6351ull, 0xd5a79147c6e00bf3ull);
113  ShiftMessageB(m2, m3, m0);
114  QuadRound(s0, s1, m0, 0x53380d134d2c6dfcull, 0x2e1b213827b70a85ull);
115  ShiftMessageB(m3, m0, m1);
116  QuadRound(s0, s1, m1, 0x92722c8581c2c92eull, 0x766a0abb650a7354ull);
117  ShiftMessageB(m0, m1, m2);
118  QuadRound(s0, s1, m2, 0xc76c51A3c24b8b70ull, 0xa81a664ba2bfe8a1ull);
119  ShiftMessageB(m1, m2, m3);
120  QuadRound(s0, s1, m3, 0x106aa070f40e3585ull, 0xd6990624d192e819ull);
121  ShiftMessageB(m2, m3, m0);
122  QuadRound(s0, s1, m0, 0x34b0bcb52748774cull, 0x1e376c0819a4c116ull);
123  ShiftMessageB(m3, m0, m1);
124  QuadRound(s0, s1, m1, 0x682e6ff35b9cca4full, 0x4ed8aa4a391c0cb3ull);
125  ShiftMessageC(m0, m1, m2);
126  QuadRound(s0, s1, m2, 0x8cc7020884c87814ull, 0x78a5636f748f82eeull);
127  ShiftMessageC(m1, m2, m3);
128  QuadRound(s0, s1, m3, 0xc67178f2bef9A3f7ull, 0xa4506ceb90befffaull);
129 
130  /* Combine with old state */
131  s0 = _mm_add_epi32(s0, so0);
132  s1 = _mm_add_epi32(s1, so1);
133 
134  /* Advance */
135  chunk += 64;
136  }
137 
138  Unshuffle(s0, s1);
139  _mm_storeu_si128((__m128i*)s, s0);
140  _mm_storeu_si128((__m128i*)(s + 4), s1);
141 }
142 }
143 
144 namespace sha256d64_x86_shani {
145 
146 void Transform_2way(unsigned char* out, const unsigned char* in)
147 {
148  __m128i am0, am1, am2, am3, as0, as1, aso0, aso1;
149  __m128i bm0, bm1, bm2, bm3, bs0, bs1, bso0, bso1;
150 
151  /* Transform 1 */
152  bs0 = as0 = _mm_load_si128((const __m128i*)INIT0);
153  bs1 = as1 = _mm_load_si128((const __m128i*)INIT1);
154  am0 = Load(in);
155  bm0 = Load(in + 64);
156  QuadRound(as0, as1, am0, 0xe9b5dba5b5c0fbcfull, 0x71374491428a2f98ull);
157  QuadRound(bs0, bs1, bm0, 0xe9b5dba5b5c0fbcfull, 0x71374491428a2f98ull);
158  am1 = Load(in + 16);
159  bm1 = Load(in + 80);
160  QuadRound(as0, as1, am1, 0xab1c5ed5923f82a4ull, 0x59f111f13956c25bull);
161  QuadRound(bs0, bs1, bm1, 0xab1c5ed5923f82a4ull, 0x59f111f13956c25bull);
162  ShiftMessageA(am0, am1);
163  ShiftMessageA(bm0, bm1);
164  am2 = Load(in + 32);
165  bm2 = Load(in + 96);
166  QuadRound(as0, as1, am2, 0x550c7dc3243185beull, 0x12835b01d807aa98ull);
167  QuadRound(bs0, bs1, bm2, 0x550c7dc3243185beull, 0x12835b01d807aa98ull);
168  ShiftMessageA(am1, am2);
169  ShiftMessageA(bm1, bm2);
170  am3 = Load(in + 48);
171  bm3 = Load(in + 112);
172  QuadRound(as0, as1, am3, 0xc19bf1749bdc06a7ull, 0x80deb1fe72be5d74ull);
173  QuadRound(bs0, bs1, bm3, 0xc19bf1749bdc06a7ull, 0x80deb1fe72be5d74ull);
174  ShiftMessageB(am2, am3, am0);
175  ShiftMessageB(bm2, bm3, bm0);
176  QuadRound(as0, as1, am0, 0x240ca1cc0fc19dc6ull, 0xefbe4786E49b69c1ull);
177  QuadRound(bs0, bs1, bm0, 0x240ca1cc0fc19dc6ull, 0xefbe4786E49b69c1ull);
178  ShiftMessageB(am3, am0, am1);
179  ShiftMessageB(bm3, bm0, bm1);
180  QuadRound(as0, as1, am1, 0x76f988da5cb0a9dcull, 0x4a7484aa2de92c6full);
181  QuadRound(bs0, bs1, bm1, 0x76f988da5cb0a9dcull, 0x4a7484aa2de92c6full);
182  ShiftMessageB(am0, am1, am2);
183  ShiftMessageB(bm0, bm1, bm2);
184  QuadRound(as0, as1, am2, 0xbf597fc7b00327c8ull, 0xa831c66d983e5152ull);
185  QuadRound(bs0, bs1, bm2, 0xbf597fc7b00327c8ull, 0xa831c66d983e5152ull);
186  ShiftMessageB(am1, am2, am3);
187  ShiftMessageB(bm1, bm2, bm3);
188  QuadRound(as0, as1, am3, 0x1429296706ca6351ull, 0xd5a79147c6e00bf3ull);
189  QuadRound(bs0, bs1, bm3, 0x1429296706ca6351ull, 0xd5a79147c6e00bf3ull);
190  ShiftMessageB(am2, am3, am0);
191  ShiftMessageB(bm2, bm3, bm0);
192  QuadRound(as0, as1, am0, 0x53380d134d2c6dfcull, 0x2e1b213827b70a85ull);
193  QuadRound(bs0, bs1, bm0, 0x53380d134d2c6dfcull, 0x2e1b213827b70a85ull);
194  ShiftMessageB(am3, am0, am1);
195  ShiftMessageB(bm3, bm0, bm1);
196  QuadRound(as0, as1, am1, 0x92722c8581c2c92eull, 0x766a0abb650a7354ull);
197  QuadRound(bs0, bs1, bm1, 0x92722c8581c2c92eull, 0x766a0abb650a7354ull);
198  ShiftMessageB(am0, am1, am2);
199  ShiftMessageB(bm0, bm1, bm2);
200  QuadRound(as0, as1, am2, 0xc76c51A3c24b8b70ull, 0xa81a664ba2bfe8a1ull);
201  QuadRound(bs0, bs1, bm2, 0xc76c51A3c24b8b70ull, 0xa81a664ba2bfe8a1ull);
202  ShiftMessageB(am1, am2, am3);
203  ShiftMessageB(bm1, bm2, bm3);
204  QuadRound(as0, as1, am3, 0x106aa070f40e3585ull, 0xd6990624d192e819ull);
205  QuadRound(bs0, bs1, bm3, 0x106aa070f40e3585ull, 0xd6990624d192e819ull);
206  ShiftMessageB(am2, am3, am0);
207  ShiftMessageB(bm2, bm3, bm0);
208  QuadRound(as0, as1, am0, 0x34b0bcb52748774cull, 0x1e376c0819a4c116ull);
209  QuadRound(bs0, bs1, bm0, 0x34b0bcb52748774cull, 0x1e376c0819a4c116ull);
210  ShiftMessageB(am3, am0, am1);
211  ShiftMessageB(bm3, bm0, bm1);
212  QuadRound(as0, as1, am1, 0x682e6ff35b9cca4full, 0x4ed8aa4a391c0cb3ull);
213  QuadRound(bs0, bs1, bm1, 0x682e6ff35b9cca4full, 0x4ed8aa4a391c0cb3ull);
214  ShiftMessageC(am0, am1, am2);
215  ShiftMessageC(bm0, bm1, bm2);
216  QuadRound(as0, as1, am2, 0x8cc7020884c87814ull, 0x78a5636f748f82eeull);
217  QuadRound(bs0, bs1, bm2, 0x8cc7020884c87814ull, 0x78a5636f748f82eeull);
218  ShiftMessageC(am1, am2, am3);
219  ShiftMessageC(bm1, bm2, bm3);
220  QuadRound(as0, as1, am3, 0xc67178f2bef9A3f7ull, 0xa4506ceb90befffaull);
221  QuadRound(bs0, bs1, bm3, 0xc67178f2bef9A3f7ull, 0xa4506ceb90befffaull);
222  as0 = _mm_add_epi32(as0, _mm_load_si128((const __m128i*)INIT0));
223  bs0 = _mm_add_epi32(bs0, _mm_load_si128((const __m128i*)INIT0));
224  as1 = _mm_add_epi32(as1, _mm_load_si128((const __m128i*)INIT1));
225  bs1 = _mm_add_epi32(bs1, _mm_load_si128((const __m128i*)INIT1));
226 
227  /* Transform 2 */
228  aso0 = as0;
229  bso0 = bs0;
230  aso1 = as1;
231  bso1 = bs1;
232  QuadRound(as0, as1, 0xe9b5dba5b5c0fbcfull, 0x71374491c28a2f98ull);
233  QuadRound(bs0, bs1, 0xe9b5dba5b5c0fbcfull, 0x71374491c28a2f98ull);
234  QuadRound(as0, as1, 0xab1c5ed5923f82a4ull, 0x59f111f13956c25bull);
235  QuadRound(bs0, bs1, 0xab1c5ed5923f82a4ull, 0x59f111f13956c25bull);
236  QuadRound(as0, as1, 0x550c7dc3243185beull, 0x12835b01d807aa98ull);
237  QuadRound(bs0, bs1, 0x550c7dc3243185beull, 0x12835b01d807aa98ull);
238  QuadRound(as0, as1, 0xc19bf3749bdc06a7ull, 0x80deb1fe72be5d74ull);
239  QuadRound(bs0, bs1, 0xc19bf3749bdc06a7ull, 0x80deb1fe72be5d74ull);
240  QuadRound(as0, as1, 0x240cf2540fe1edc6ull, 0xf0fe4786649b69c1ull);
241  QuadRound(bs0, bs1, 0x240cf2540fe1edc6ull, 0xf0fe4786649b69c1ull);
242  QuadRound(as0, as1, 0x16f988fa61b9411eull, 0x6cc984be4fe9346full);
243  QuadRound(bs0, bs1, 0x16f988fa61b9411eull, 0x6cc984be4fe9346full);
244  QuadRound(as0, as1, 0xb9d99ec7b019fc65ull, 0xa88e5a6df2c65152ull);
245  QuadRound(bs0, bs1, 0xb9d99ec7b019fc65ull, 0xa88e5a6df2c65152ull);
246  QuadRound(as0, as1, 0xc7353eb0fdb1232bull, 0xe70eeaa09a1231c3ull);
247  QuadRound(bs0, bs1, 0xc7353eb0fdb1232bull, 0xe70eeaa09a1231c3ull);
248  QuadRound(as0, as1, 0xdc1eeefd5a0f118full, 0xcb976d5f3069bad5ull);
249  QuadRound(bs0, bs1, 0xdc1eeefd5a0f118full, 0xcb976d5f3069bad5ull);
250  QuadRound(as0, as1, 0xe15d5b1658f4ca9dull, 0xde0b7a040a35b689ull);
251  QuadRound(bs0, bs1, 0xe15d5b1658f4ca9dull, 0xde0b7a040a35b689ull);
252  QuadRound(as0, as1, 0x6fab9537a507ea32ull, 0x37088980007f3e86ull);
253  QuadRound(bs0, bs1, 0x6fab9537a507ea32ull, 0x37088980007f3e86ull);
254  QuadRound(as0, as1, 0xc0bbbe37cdaa3b6dull, 0x0d8cd6f117406110ull);
255  QuadRound(bs0, bs1, 0xc0bbbe37cdaa3b6dull, 0x0d8cd6f117406110ull);
256  QuadRound(as0, as1, 0x6fd15ca70b02e931ull, 0xdb48a36383613bdaull);
257  QuadRound(bs0, bs1, 0x6fd15ca70b02e931ull, 0xdb48a36383613bdaull);
258  QuadRound(as0, as1, 0x6d4378906ed41a95ull, 0x31338431521afacaull);
259  QuadRound(bs0, bs1, 0x6d4378906ed41a95ull, 0x31338431521afacaull);
260  QuadRound(as0, as1, 0x532fb63cb5c9a0e6ull, 0x9eccabbdc39c91f2ull);
261  QuadRound(bs0, bs1, 0x532fb63cb5c9a0e6ull, 0x9eccabbdc39c91f2ull);
262  QuadRound(as0, as1, 0x4c191d76a4954b68ull, 0x07237ea3d2c741c6ull);
263  QuadRound(bs0, bs1, 0x4c191d76a4954b68ull, 0x07237ea3d2c741c6ull);
264  as0 = _mm_add_epi32(as0, aso0);
265  bs0 = _mm_add_epi32(bs0, bso0);
266  as1 = _mm_add_epi32(as1, aso1);
267  bs1 = _mm_add_epi32(bs1, bso1);
268 
269  /* Extract hash */
270  Unshuffle(as0, as1);
271  Unshuffle(bs0, bs1);
272  am0 = as0;
273  bm0 = bs0;
274  am1 = as1;
275  bm1 = bs1;
276 
277  /* Transform 3 */
278  bs0 = as0 = _mm_load_si128((const __m128i*)INIT0);
279  bs1 = as1 = _mm_load_si128((const __m128i*)INIT1);
280  QuadRound(as0, as1, am0, 0xe9b5dba5B5c0fbcfull, 0x71374491428a2f98ull);
281  QuadRound(bs0, bs1, bm0, 0xe9b5dba5B5c0fbcfull, 0x71374491428a2f98ull);
282  QuadRound(as0, as1, am1, 0xab1c5ed5923f82a4ull, 0x59f111f13956c25bull);
283  QuadRound(bs0, bs1, bm1, 0xab1c5ed5923f82a4ull, 0x59f111f13956c25bull);
284  ShiftMessageA(am0, am1);
285  ShiftMessageA(bm0, bm1);
286  bm2 = am2 = _mm_set_epi64x(0x0ull, 0x80000000ull);
287  QuadRound(as0, as1, 0x550c7dc3243185beull, 0x12835b015807aa98ull);
288  QuadRound(bs0, bs1, 0x550c7dc3243185beull, 0x12835b015807aa98ull);
289  ShiftMessageA(am1, am2);
290  ShiftMessageA(bm1, bm2);
291  bm3 = am3 = _mm_set_epi64x(0x10000000000ull, 0x0ull);
292  QuadRound(as0, as1, 0xc19bf2749bdc06a7ull, 0x80deb1fe72be5d74ull);
293  QuadRound(bs0, bs1, 0xc19bf2749bdc06a7ull, 0x80deb1fe72be5d74ull);
294  ShiftMessageB(am2, am3, am0);
295  ShiftMessageB(bm2, bm3, bm0);
296  QuadRound(as0, as1, am0, 0x240ca1cc0fc19dc6ull, 0xefbe4786e49b69c1ull);
297  QuadRound(bs0, bs1, bm0, 0x240ca1cc0fc19dc6ull, 0xefbe4786e49b69c1ull);
298  ShiftMessageB(am3, am0, am1);
299  ShiftMessageB(bm3, bm0, bm1);
300  QuadRound(as0, as1, am1, 0x76f988da5cb0a9dcull, 0x4a7484aa2de92c6full);
301  QuadRound(bs0, bs1, bm1, 0x76f988da5cb0a9dcull, 0x4a7484aa2de92c6full);
302  ShiftMessageB(am0, am1, am2);
303  ShiftMessageB(bm0, bm1, bm2);
304  QuadRound(as0, as1, am2, 0xbf597fc7b00327c8ull, 0xa831c66d983e5152ull);
305  QuadRound(bs0, bs1, bm2, 0xbf597fc7b00327c8ull, 0xa831c66d983e5152ull);
306  ShiftMessageB(am1, am2, am3);
307  ShiftMessageB(bm1, bm2, bm3);
308  QuadRound(as0, as1, am3, 0x1429296706ca6351ull, 0xd5a79147c6e00bf3ull);
309  QuadRound(bs0, bs1, bm3, 0x1429296706ca6351ull, 0xd5a79147c6e00bf3ull);
310  ShiftMessageB(am2, am3, am0);
311  ShiftMessageB(bm2, bm3, bm0);
312  QuadRound(as0, as1, am0, 0x53380d134d2c6dfcull, 0x2e1b213827b70a85ull);
313  QuadRound(bs0, bs1, bm0, 0x53380d134d2c6dfcull, 0x2e1b213827b70a85ull);
314  ShiftMessageB(am3, am0, am1);
315  ShiftMessageB(bm3, bm0, bm1);
316  QuadRound(as0, as1, am1, 0x92722c8581c2c92eull, 0x766a0abb650a7354ull);
317  QuadRound(bs0, bs1, bm1, 0x92722c8581c2c92eull, 0x766a0abb650a7354ull);
318  ShiftMessageB(am0, am1, am2);
319  ShiftMessageB(bm0, bm1, bm2);
320  QuadRound(as0, as1, am2, 0xc76c51a3c24b8b70ull, 0xa81a664ba2bfe8A1ull);
321  QuadRound(bs0, bs1, bm2, 0xc76c51a3c24b8b70ull, 0xa81a664ba2bfe8A1ull);
322  ShiftMessageB(am1, am2, am3);
323  ShiftMessageB(bm1, bm2, bm3);
324  QuadRound(as0, as1, am3, 0x106aa070f40e3585ull, 0xd6990624d192e819ull);
325  QuadRound(bs0, bs1, bm3, 0x106aa070f40e3585ull, 0xd6990624d192e819ull);
326  ShiftMessageB(am2, am3, am0);
327  ShiftMessageB(bm2, bm3, bm0);
328  QuadRound(as0, as1, am0, 0x34b0bcb52748774cull, 0x1e376c0819a4c116ull);
329  QuadRound(bs0, bs1, bm0, 0x34b0bcb52748774cull, 0x1e376c0819a4c116ull);
330  ShiftMessageB(am3, am0, am1);
331  ShiftMessageB(bm3, bm0, bm1);
332  QuadRound(as0, as1, am1, 0x682e6ff35b9cca4full, 0x4ed8aa4a391c0cb3ull);
333  QuadRound(bs0, bs1, bm1, 0x682e6ff35b9cca4full, 0x4ed8aa4a391c0cb3ull);
334  ShiftMessageC(am0, am1, am2);
335  ShiftMessageC(bm0, bm1, bm2);
336  QuadRound(as0, as1, am2, 0x8cc7020884c87814ull, 0x78a5636f748f82eeull);
337  QuadRound(bs0, bs1, bm2, 0x8cc7020884c87814ull, 0x78a5636f748f82eeull);
338  ShiftMessageC(am1, am2, am3);
339  ShiftMessageC(bm1, bm2, bm3);
340  QuadRound(as0, as1, am3, 0xc67178f2bef9a3f7ull, 0xa4506ceb90befffaull);
341  QuadRound(bs0, bs1, bm3, 0xc67178f2bef9a3f7ull, 0xa4506ceb90befffaull);
342  as0 = _mm_add_epi32(as0, _mm_load_si128((const __m128i*)INIT0));
343  bs0 = _mm_add_epi32(bs0, _mm_load_si128((const __m128i*)INIT0));
344  as1 = _mm_add_epi32(as1, _mm_load_si128((const __m128i*)INIT1));
345  bs1 = _mm_add_epi32(bs1, _mm_load_si128((const __m128i*)INIT1));
346 
347  /* Extract hash into out */
348  Unshuffle(as0, as1);
349  Unshuffle(bs0, bs1);
350  Save(out, as0);
351  Save(out + 16, as1);
352  Save(out + 32, bs0);
353  Save(out + 48, bs1);
354 }
355 
356 }
357 
358 #endif
void Transform(uint32_t *s, const unsigned char *chunk, size_t blocks)
void Transform_2way(unsigned char *out, const unsigned char *in)
void Shuffle(I first, I last, R &&rng)
More efficient than using std::shuffle on a FastRandomContext.
Definition: random.h:265