2 The MIT License (MIT)
\r
4 Copyright (c) 2016 kste
\r
6 Permission is hereby granted, free of charge, to any person obtaining a copy
\r
7 of this software and associated documentation files (the "Software"), to deal
\r
8 in the Software without restriction, including without limitation the rights
\r
9 to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
\r
10 copies of the Software, and to permit persons to whom the Software is
\r
11 furnished to do so, subject to the following conditions:
\r
13 The above copyright notice and this permission notice shall be included in all
\r
14 copies or substantial portions of the Software.
\r
16 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
\r
17 IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
\r
18 FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
\r
19 AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
\r
20 LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
\r
21 OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
\r
24 Optimized Implementations for Haraka256 and Haraka512
\r
33 void load_constants() {
\r
34 //MIX_4 = _mm512_set_epi32(3,11,7,15,8,0,12,4,9,1,13,5,2,10,6,14); stupid thing goes the wrong way
\r
35 MIX_4 = _mm512_set_epi32(14,6,10,2,5,13,1,9,4,12,0,8,15,7,11,3);
\r
37 rc[0] = _mm512_set_epi32(0x0ed6eae6,0x2e7b4f08,0xbbf3bcaf,0xfd5b4f79,0x3402de2d,0x53f28498,0xcf029d60,0x9f029114,0x8b66b4e1,0x88f3a06b,0x640f6ba4,0x2f08f717,0x0684704c,0xe620c00a,0xb2c5fef0,0x75817b9d);
\r
38 rc[1] = _mm512_set_epi32(0x2924d9b0,0xafcacc07,0x675ffde2,0x1fc70b3b,0x67c28f43,0x5e2e7cd0,0xe2412761,0xda4fef1b,0x7eeacdee,0x6e9032b7,0x8d5335ed,0x2b8a057b,0xcbcfb0cb,0x4872448b,0x79eecd1c,0xbe397044);
\r
39 rc[2] = _mm512_set_epi32(0xfa0478a6,0xde6f5572,0x4aaa9ec8,0x5c9d2d8a,0xb2cc0bb9,0x941723bf,0x69028b2e,0x8df69800,0x1c30bf84,0xd4b7cd64,0x5b2a404f,0xad037e33,0xab4d63f1,0xe6867fe9,0xecdb8fca,0xb9d465ee);
\r
40 rc[3] = _mm512_set_epi32(0x21025ed8,0x9d199c4f,0x78a2c7e3,0x27e593ec,0xaf044988,0x4b050084,0x5f9600c9,0x9ca8eca6,0x1ea10344,0xf449a236,0x32d611ae,0xbb6a12ee,0xdfb49f2b,0x6b772a12,0x0efa4f2e,0x29129fd4);
\r
41 rc[4] = _mm512_set_epi32(0x9223973c,0x226b68bb,0x2caf92e8,0x36d1943a,0x5aca45c2,0x21300443,0x81c29153,0xf6fc9ac6,0x6260700d,0x6186b017,0x37f2efd9,0x10307d6b,0xbf3aaaf8,0xa759c9b7,0xb9282ecd,0x82d40173);
\r
42 rc[5] = _mm512_set_epi32(0x734bd3dc,0xe2e4d19c,0x2db91a4e,0xc72bf77d,0xbb606268,0xffeba09c,0x83e48de3,0xcb2212b1,0xdb863ce5,0xaef0c677,0x933dfddd,0x24e1128d,0xd3bf9238,0x225886eb,0x6cbab958,0xe51071b4);
\r
43 rc[6] = _mm512_set_epi32(0xcda75a17,0xd6de7d77,0x6d1be5b9,0xb88617f9,0x6df3614b,0x3c755977,0x8e5e2302,0x7eca472c,0xdba775a8,0xe707eff6,0x03b231dd,0x16eb6899,0x43bb47c3,0x61301b43,0x4b1415c4,0x2cb3924e);
\r
44 rc[7] = _mm512_set_epi32(0xf0b1a5a1,0x96e90cab,0x80bbbabc,0x63a4a350,0x2cee0c75,0x00da619c,0xe4ed0353,0x600ed0d9,0xcb1e6950,0xf957332b,0xa2531159,0x3bf327c1,0xec6b43f0,0x6ba8e9aa,0x9d6c069d,0xa946ee5d);
\r
45 rc[8] = _mm512_set_epi32(0x26f65241,0xcbe55438,0x43ce5918,0xffbaafde,0x34bb8a5b,0x5f427fd7,0xaeb6b779,0x360a16f6,0x17bb8f38,0xd554a40b,0x8814f3a8,0x2e75b442,0xae3db102,0x5e962988,0xab0dde30,0x938dca39);
\r
46 rc[9] = _mm512_set_epi32(0x756acc03,0x02288288,0x4ad6bdfd,0xe9c59da1,0xa0c1613c,0xba7ed22b,0xc173bc0f,0x48a659cf,0xae51a51a,0x1bdff7be,0x40c06e28,0x22901235,0x4ce99a54,0xb9f3026a,0xa2ca9cf7,0x839ec978);
\r
47 }//duck its endien ness // good programers make programs to make their programs i fixed it
\r
50 int test_implementations() {
\r
51 alignas(64) unsigned char in[64*8];
\r
52 alignas(64) unsigned char out256[32*8];
\r
53 alignas(64) unsigned char out512[32*8];
\r
54 alignas(64) unsigned char testvector256[32] = {0x80, 0x27, 0xcc, 0xb8, 0x79, 0x49, 0x77, 0x4b,
\r
55 0x78, 0xd0, 0x54, 0x5f, 0xb7, 0x2b, 0xf7, 0x0c,
\r
56 0x69, 0x5c, 0x2a, 0x09, 0x23, 0xcb, 0xd4, 0x7b,
\r
57 0xba, 0x11, 0x59, 0xef, 0xbf, 0x2b, 0x2c, 0x1c};
\r
59 alignas(64) unsigned char testvector512[32] = {0xbe, 0x7f, 0x72, 0x3b, 0x4e, 0x80, 0xa9, 0x98,
\r
60 0x13, 0xb2, 0x92, 0x28, 0x7f, 0x30, 0x6f, 0x62,
\r
61 0x5a, 0x6d, 0x57, 0x33, 0x1c, 0xae, 0x5f, 0x34,
\r
62 0xdd, 0x92, 0x77, 0xb0, 0x94, 0x5b, 0xe2, 0xaa};
\r
68 // Input for testvector
\r
69 for(i = 0; i < 512; i++) {
\r
74 haraka512_8x(out512, in);
\r
77 for(i = 0; i < 32; i++) {
\r
78 if (out512[i % 32] != testvector512[i]) {
\r
79 printf("Error: testvector incorrect.\n");
\r
86 void haraka256(unsigned char *out, const unsigned char *in) {
\r
90 s[1] = LOAD(in + 16);
\r
92 AES2(s[0], s[1], 0);
\r
95 AES2(s[0], s[1], 4);
\r
98 AES2(s[0], s[1], 8);
\r
101 AES2(s[0], s[1], 12);
\r
104 AES2(s[0], s[1], 16);
\r
107 s[0] = _mm_xor_si128(s[0], LOAD(in));
\r
108 s[1] = _mm_xor_si128(s[1], LOAD(in + 16));
\r
111 STORE(out + 16, s[1]);
\r
114 void haraka256_keyed(unsigned char *out, const unsigned char *in, const u128 *rc) {
\r
118 s[1] = LOAD(in + 16);
\r
121 s[0] = _mm_xor_si128(s[0], LOAD(in));
\r
122 s[1] = _mm_xor_si128(s[1], LOAD(in + 16));
\r
125 STORE(out + 16, s[1]);
\r
128 void haraka256_4x(unsigned char *out, const unsigned char *in) {
\r
129 __m128i s[4][2], tmp;
\r
131 s[0][0] = LOAD(in);
\r
132 s[0][1] = LOAD(in + 16);
\r
133 s[1][0] = LOAD(in + 32);
\r
134 s[1][1] = LOAD(in + 48);
\r
135 s[2][0] = LOAD(in + 64);
\r
136 s[2][1] = LOAD(in + 80);
\r
137 s[3][0] = LOAD(in + 96);
\r
140 MIX2(s[0][0], s[0][1]);
\r
141 MIX2(s[1][0], s[1][1]);
\r
142 MIX2(s[2][0], s[2][1]);
\r
143 MIX2(s[3][0], s[3][1]);
\r
146 s[0][0] = _mm_xor_si128(s[0][0], LOAD(in));
\r
147 s[0][1] = _mm_xor_si128(s[0][1], LOAD(in + 16));
\r
148 s[1][0] = _mm_xor_si128(s[1][0], LOAD(in + 32));
\r
149 s[1][1] = _mm_xor_si128(s[1][1], LOAD(in + 48));
\r
150 s[2][0] = _mm_xor_si128(s[2][0], LOAD(in + 64));
\r
151 s[2][1] = _mm_xor_si128(s[2][1], LOAD(in + 80));
\r
152 s[3][0] = _mm_xor_si128(s[3][0], LOAD(in + 96));
\r
153 s[3][1] = _mm_xor_si128(s[3][1], LOAD(in + 112));
\r
155 STORE(out, s[0][0]);
\r
156 STORE(out + 16, s[0][1]);
\r
157 STORE(out + 32, s[1][0]);
\r
158 STORE(out + 48, s[1][1]);
\r
159 STORE(out + 64, s[2][0]);
\r
160 STORE(out + 80, s[2][1]);
\r
161 STORE(out + 96, s[3][0]);
\r
162 STORE(out + 112, s[3][1]);
\r
165 void haraka256_8x(unsigned char *out, const unsigned char *in) {
\r
166 // This is faster on Skylake, the code below is faster on Haswell.
\r
167 haraka256_4x(out, in);
\r
168 haraka256_4x(out + 128, in + 128);
\r
170 // __m128i s[8][2], tmp;
\r
174 // s[0][0] = LOAD(in);
\r
175 // s[0][1] = LOAD(in + 16);
\r
176 // s[1][0] = LOAD(in + 32);
\r
177 // s[1][1] = LOAD(in + 48);
\r
178 // s[2][0] = LOAD(in + 64);
\r
179 // s[2][1] = LOAD(in + 80);
\r
180 // s[3][0] = LOAD(in + 96);
\r
181 // s[3][1] = LOAD(in + 112);
\r
182 // s[4][0] = LOAD(in + 128);
\r
183 // s[4][1] = LOAD(in + 144);
\r
184 // s[5][0] = LOAD(in + 160);
\r
185 // s[5][1] = LOAD(in + 176);
\r
186 // s[6][0] = LOAD(in + 192);
\r
187 // s[6][1] = LOAD(in + 208);
\r
188 // s[7][0] = LOAD(in + 224);
\r
189 // s[7][1] = LOAD(in + 240);
\r
192 // AES2_8x(s[0], s[1], s[2], s[3], s[4], s[5], s[6], s[7], 0);
\r
194 // MIX2(s[0][0], s[0][1]);
\r
195 // MIX2(s[1][0], s[1][1]);
\r
196 // MIX2(s[2][0], s[2][1]);
\r
197 // MIX2(s[3][0], s[3][1]);
\r
198 // MIX2(s[4][0], s[4][1]);
\r
199 // MIX2(s[5][0], s[5][1]);
\r
200 // MIX2(s[6][0], s[6][1]);
\r
201 // MIX2(s[7][0], s[7][1]);
\r
205 // AES2_8x(s[0], s[1], s[2], s[3], s[4], s[5], s[6], s[7], 4);
\r
207 // MIX2(s[0][0], s[0][1]);
\r
208 // MIX2(s[1][0], s[1][1]);
\r
209 // MIX2(s[2][0], s[2][1]);
\r
210 // MIX2(s[3][0], s[3][1]);
\r
211 // MIX2(s[4][0], s[4][1]);
\r
212 // MIX2(s[5][0], s[5][1]);
\r
213 // MIX2(s[6][0], s[6][1]);
\r
214 // MIX2(s[7][0], s[7][1]);
\r
217 // AES2_8x(s[0], s[1], s[2], s[3], s[4], s[5], s[6], s[7], 8);
\r
219 // MIX2(s[0][0], s[0][1]);
\r
220 // MIX2(s[1][0], s[1][1]);
\r
221 // MIX2(s[2][0], s[2][1]);
\r
222 // MIX2(s[3][0], s[3][1]);
\r
223 // MIX2(s[4][0], s[4][1]);
\r
224 // MIX2(s[5][0], s[5][1]);
\r
225 // MIX2(s[6][0], s[6][1]);
\r
226 // MIX2(s[7][0], s[7][1]);
\r
229 // AES2_8x(s[0], s[1], s[2], s[3], s[4], s[5], s[6], s[7], 12);
\r
231 // MIX2(s[0][0], s[0][1]);
\r
232 // MIX2(s[1][0], s[1][1]);
\r
233 // MIX2(s[2][0], s[2][1]);
\r
234 // MIX2(s[3][0], s[3][1]);
\r
235 // MIX2(s[4][0], s[4][1]);
\r
236 // MIX2(s[5][0], s[5][1]);
\r
237 // MIX2(s[6][0], s[6][1]);
\r
238 // MIX2(s[7][0], s[7][1]);
\r
241 // AES2_8x(s[0], s[1], s[2], s[3], s[4], s[5], s[6], s[7], 16);
\r
243 // MIX2(s[0][0], s[0][1]);
\r
244 // MIX2(s[1][0], s[1][1]);
\r
245 // MIX2(s[2][0], s[2][1]);
\r
246 // MIX2(s[3][0], s[3][1]);
\r
247 // MIX2(s[4][0], s[4][1]);
\r
248 // MIX2(s[5][0], s[5][1]);
\r
249 // MIX2(s[6][0], s[6][1]);
\r
250 // MIX2(s[7][0], s[7][1]);
\r
253 // s[0][0] = _mm_xor_si128(s[0][0], LOAD(in));
\r
254 // s[0][1] = _mm_xor_si128(s[0][1], LOAD(in + 16));
\r
255 // s[1][0] = _mm_xor_si128(s[1][0], LOAD(in + 32));
\r
256 // s[1][1] = _mm_xor_si128(s[1][1], LOAD(in + 48));
\r
257 // s[2][0] = _mm_xor_si128(s[2][0], LOAD(in + 64));
\r
258 // s[2][1] = _mm_xor_si128(s[2][1], LOAD(in + 80));
\r
259 // s[3][0] = _mm_xor_si128(s[3][0], LOAD(in + 96));
\r
260 // s[3][1] = _mm_xor_si128(s[3][1], LOAD(in + 112));
\r
261 // s[4][0] = _mm_xor_si128(s[4][0], LOAD(in + 128));
\r
262 // s[4][1] = _mm_xor_si128(s[4][1], LOAD(in + 144));
\r
263 // s[5][0] = _mm_xor_si128(s[5][0], LOAD(in + 160));
\r
264 // s[5][1] = _mm_xor_si128(s[5][1], LOAD(in + 176));
\r
265 // s[6][0] = _mm_xor_si128(s[6][0], LOAD(in + 192));
\r
266 // s[6][1] = _mm_xor_si128(s[6][1], LOAD(in + 208));
\r
267 // s[7][0] = _mm_xor_si128(s[7][0], LOAD(in + 224));
\r
268 // s[7][1] = _mm_xor_si128(s[7][1], LOAD(in + 240));
\r
270 // STORE(out, s[0][0]);
\r
271 // STORE(out + 16, s[0][1]);
\r
272 // STORE(out + 32, s[1][0]);
\r
273 // STORE(out + 48, s[1][1]);
\r
274 // STORE(out + 64, s[2][0]);
\r
275 // STORE(out + 80, s[2][1]);
\r
276 // STORE(out + 96, s[3][0]);
\r
277 // STORE(out + 112, s[3][1]);
\r
278 // STORE(out + 128, s[4][0]);
\r
279 // STORE(out + 144, s[4][1]);
\r
280 // STORE(out + 160, s[5][0]);
\r
281 // STORE(out + 176, s[5][1]);
\r
282 // STORE(out + 192, s[6][0]);
\r
283 // STORE(out + 208, s[6][1]);
\r
284 // STORE(out + 224, s[7][0]);
\r
285 // STORE(out + 240, s[7][1]);
\r
288 static void phex(uint8_t* str)
\r
293 for (i = 0; i < len; ++i)
\r
294 printf("%hx, ", str[i]);
\r
297 void haraka512(unsigned char *out, const unsigned char *in) {
\r
317 s = _mm512_xor_si512(s, i);
\r
319 TRUNCSTORE(out, s);
\r
322 void haraka512_zero(unsigned char *out, const unsigned char *in) {
\r
341 s = _mm512_xor_si512(s, i);
\r
343 TRUNCSTORE(out, s);
\r
346 void haraka512_4x(unsigned char *out, const unsigned char *in) {
\r
350 s[1] = LOAD(in + 64);
\r
351 s[2] = LOAD(in + 128);
\r
352 s[3] = LOAD(in + 192);
\r
391 s[0] = _mm512_xor_si512(s[0], i[0]);
\r
392 s[1] = _mm512_xor_si512(s[1], i[1]);
\r
393 s[2] = _mm512_xor_si512(s[2], i[2]);
\r
394 s[3] = _mm512_xor_si512(s[3], i[3]);
\r
397 TRUNCSTORE(out, s[0]);
\r
398 TRUNCSTORE(out + 32, s[1]);
\r
399 TRUNCSTORE(out + 64, s[2]);
\r
400 TRUNCSTORE(out + 96, s[3]);
\r
403 void haraka512_8x(unsigned char *out, const unsigned char *in) {
\r
404 // This is faster on Skylake, the code below is faster on Haswell.
\r
405 haraka512_4x(out, in);
\r
406 haraka512_4x(out + 128, in + 256);
\r
408 // u128 s[8][4], tmp;
\r
410 // s[0][0] = LOAD(in);
\r
411 // s[0][1] = LOAD(in + 16);
\r
412 // s[0][2] = LOAD(in + 32);
\r
413 // s[0][3] = LOAD(in + 48);
\r
414 // s[1][0] = LOAD(in + 64);
\r
415 // s[1][1] = LOAD(in + 80);
\r
416 // s[1][2] = LOAD(in + 96);
\r
417 // s[1][3] = LOAD(in + 112);
\r
418 // s[2][0] = LOAD(in + 128);
\r
419 // s[2][1] = LOAD(in + 144);
\r
420 // s[2][2] = LOAD(in + 160);
\r
421 // s[2][3] = LOAD(in + 176);
\r
422 // s[3][0] = LOAD(in + 192);
\r
423 // s[3][1] = LOAD(in + 208);
\r
424 // s[3][2] = LOAD(in + 224);
\r
425 // s[3][3] = LOAD(in + 240);
\r
426 // s[4][0] = LOAD(in + 256);
\r
427 // s[4][1] = LOAD(in + 272);
\r
428 // s[4][2] = LOAD(in + 288);
\r
429 // s[4][3] = LOAD(in + 304);
\r
430 // s[5][0] = LOAD(in + 320);
\r
431 // s[5][1] = LOAD(in + 336);
\r
432 // s[5][2] = LOAD(in + 352);
\r
433 // s[5][3] = LOAD(in + 368);
\r
434 // s[6][0] = LOAD(in + 384);
\r
435 // s[6][1] = LOAD(in + 400);
\r
436 // s[6][2] = LOAD(in + 416);
\r
437 // s[6][3] = LOAD(in + 432);
\r
438 // s[7][0] = LOAD(in + 448);
\r
439 // s[7][1] = LOAD(in + 464);
\r
440 // s[7][2] = LOAD(in + 480);
\r
441 // s[7][3] = LOAD(in + 496);
\r
443 // AES4_8x(s[0], s[1], s[2], s[3], s[4], s[5], s[6], s[7], 0);
\r
444 // MIX4(s[0][0], s[0][1], s[0][2], s[0][3]);
\r
445 // MIX4(s[1][0], s[1][1], s[1][2], s[1][3]);
\r
446 // MIX4(s[2][0], s[2][1], s[2][2], s[2][3]);
\r
447 // MIX4(s[3][0], s[3][1], s[3][2], s[3][3]);
\r
448 // MIX4(s[4][0], s[4][1], s[4][2], s[4][3]);
\r
449 // MIX4(s[5][0], s[5][1], s[5][2], s[5][3]);
\r
450 // MIX4(s[6][0], s[6][1], s[6][2], s[6][3]);
\r
451 // MIX4(s[7][0], s[7][1], s[7][2], s[7][3]);
\r
453 // AES4_8x(s[0], s[1], s[2], s[3], s[4], s[5], s[6], s[7], 8);
\r
454 // MIX4(s[0][0], s[0][1], s[0][2], s[0][3]);
\r
455 // MIX4(s[1][0], s[1][1], s[1][2], s[1][3]);
\r
456 // MIX4(s[2][0], s[2][1], s[2][2], s[2][3]);
\r
457 // MIX4(s[3][0], s[3][1], s[3][2], s[3][3]);
\r
458 // MIX4(s[4][0], s[4][1], s[4][2], s[4][3]);
\r
459 // MIX4(s[5][0], s[5][1], s[5][2], s[5][3]);
\r
460 // MIX4(s[6][0], s[6][1], s[6][2], s[6][3]);
\r
461 // MIX4(s[7][0], s[7][1], s[7][2], s[7][3]);
\r
463 // AES4_8x(s[0], s[1], s[2], s[3], s[4], s[5], s[6], s[7], 16);
\r
464 // MIX4(s[0][0], s[0][1], s[0][2], s[0][3]);
\r
465 // MIX4(s[1][0], s[1][1], s[1][2], s[1][3]);
\r
466 // MIX4(s[2][0], s[2][1], s[2][2], s[2][3]);
\r
467 // MIX4(s[3][0], s[3][1], s[3][2], s[3][3]);
\r
468 // MIX4(s[4][0], s[4][1], s[4][2], s[4][3]);
\r
469 // MIX4(s[5][0], s[5][1], s[5][2], s[5][3]);
\r
470 // MIX4(s[6][0], s[6][1], s[6][2], s[6][3]);
\r
471 // MIX4(s[7][0], s[7][1], s[7][2], s[7][3]);
\r
473 // AES4_8x(s[0], s[1], s[2], s[3], s[4], s[5], s[6], s[7], 24);
\r
474 // MIX4(s[0][0], s[0][1], s[0][2], s[0][3]);
\r
475 // MIX4(s[1][0], s[1][1], s[1][2], s[1][3]);
\r
476 // MIX4(s[2][0], s[2][1], s[2][2], s[2][3]);
\r
477 // MIX4(s[3][0], s[3][1], s[3][2], s[3][3]);
\r
478 // MIX4(s[4][0], s[4][1], s[4][2], s[4][3]);
\r
479 // MIX4(s[5][0], s[5][1], s[5][2], s[5][3]);
\r
480 // MIX4(s[6][0], s[6][1], s[6][2], s[6][3]);
\r
481 // MIX4(s[7][0], s[7][1], s[7][2], s[7][3]);
\r
483 // AES4_8x(s[0], s[1], s[2], s[3], s[4], s[5], s[6], s[7], 32);
\r
484 // MIX4(s[0][0], s[0][1], s[0][2], s[0][3]);
\r
485 // MIX4(s[1][0], s[1][1], s[1][2], s[1][3]);
\r
486 // MIX4(s[2][0], s[2][1], s[2][2], s[2][3]);
\r
487 // MIX4(s[3][0], s[3][1], s[3][2], s[3][3]);
\r
488 // MIX4(s[4][0], s[4][1], s[4][2], s[4][3]);
\r
489 // MIX4(s[5][0], s[5][1], s[5][2], s[5][3]);
\r
490 // MIX4(s[6][0], s[6][1], s[6][2], s[6][3]);
\r
491 // MIX4(s[7][0], s[7][1], s[7][2], s[7][3]);
\r
494 // s[0][0] = _mm_xor_si128(s[0][0], LOAD(in));
\r
495 // s[0][1] = _mm_xor_si128(s[0][1], LOAD(in + 16));
\r
496 // s[0][2] = _mm_xor_si128(s[0][2], LOAD(in + 32));
\r
497 // s[0][3] = _mm_xor_si128(s[0][3], LOAD(in + 48));
\r
498 // s[1][0] = _mm_xor_si128(s[1][0], LOAD(in + 64));
\r
499 // s[1][1] = _mm_xor_si128(s[1][1], LOAD(in + 80));
\r
500 // s[1][2] = _mm_xor_si128(s[1][2], LOAD(in + 96));
\r
501 // s[1][3] = _mm_xor_si128(s[1][3], LOAD(in + 112));
\r
502 // s[2][0] = _mm_xor_si128(s[2][0], LOAD(in + 128));
\r
503 // s[2][1] = _mm_xor_si128(s[2][1], LOAD(in + 144));
\r
504 // s[2][2] = _mm_xor_si128(s[2][2], LOAD(in + 160));
\r
505 // s[2][3] = _mm_xor_si128(s[2][3], LOAD(in + 176));
\r
506 // s[3][0] = _mm_xor_si128(s[3][0], LOAD(in + 192));
\r
507 // s[3][1] = _mm_xor_si128(s[3][1], LOAD(in + 208));
\r
508 // s[3][2] = _mm_xor_si128(s[3][2], LOAD(in + 224));
\r
509 // s[3][3] = _mm_xor_si128(s[3][3], LOAD(in + 240));
\r
510 // s[4][0] = _mm_xor_si128(s[4][0], LOAD(in + 256));
\r
511 // s[4][1] = _mm_xor_si128(s[4][1], LOAD(in + 272));
\r
512 // s[4][2] = _mm_xor_si128(s[4][2], LOAD(in + 288));
\r
513 // s[4][3] = _mm_xor_si128(s[4][3], LOAD(in + 304));
\r
514 // s[5][0] = _mm_xor_si128(s[5][0], LOAD(in + 320));
\r
515 // s[5][1] = _mm_xor_si128(s[5][1], LOAD(in + 336));
\r
516 // s[5][2] = _mm_xor_si128(s[5][2], LOAD(in + 352));
\r
517 // s[5][3] = _mm_xor_si128(s[5][3], LOAD(in + 368));
\r
518 // s[6][0] = _mm_xor_si128(s[6][0], LOAD(in + 384));
\r
519 // s[6][1] = _mm_xor_si128(s[6][1], LOAD(in + 400));
\r
520 // s[6][2] = _mm_xor_si128(s[6][2], LOAD(in + 416));
\r
521 // s[6][3] = _mm_xor_si128(s[6][3], LOAD(in + 432));
\r
522 // s[7][0] = _mm_xor_si128(s[7][0], LOAD(in + 448));
\r
523 // s[7][1] = _mm_xor_si128(s[7][1], LOAD(in + 464));
\r
524 // s[7][2] = _mm_xor_si128(s[7][2], LOAD(in + 480));
\r
525 // s[7][3] = _mm_xor_si128(s[7][3], LOAD(in + 496));
\r
527 // TRUNCSTORE(out, s[0][0], s[0][1], s[0][2], s[0][3]);
\r
528 // TRUNCSTORE(out + 32, s[1][0], s[1][1], s[1][2], s[1][3]);
\r
529 // TRUNCSTORE(out + 64, s[2][0], s[2][1], s[2][2], s[2][3]);
\r
530 // TRUNCSTORE(out + 96, s[3][0], s[3][1], s[3][2], s[3][3]);
\r
531 // TRUNCSTORE(out + 128, s[4][0], s[4][1], s[4][2], s[4][3]);
\r
532 // TRUNCSTORE(out + 160, s[5][0], s[5][1], s[5][2], s[5][3]);
\r
533 // TRUNCSTORE(out + 192, s[6][0], s[6][1], s[6][2], s[6][3]);
\r
534 // TRUNCSTORE(out + 224, s[7][0], s[7][1], s[7][2], s[7][3]);
\r