Blame view

Socket/cryptopp/rijndael.cpp 36.5 KB
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
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
// rijndael.cpp - modified by Chris Morgan <cmorgan@wpi.edu>
// and Wei Dai from Paulo Baretto's Rijndael implementation
// The original code and all modifications are in the public domain.

// use "cl /EP /P /DCRYPTOPP_GENERATE_X64_MASM rijndael.cpp" to generate MASM code

/*
July 2010: Added support for AES-NI instructions via compiler intrinsics.
*/

/*
Feb 2009: The x86/x64 assembly code was rewritten in by Wei Dai to do counter mode 
caching, which was invented by Hongjun Wu and popularized by Daniel J. Bernstein 
and Peter Schwabe in their paper "New AES software speed records". The round 
function was also modified to include a trick similar to one in Brian Gladman's 
x86 assembly code, doing an 8-bit register move to minimize the number of 
register spills. Also switched to compressed tables and copying round keys to 
the stack.

The C++ implementation now uses compressed tables if 
CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS is defined.
*/

/*
July 2006: Defense against timing attacks was added in by Wei Dai.

The code now uses smaller tables in the first and last rounds,
and preloads them into L1 cache before usage (by loading at least 
one element in each cache line). 

We try to delay subsequent accesses to each table (used in the first 
and last rounds) until all of the table has been preloaded. Hopefully
the compiler isn't smart enough to optimize that code away.

After preloading the table, we also try not to access any memory location
other than the table and the stack, in order to prevent table entries from 
being unloaded from L1 cache, until that round is finished.
(Some popular CPUs have 2-way associative caches.)
*/

// This is the original introductory comment:

/**
 * version 3.0 (December 2000)
 *
 * Optimised ANSI C code for the Rijndael cipher (now AES)
 *
 * author Vincent Rijmen <vincent.rijmen@esat.kuleuven.ac.be>
 * author Antoon Bosselaers <antoon.bosselaers@esat.kuleuven.ac.be>
 * author Paulo Barreto <paulo.barreto@terra.com.br>
 *
 * This code is hereby placed in the public domain.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS ''AS IS'' AND ANY EXPRESS
 * OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE
 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE
 * OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE,
 * EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 */

#include "pch.h"

#ifndef CRYPTOPP_IMPORTS
#ifndef CRYPTOPP_GENERATE_X64_MASM

#include "rijndael.h"
#include "misc.h"
#include "cpu.h"

NAMESPACE_BEGIN(CryptoPP)

#ifdef CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS
#if CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE || defined(CRYPTOPP_X64_MASM_AVAILABLE)
namespace rdtable {CRYPTOPP_ALIGN_DATA(16) word64 Te[256+2];}
using namespace rdtable;
#else
static word64 Te[256];
#endif
static word64 Td[256];
#else
static word32 Te[256*4], Td[256*4];
#endif
static volatile bool s_TeFilled = false, s_TdFilled = false;

// ************************* Portable Code ************************************

#define QUARTER_ROUND(L, T, t, a, b, c, d)	\
	a ^= L(T, 3, byte(t)); t >>= 8;\
	b ^= L(T, 2, byte(t)); t >>= 8;\
	c ^= L(T, 1, byte(t)); t >>= 8;\
	d ^= L(T, 0, t);

#define QUARTER_ROUND_LE(t, a, b, c, d)	\
	tempBlock[a] = ((byte *)(Te+byte(t)))[1]; t >>= 8;\
	tempBlock[b] = ((byte *)(Te+byte(t)))[1]; t >>= 8;\
	tempBlock[c] = ((byte *)(Te+byte(t)))[1]; t >>= 8;\
	tempBlock[d] = ((byte *)(Te+t))[1];

#ifdef CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS
	#define QUARTER_ROUND_LD(t, a, b, c, d)	\
		tempBlock[a] = ((byte *)(Td+byte(t)))[GetNativeByteOrder()*7]; t >>= 8;\
		tempBlock[b] = ((byte *)(Td+byte(t)))[GetNativeByteOrder()*7]; t >>= 8;\
		tempBlock[c] = ((byte *)(Td+byte(t)))[GetNativeByteOrder()*7]; t >>= 8;\
		tempBlock[d] = ((byte *)(Td+t))[GetNativeByteOrder()*7];
#else
	#define QUARTER_ROUND_LD(t, a, b, c, d)	\
		tempBlock[a] = Sd[byte(t)]; t >>= 8;\
		tempBlock[b] = Sd[byte(t)]; t >>= 8;\
		tempBlock[c] = Sd[byte(t)]; t >>= 8;\
		tempBlock[d] = Sd[t];
#endif

#define QUARTER_ROUND_E(t, a, b, c, d)		QUARTER_ROUND(TL_M, Te, t, a, b, c, d)
#define QUARTER_ROUND_D(t, a, b, c, d)		QUARTER_ROUND(TL_M, Td, t, a, b, c, d)

#ifdef IS_LITTLE_ENDIAN
	#define QUARTER_ROUND_FE(t, a, b, c, d)		QUARTER_ROUND(TL_F, Te, t, d, c, b, a)
	#define QUARTER_ROUND_FD(t, a, b, c, d)		QUARTER_ROUND(TL_F, Td, t, d, c, b, a)
	#ifdef CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS
		#define TL_F(T, i, x)	(*(word32 *)((byte *)T + x*8 + (6-i)%4+1))
		#define TL_M(T, i, x)	(*(word32 *)((byte *)T + x*8 + (i+3)%4+1))
	#else
		#define TL_F(T, i, x)	rotrFixed(T[x], (3-i)*8)
		#define TL_M(T, i, x)	T[i*256 + x]
	#endif
#else
	#define QUARTER_ROUND_FE(t, a, b, c, d)		QUARTER_ROUND(TL_F, Te, t, a, b, c, d)
	#define QUARTER_ROUND_FD(t, a, b, c, d)		QUARTER_ROUND(TL_F, Td, t, a, b, c, d)
	#ifdef CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS
		#define TL_F(T, i, x)	(*(word32 *)((byte *)T + x*8 + (4-i)%4))
		#define TL_M			TL_F
	#else
		#define TL_F(T, i, x)	rotrFixed(T[x], i*8)
		#define TL_M(T, i, x)	T[i*256 + x]
	#endif
#endif


#define f2(x)   ((x<<1)^(((x>>7)&1)*0x11b))
#define f4(x)   ((x<<2)^(((x>>6)&1)*0x11b)^(((x>>6)&2)*0x11b))
#define f8(x)   ((x<<3)^(((x>>5)&1)*0x11b)^(((x>>5)&2)*0x11b)^(((x>>5)&4)*0x11b))

#define f3(x)   (f2(x) ^ x)
#define f9(x)   (f8(x) ^ x)
#define fb(x)   (f8(x) ^ f2(x) ^ x)
#define fd(x)   (f8(x) ^ f4(x) ^ x)
#define fe(x)   (f8(x) ^ f4(x) ^ f2(x))

void Rijndael::Base::FillEncTable()
{
	for (int i=0; i<256; i++)
	{
		byte x = Se[i];
#ifdef CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS
		word32 y = word32(x)<<8 | word32(x)<<16 | word32(f2(x))<<24;
		Te[i] = word64(y | f3(x))<<32 | y;
#else
		word32 y = f3(x) | word32(x)<<8 | word32(x)<<16 | word32(f2(x))<<24;
		for (int j=0; j<4; j++)
		{
			Te[i+j*256] = y;
			y = rotrFixed(y, 8);
		}
#endif
	}
#if CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE || defined(CRYPTOPP_X64_MASM_AVAILABLE)
	Te[256] = Te[257] = 0;
#endif
	s_TeFilled = true;
}

void Rijndael::Base::FillDecTable()
{
	for (int i=0; i<256; i++)
	{
		byte x = Sd[i];
#ifdef CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS
		word32 y = word32(fd(x))<<8 | word32(f9(x))<<16 | word32(fe(x))<<24;
		Td[i] = word64(y | fb(x))<<32 | y | x;
#else
		word32 y = fb(x) | word32(fd(x))<<8 | word32(f9(x))<<16 | word32(fe(x))<<24;;
		for (int j=0; j<4; j++)
		{
			Td[i+j*256] = y;
			y = rotrFixed(y, 8);
		}
#endif
	}
	s_TdFilled = true;
}

void Rijndael::Base::UncheckedSetKey(const byte *userKey, unsigned int keylen, const NameValuePairs &)
{
	AssertValidKeyLength(keylen);

	m_rounds = keylen/4 + 6;
	m_key.New(4*(m_rounds+1));

	word32 *rk = m_key;

#if CRYPTOPP_BOOL_AESNI_INTRINSICS_AVAILABLE && (!defined(_MSC_VER) || _MSC_VER >= 1600 || CRYPTOPP_BOOL_X86)
	// MSVC 2008 SP1 generates bad code for _mm_extract_epi32() when compiling for X64
	if (HasAESNI())
	{
		static const word32 rcLE[] = {
			0x01, 0x02, 0x04, 0x08,
			0x10, 0x20, 0x40, 0x80,
			0x1B, 0x36, /* for 128-bit blocks, Rijndael never uses more than 10 rcon values */
		};
		const word32 *rc = rcLE;

		__m128i temp = _mm_loadu_si128((__m128i *)(userKey+keylen-16));
		memcpy(rk, userKey, keylen);

		while (true)
		{
			rk[keylen/4] = rk[0] ^ _mm_extract_epi32(_mm_aeskeygenassist_si128(temp, 0), 3) ^ *(rc++);
			rk[keylen/4+1] = rk[1] ^ rk[keylen/4];
			rk[keylen/4+2] = rk[2] ^ rk[keylen/4+1];
			rk[keylen/4+3] = rk[3] ^ rk[keylen/4+2];

			if (rk + keylen/4 + 4 == m_key.end())
				break;

			if (keylen == 24)
			{
				rk[10] = rk[ 4] ^ rk[ 9];
				rk[11] = rk[ 5] ^ rk[10];
				temp = _mm_insert_epi32(temp, rk[11], 3);
			}
			else if (keylen == 32)
			{
				temp = _mm_insert_epi32(temp, rk[11], 3);
    			rk[12] = rk[ 4] ^ _mm_extract_epi32(_mm_aeskeygenassist_si128(temp, 0), 2);
    			rk[13] = rk[ 5] ^ rk[12];
    			rk[14] = rk[ 6] ^ rk[13];
    			rk[15] = rk[ 7] ^ rk[14];
				temp = _mm_insert_epi32(temp, rk[15], 3);
			}
			else
				temp = _mm_insert_epi32(temp, rk[7], 3);

			rk += keylen/4;
		}

		if (!IsForwardTransformation())
		{
			rk = m_key;
			unsigned int i, j;

			std::swap(*(__m128i *)(rk), *(__m128i *)(rk+4*m_rounds));

			for (i = 4, j = 4*m_rounds-4; i < j; i += 4, j -= 4)
			{
				temp = _mm_aesimc_si128(*(__m128i *)(rk+i));
				*(__m128i *)(rk+i) = _mm_aesimc_si128(*(__m128i *)(rk+j));
				*(__m128i *)(rk+j) = temp;
			}

			*(__m128i *)(rk+i) = _mm_aesimc_si128(*(__m128i *)(rk+i));
		}

		return;
	}
#endif

	GetUserKey(BIG_ENDIAN_ORDER, rk, keylen/4, userKey, keylen);
	const word32 *rc = rcon;
	word32 temp;

	while (true)
	{
		temp  = rk[keylen/4-1];
		word32 x = (word32(Se[GETBYTE(temp, 2)]) << 24) ^ (word32(Se[GETBYTE(temp, 1)]) << 16) ^ (word32(Se[GETBYTE(temp, 0)]) << 8) ^ Se[GETBYTE(temp, 3)];
		rk[keylen/4] = rk[0] ^ x ^ *(rc++);
		rk[keylen/4+1] = rk[1] ^ rk[keylen/4];
		rk[keylen/4+2] = rk[2] ^ rk[keylen/4+1];
		rk[keylen/4+3] = rk[3] ^ rk[keylen/4+2];

		if (rk + keylen/4 + 4 == m_key.end())
			break;

		if (keylen == 24)
		{
			rk[10] = rk[ 4] ^ rk[ 9];
			rk[11] = rk[ 5] ^ rk[10];
		}
		else if (keylen == 32)
		{
    		temp = rk[11];
    		rk[12] = rk[ 4] ^ (word32(Se[GETBYTE(temp, 3)]) << 24) ^ (word32(Se[GETBYTE(temp, 2)]) << 16) ^ (word32(Se[GETBYTE(temp, 1)]) << 8) ^ Se[GETBYTE(temp, 0)];
    		rk[13] = rk[ 5] ^ rk[12];
    		rk[14] = rk[ 6] ^ rk[13];
    		rk[15] = rk[ 7] ^ rk[14];
		}
		rk += keylen/4;
	}

	rk = m_key;

	if (IsForwardTransformation())
	{
		if (!s_TeFilled)
			FillEncTable();

		ConditionalByteReverse(BIG_ENDIAN_ORDER, rk, rk, 16);
		ConditionalByteReverse(BIG_ENDIAN_ORDER, rk + m_rounds*4, rk + m_rounds*4, 16);
	}
	else
	{
		if (!s_TdFilled)
			FillDecTable();

		unsigned int i, j;

#define InverseMixColumn(x)		TL_M(Td, 0, Se[GETBYTE(x, 3)]) ^ TL_M(Td, 1, Se[GETBYTE(x, 2)]) ^ TL_M(Td, 2, Se[GETBYTE(x, 1)]) ^ TL_M(Td, 3, Se[GETBYTE(x, 0)])

		for (i = 4, j = 4*m_rounds-4; i < j; i += 4, j -= 4)
		{
			temp = InverseMixColumn(rk[i    ]); rk[i    ] = InverseMixColumn(rk[j    ]); rk[j    ] = temp;
			temp = InverseMixColumn(rk[i + 1]); rk[i + 1] = InverseMixColumn(rk[j + 1]); rk[j + 1] = temp;
			temp = InverseMixColumn(rk[i + 2]); rk[i + 2] = InverseMixColumn(rk[j + 2]); rk[j + 2] = temp;
			temp = InverseMixColumn(rk[i + 3]); rk[i + 3] = InverseMixColumn(rk[j + 3]); rk[j + 3] = temp;
		}

		rk[i+0] = InverseMixColumn(rk[i+0]);
		rk[i+1] = InverseMixColumn(rk[i+1]);
		rk[i+2] = InverseMixColumn(rk[i+2]);
		rk[i+3] = InverseMixColumn(rk[i+3]);

		temp = ConditionalByteReverse(BIG_ENDIAN_ORDER, rk[0]); rk[0] = ConditionalByteReverse(BIG_ENDIAN_ORDER, rk[4*m_rounds+0]); rk[4*m_rounds+0] = temp;
		temp = ConditionalByteReverse(BIG_ENDIAN_ORDER, rk[1]); rk[1] = ConditionalByteReverse(BIG_ENDIAN_ORDER, rk[4*m_rounds+1]); rk[4*m_rounds+1] = temp;
		temp = ConditionalByteReverse(BIG_ENDIAN_ORDER, rk[2]); rk[2] = ConditionalByteReverse(BIG_ENDIAN_ORDER, rk[4*m_rounds+2]); rk[4*m_rounds+2] = temp;
		temp = ConditionalByteReverse(BIG_ENDIAN_ORDER, rk[3]); rk[3] = ConditionalByteReverse(BIG_ENDIAN_ORDER, rk[4*m_rounds+3]); rk[4*m_rounds+3] = temp;
	}

#if CRYPTOPP_BOOL_AESNI_INTRINSICS_AVAILABLE
	if (HasAESNI())
		ConditionalByteReverse(BIG_ENDIAN_ORDER, rk+4, rk+4, (m_rounds-1)*16);
#endif
}

void Rijndael::Enc::ProcessAndXorBlock(const byte *inBlock, const byte *xorBlock, byte *outBlock) const
{
#if CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE || defined(CRYPTOPP_X64_MASM_AVAILABLE) || CRYPTOPP_BOOL_AESNI_INTRINSICS_AVAILABLE
#if CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE || defined(CRYPTOPP_X64_MASM_AVAILABLE)
	if (HasSSE2())
#else
	if (HasAESNI())
#endif
	{
		Rijndael::Enc::AdvancedProcessBlocks(inBlock, xorBlock, outBlock, 16, 0);
		return;
	}
#endif

	typedef BlockGetAndPut<word32, NativeByteOrder> Block;

	word32 s0, s1, s2, s3, t0, t1, t2, t3;
	Block::Get(inBlock)(s0)(s1)(s2)(s3);

	const word32 *rk = m_key;
	s0 ^= rk[0];
	s1 ^= rk[1];
	s2 ^= rk[2];
	s3 ^= rk[3];
	t0 = rk[4];
	t1 = rk[5];
	t2 = rk[6];
	t3 = rk[7];
	rk += 8;

	// timing attack countermeasure. see comments at top for more details
	const int cacheLineSize = GetCacheLineSize();
	unsigned int i;
	word32 u = 0;
#ifdef CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS
	for (i=0; i<2048; i+=cacheLineSize)
#else
	for (i=0; i<1024; i+=cacheLineSize)
#endif
		u &= *(const word32 *)(((const byte *)Te)+i);
	u &= Te[255];
	s0 |= u; s1 |= u; s2 |= u; s3 |= u;

	QUARTER_ROUND_FE(s3, t0, t1, t2, t3)
	QUARTER_ROUND_FE(s2, t3, t0, t1, t2)
	QUARTER_ROUND_FE(s1, t2, t3, t0, t1)
	QUARTER_ROUND_FE(s0, t1, t2, t3, t0)

	// Nr - 2 full rounds:
    unsigned int r = m_rounds/2 - 1;
    do
	{
		s0 = rk[0]; s1 = rk[1]; s2 = rk[2]; s3 = rk[3];

		QUARTER_ROUND_E(t3, s0, s1, s2, s3)
		QUARTER_ROUND_E(t2, s3, s0, s1, s2)
		QUARTER_ROUND_E(t1, s2, s3, s0, s1)
		QUARTER_ROUND_E(t0, s1, s2, s3, s0)

		t0 = rk[4]; t1 = rk[5]; t2 = rk[6]; t3 = rk[7];

		QUARTER_ROUND_E(s3, t0, t1, t2, t3)
		QUARTER_ROUND_E(s2, t3, t0, t1, t2)
		QUARTER_ROUND_E(s1, t2, t3, t0, t1)
		QUARTER_ROUND_E(s0, t1, t2, t3, t0)

        rk += 8;
    } while (--r);

	word32 tbw[4];
	byte *const tempBlock = (byte *)tbw;

	QUARTER_ROUND_LE(t2, 15, 2, 5, 8)
	QUARTER_ROUND_LE(t1, 11, 14, 1, 4)
	QUARTER_ROUND_LE(t0, 7, 10, 13, 0)
	QUARTER_ROUND_LE(t3, 3, 6, 9, 12)

	Block::Put(xorBlock, outBlock)(tbw[0]^rk[0])(tbw[1]^rk[1])(tbw[2]^rk[2])(tbw[3]^rk[3]);
}

void Rijndael::Dec::ProcessAndXorBlock(const byte *inBlock, const byte *xorBlock, byte *outBlock) const
{
#if CRYPTOPP_BOOL_AESNI_INTRINSICS_AVAILABLE
	if (HasAESNI())
	{
		Rijndael::Dec::AdvancedProcessBlocks(inBlock, xorBlock, outBlock, 16, 0);
		return;
	}
#endif

	typedef BlockGetAndPut<word32, NativeByteOrder> Block;

	word32 s0, s1, s2, s3, t0, t1, t2, t3;
	Block::Get(inBlock)(s0)(s1)(s2)(s3);

	const word32 *rk = m_key;
	s0 ^= rk[0];
	s1 ^= rk[1];
	s2 ^= rk[2];
	s3 ^= rk[3];
	t0 = rk[4];
	t1 = rk[5];
	t2 = rk[6];
	t3 = rk[7];
	rk += 8;

	// timing attack countermeasure. see comments at top for more details
	const int cacheLineSize = GetCacheLineSize();
	unsigned int i;
	word32 u = 0;
#ifdef CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS
	for (i=0; i<2048; i+=cacheLineSize)
#else
	for (i=0; i<1024; i+=cacheLineSize)
#endif
		u &= *(const word32 *)(((const byte *)Td)+i);
	u &= Td[255];
	s0 |= u; s1 |= u; s2 |= u; s3 |= u;

	QUARTER_ROUND_FD(s3, t2, t1, t0, t3)
	QUARTER_ROUND_FD(s2, t1, t0, t3, t2)
	QUARTER_ROUND_FD(s1, t0, t3, t2, t1)
	QUARTER_ROUND_FD(s0, t3, t2, t1, t0)

	// Nr - 2 full rounds:
    unsigned int r = m_rounds/2 - 1;
    do
	{
		s0 = rk[0]; s1 = rk[1]; s2 = rk[2]; s3 = rk[3];

		QUARTER_ROUND_D(t3, s2, s1, s0, s3)
		QUARTER_ROUND_D(t2, s1, s0, s3, s2)
		QUARTER_ROUND_D(t1, s0, s3, s2, s1)
		QUARTER_ROUND_D(t0, s3, s2, s1, s0)

		t0 = rk[4]; t1 = rk[5]; t2 = rk[6]; t3 = rk[7];

		QUARTER_ROUND_D(s3, t2, t1, t0, t3)
		QUARTER_ROUND_D(s2, t1, t0, t3, t2)
		QUARTER_ROUND_D(s1, t0, t3, t2, t1)
		QUARTER_ROUND_D(s0, t3, t2, t1, t0)

        rk += 8;
    } while (--r);

#ifndef CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS
	// timing attack countermeasure. see comments at top for more details
	// If CRYPTOPP_ALLOW_UNALIGNED_DATA_ACCESS is defined, 
	// QUARTER_ROUND_LD will use Td, which is already preloaded.
	u = 0;
	for (i=0; i<256; i+=cacheLineSize)
		u &= *(const word32 *)(Sd+i);
	u &= *(const word32 *)(Sd+252);
	t0 |= u; t1 |= u; t2 |= u; t3 |= u;
#endif

	word32 tbw[4];
	byte *const tempBlock = (byte *)tbw;

	QUARTER_ROUND_LD(t2, 7, 2, 13, 8)
	QUARTER_ROUND_LD(t1, 3, 14, 9, 4)
	QUARTER_ROUND_LD(t0, 15, 10, 5, 0)
	QUARTER_ROUND_LD(t3, 11, 6, 1, 12)

	Block::Put(xorBlock, outBlock)(tbw[0]^rk[0])(tbw[1]^rk[1])(tbw[2]^rk[2])(tbw[3]^rk[3]);
}

// ************************* Assembly Code ************************************

#pragma warning(disable: 4731)	// frame pointer register 'ebp' modified by inline assembly code

#endif	// #ifndef CRYPTOPP_GENERATE_X64_MASM

#if CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE

CRYPTOPP_NAKED void CRYPTOPP_FASTCALL Rijndael_Enc_AdvancedProcessBlocks(void *locals, const word32 *k)
{
#if CRYPTOPP_BOOL_X86

#define L_REG			esp
#define L_INDEX(i)		(L_REG+768+i)
#define L_INXORBLOCKS	L_INBLOCKS+4
#define L_OUTXORBLOCKS	L_INBLOCKS+8
#define L_OUTBLOCKS		L_INBLOCKS+12
#define L_INCREMENTS	L_INDEX(16*15)
#define L_SP			L_INDEX(16*16)
#define L_LENGTH		L_INDEX(16*16+4)
#define L_KEYS_BEGIN	L_INDEX(16*16+8)

#define MOVD			movd
#define MM(i)			mm##i

#define MXOR(a,b,c)	\
	AS2(	movzx	esi, b)\
	AS2(	movd	mm7, DWORD PTR [AS_REG_7+8*WORD_REG(si)+MAP0TO4(c)])\
	AS2(	pxor	MM(a), mm7)\

#define MMOV(a,b,c)	\
	AS2(	movzx	esi, b)\
	AS2(	movd	MM(a), DWORD PTR [AS_REG_7+8*WORD_REG(si)+MAP0TO4(c)])\

#else

#define L_REG			r8
#define L_INDEX(i)		(L_REG+i)
#define L_INXORBLOCKS	L_INBLOCKS+8
#define L_OUTXORBLOCKS	L_INBLOCKS+16
#define L_OUTBLOCKS		L_INBLOCKS+24
#define L_INCREMENTS	L_INDEX(16*16)
#define L_LENGTH		L_INDEX(16*18+8)
#define L_KEYS_BEGIN	L_INDEX(16*19)

#define MOVD			mov
#define MM_0			r9d
#define MM_1			r12d
#ifdef __GNUC__
#define MM_2			r11d
#else
#define MM_2			r10d
#endif
#define MM(i)			MM_##i

#define MXOR(a,b,c)	\
	AS2(	movzx	esi, b)\
	AS2(	xor		MM(a), DWORD PTR [AS_REG_7+8*WORD_REG(si)+MAP0TO4(c)])\

#define MMOV(a,b,c)	\
	AS2(	movzx	esi, b)\
	AS2(	mov		MM(a), DWORD PTR [AS_REG_7+8*WORD_REG(si)+MAP0TO4(c)])\

#endif

#define L_SUBKEYS		L_INDEX(0)
#define L_SAVED_X		L_SUBKEYS
#define L_KEY12			L_INDEX(16*12)
#define L_LASTROUND		L_INDEX(16*13)
#define L_INBLOCKS		L_INDEX(16*14)
#define MAP0TO4(i)		(ASM_MOD(i+3,4)+1)

#define XOR(a,b,c)	\
	AS2(	movzx	esi, b)\
	AS2(	xor		a, DWORD PTR [AS_REG_7+8*WORD_REG(si)+MAP0TO4(c)])\

#define MOV(a,b,c)	\
	AS2(	movzx	esi, b)\
	AS2(	mov		a, DWORD PTR [AS_REG_7+8*WORD_REG(si)+MAP0TO4(c)])\

#ifdef CRYPTOPP_GENERATE_X64_MASM
		ALIGN   8
	Rijndael_Enc_AdvancedProcessBlocks	PROC FRAME
		rex_push_reg rsi
		push_reg rdi
		push_reg rbx
		push_reg r12
		.endprolog
		mov L_REG, rcx
		mov AS_REG_7, ?Te@rdtable@CryptoPP@@3PA_KA
		mov edi, DWORD PTR [?g_cacheLineSize@CryptoPP@@3IA]
#elif defined(__GNUC__)
	__asm__ __volatile__
	(
	".intel_syntax noprefix;"
	#if CRYPTOPP_BOOL_X64
	AS2(	mov		L_REG, rcx)
	#endif
	AS_PUSH_IF86(bx)
	AS_PUSH_IF86(bp)
	AS2(	mov		AS_REG_7, WORD_REG(si))
#else
	AS_PUSH_IF86(si)
	AS_PUSH_IF86(di)
	AS_PUSH_IF86(bx)
	AS_PUSH_IF86(bp)
	AS2(	lea		AS_REG_7, [Te])
	AS2(	mov		edi, [g_cacheLineSize])
#endif

#if CRYPTOPP_BOOL_X86
	AS2(	mov		[ecx+16*12+16*4], esp)	// save esp to L_SP
	AS2(	lea		esp, [ecx-768])
#endif

	// copy subkeys to stack
	AS2(	mov		WORD_REG(si), [L_KEYS_BEGIN])
	AS2(	mov		WORD_REG(ax), 16)
	AS2(	and		WORD_REG(ax), WORD_REG(si))
	AS2(	movdqa	xmm3, XMMWORD_PTR [WORD_REG(dx)+16+WORD_REG(ax)])	// subkey 1 (non-counter) or 2 (counter)
	AS2(	movdqa	[L_KEY12], xmm3)
	AS2(	lea		WORD_REG(ax), [WORD_REG(dx)+WORD_REG(ax)+2*16])
	AS2(	sub		WORD_REG(ax), WORD_REG(si))
	ASL(0)
	AS2(	movdqa	xmm0, [WORD_REG(ax)+WORD_REG(si)])
	AS2(	movdqa	XMMWORD_PTR [L_SUBKEYS+WORD_REG(si)], xmm0)
	AS2(	add		WORD_REG(si), 16)
	AS2(	cmp		WORD_REG(si), 16*12)
	ASJ(	jl,		0, b)

	// read subkeys 0, 1 and last
	AS2(	movdqa	xmm4, [WORD_REG(ax)+WORD_REG(si)])	// last subkey
	AS2(	movdqa	xmm1, [WORD_REG(dx)])			// subkey 0
	AS2(	MOVD	MM(1), [WORD_REG(dx)+4*4])		// 0,1,2,3
	AS2(	mov		ebx, [WORD_REG(dx)+5*4])		// 4,5,6,7
	AS2(	mov		ecx, [WORD_REG(dx)+6*4])		// 8,9,10,11
	AS2(	mov		edx, [WORD_REG(dx)+7*4])		// 12,13,14,15

	// load table into cache
	AS2(	xor		WORD_REG(ax), WORD_REG(ax))
	ASL(9)
	AS2(	mov		esi, [AS_REG_7+WORD_REG(ax)])
	AS2(	add		WORD_REG(ax), WORD_REG(di))
	AS2(	mov		esi, [AS_REG_7+WORD_REG(ax)])
	AS2(	add		WORD_REG(ax), WORD_REG(di))
	AS2(	mov		esi, [AS_REG_7+WORD_REG(ax)])
	AS2(	add		WORD_REG(ax), WORD_REG(di))
	AS2(	mov		esi, [AS_REG_7+WORD_REG(ax)])
	AS2(	add		WORD_REG(ax), WORD_REG(di))
	AS2(	cmp		WORD_REG(ax), 2048)
	ASJ(	jl,		9, b)
	AS1(	lfence)

	AS2(	test	DWORD PTR [L_LENGTH], 1)
	ASJ(	jz,		8, f)

	// counter mode one-time setup
	AS2(	mov		WORD_REG(si), [L_INBLOCKS])
	AS2(	movdqu	xmm2, [WORD_REG(si)])	// counter
	AS2(	pxor	xmm2, xmm1)
	AS2(	psrldq	xmm1, 14)
	AS2(	movd	eax, xmm1)
	AS2(	mov		al, BYTE PTR [WORD_REG(si)+15])
	AS2(	MOVD	MM(2), eax)
#if CRYPTOPP_BOOL_X86
	AS2(	mov		eax, 1)
	AS2(	movd	mm3, eax)
#endif

	// partial first round, in: xmm2(15,14,13,12;11,10,9,8;7,6,5,4;3,2,1,0), out: mm1, ebx, ecx, edx
	AS2(	movd	eax, xmm2)
	AS2(	psrldq	xmm2, 4)
	AS2(	movd	edi, xmm2)
	AS2(	psrldq	xmm2, 4)
		MXOR(		1, al, 0)		// 0
		XOR(		edx, ah, 1)		// 1
	AS2(	shr		eax, 16)
		XOR(		ecx, al, 2)		// 2
		XOR(		ebx, ah, 3)		// 3
	AS2(	mov		eax, edi)
	AS2(	movd	edi, xmm2)
	AS2(	psrldq	xmm2, 4)
		XOR(		ebx, al, 0)		// 4
		MXOR(		1, ah, 1)		// 5
	AS2(	shr		eax, 16)
		XOR(		edx, al, 2)		// 6
		XOR(		ecx, ah, 3)		// 7
	AS2(	mov		eax, edi)
	AS2(	movd	edi, xmm2)
		XOR(		ecx, al, 0)		// 8
		XOR(		ebx, ah, 1)		// 9
	AS2(	shr		eax, 16)
		MXOR(		1, al, 2)		// 10
		XOR(		edx, ah, 3)		// 11
	AS2(	mov		eax, edi)
		XOR(		edx, al, 0)		// 12
		XOR(		ecx, ah, 1)		// 13
	AS2(	shr		eax, 16)
		XOR(		ebx, al, 2)		// 14
	AS2(	psrldq	xmm2, 3)

	// partial second round, in: ebx(4,5,6,7), ecx(8,9,10,11), edx(12,13,14,15), out: eax, ebx, edi, mm0
	AS2(	mov		eax, [L_KEY12+0*4])
	AS2(	mov		edi, [L_KEY12+2*4])
	AS2(	MOVD	MM(0), [L_KEY12+3*4])
		MXOR(	0, cl, 3)	/* 11 */
		XOR(	edi, bl, 3)	/* 7 */
		MXOR(	0, bh, 2)	/* 6 */
	AS2(	shr ebx, 16)	/* 4,5 */
		XOR(	eax, bl, 1)	/* 5 */
		MOV(	ebx, bh, 0)	/* 4 */
	AS2(	xor		ebx, [L_KEY12+1*4])
		XOR(	eax, ch, 2)	/* 10 */
	AS2(	shr ecx, 16)	/* 8,9 */
		XOR(	eax, dl, 3)	/* 15 */
		XOR(	ebx, dh, 2)	/* 14 */
	AS2(	shr edx, 16)	/* 12,13 */
		XOR(	edi, ch, 0)	/* 8 */
		XOR(	ebx, cl, 1)	/* 9 */
		XOR(	edi, dl, 1)	/* 13 */
		MXOR(	0, dh, 0)	/* 12 */

	AS2(	movd	ecx, xmm2)
	AS2(	MOVD	edx, MM(1))
	AS2(	MOVD	[L_SAVED_X+3*4], MM(0))
	AS2(	mov		[L_SAVED_X+0*4], eax)
	AS2(	mov		[L_SAVED_X+1*4], ebx)
	AS2(	mov		[L_SAVED_X+2*4], edi)
	ASJ(	jmp,	5, f)

	ASL(3)
	// non-counter mode per-block setup
	AS2(	MOVD	MM(1), [L_KEY12+0*4])	// 0,1,2,3
	AS2(	mov		ebx, [L_KEY12+1*4])		// 4,5,6,7
	AS2(	mov		ecx, [L_KEY12+2*4])		// 8,9,10,11
	AS2(	mov		edx, [L_KEY12+3*4])		// 12,13,14,15
	ASL(8)
	AS2(	mov		WORD_REG(ax), [L_INBLOCKS])
	AS2(	movdqu	xmm2, [WORD_REG(ax)])
	AS2(	mov		WORD_REG(si), [L_INXORBLOCKS])
	AS2(	movdqu	xmm5, [WORD_REG(si)])
	AS2(	pxor	xmm2, xmm1)
	AS2(	pxor	xmm2, xmm5)

	// first round, in: xmm2(15,14,13,12;11,10,9,8;7,6,5,4;3,2,1,0), out: eax, ebx, ecx, edx
	AS2(	movd	eax, xmm2)
	AS2(	psrldq	xmm2, 4)
	AS2(	movd	edi, xmm2)
	AS2(	psrldq	xmm2, 4)
		MXOR(		1, al, 0)		// 0
		XOR(		edx, ah, 1)		// 1
	AS2(	shr		eax, 16)
		XOR(		ecx, al, 2)		// 2
		XOR(		ebx, ah, 3)		// 3
	AS2(	mov		eax, edi)
	AS2(	movd	edi, xmm2)
	AS2(	psrldq	xmm2, 4)
		XOR(		ebx, al, 0)		// 4
		MXOR(		1, ah, 1)		// 5
	AS2(	shr		eax, 16)
		XOR(		edx, al, 2)		// 6
		XOR(		ecx, ah, 3)		// 7
	AS2(	mov		eax, edi)
	AS2(	movd	edi, xmm2)
		XOR(		ecx, al, 0)		// 8
		XOR(		ebx, ah, 1)		// 9
	AS2(	shr		eax, 16)
		MXOR(		1, al, 2)		// 10
		XOR(		edx, ah, 3)		// 11
	AS2(	mov		eax, edi)
		XOR(		edx, al, 0)		// 12
		XOR(		ecx, ah, 1)		// 13
	AS2(	shr		eax, 16)
		XOR(		ebx, al, 2)		// 14
		MXOR(		1, ah, 3)		// 15
	AS2(	MOVD	eax, MM(1))

	AS2(	add		L_REG, [L_KEYS_BEGIN])
	AS2(	add		L_REG, 4*16)
	ASJ(	jmp,	2, f)

	ASL(1)
	// counter-mode per-block setup
	AS2(	MOVD	ecx, MM(2))
	AS2(	MOVD	edx, MM(1))
	AS2(	mov		eax, [L_SAVED_X+0*4])
	AS2(	mov		ebx, [L_SAVED_X+1*4])
	AS2(	xor		cl, ch)
	AS2(	and		WORD_REG(cx), 255)
	ASL(5)
#if CRYPTOPP_BOOL_X86
	AS2(	paddb	MM(2), mm3)
#else
	AS2(	add		MM(2), 1)
#endif
	// remaining part of second round, in: edx(previous round),esi(keyed counter byte) eax,ebx,[L_SAVED_X+2*4],[L_SAVED_X+3*4], out: eax,ebx,ecx,edx
	AS2(	xor		edx, DWORD PTR [AS_REG_7+WORD_REG(cx)*8+3])
		XOR(		ebx, dl, 3)
		MOV(		ecx, dh, 2)
	AS2(	shr		edx, 16)
	AS2(	xor		ecx, [L_SAVED_X+2*4])
		XOR(		eax, dh, 0)
		MOV(		edx, dl, 1)
	AS2(	xor		edx, [L_SAVED_X+3*4])

	AS2(	add		L_REG, [L_KEYS_BEGIN])
	AS2(	add		L_REG, 3*16)
	ASJ(	jmp,	4, f)

// in: eax(0,1,2,3), ebx(4,5,6,7), ecx(8,9,10,11), edx(12,13,14,15)
// out: eax, ebx, edi, mm0
#define ROUND()		\
		MXOR(	0, cl, 3)	/* 11 */\
	AS2(	mov	cl, al)		/* 8,9,10,3 */\
		XOR(	edi, ah, 2)	/* 2 */\
	AS2(	shr eax, 16)	/* 0,1 */\
		XOR(	edi, bl, 3)	/* 7 */\
		MXOR(	0, bh, 2)	/* 6 */\
	AS2(	shr ebx, 16)	/* 4,5 */\
		MXOR(	0, al, 1)	/* 1 */\
		MOV(	eax, ah, 0)	/* 0 */\
		XOR(	eax, bl, 1)	/* 5 */\
		MOV(	ebx, bh, 0)	/* 4 */\
		XOR(	eax, ch, 2)	/* 10 */\
		XOR(	ebx, cl, 3)	/* 3 */\
	AS2(	shr ecx, 16)	/* 8,9 */\
		XOR(	eax, dl, 3)	/* 15 */\
		XOR(	ebx, dh, 2)	/* 14 */\
	AS2(	shr edx, 16)	/* 12,13 */\
		XOR(	edi, ch, 0)	/* 8 */\
		XOR(	ebx, cl, 1)	/* 9 */\
		XOR(	edi, dl, 1)	/* 13 */\
		MXOR(	0, dh, 0)	/* 12 */\

	ASL(2)	// 2-round loop
	AS2(	MOVD	MM(0), [L_SUBKEYS-4*16+3*4])
	AS2(	mov		edi, [L_SUBKEYS-4*16+2*4])
	ROUND()
	AS2(	mov		ecx, edi)
	AS2(	xor		eax, [L_SUBKEYS-4*16+0*4])
	AS2(	xor		ebx, [L_SUBKEYS-4*16+1*4])
	AS2(	MOVD	edx, MM(0))

	ASL(4)
	AS2(	MOVD	MM(0), [L_SUBKEYS-4*16+7*4])
	AS2(	mov		edi, [L_SUBKEYS-4*16+6*4])
	ROUND()
	AS2(	mov		ecx, edi)
	AS2(	xor		eax, [L_SUBKEYS-4*16+4*4])
	AS2(	xor		ebx, [L_SUBKEYS-4*16+5*4])
	AS2(	MOVD	edx, MM(0))

	AS2(	add		L_REG, 32)
	AS2(	test	L_REG, 255)
	ASJ(	jnz,	2, b)
	AS2(	sub		L_REG, 16*16)

#define LAST(a, b, c)												\
	AS2(	movzx	esi, a											)\
	AS2(	movzx	edi, BYTE PTR [AS_REG_7+WORD_REG(si)*8+1]	)\
	AS2(	movzx	esi, b											)\
	AS2(	xor		edi, DWORD PTR [AS_REG_7+WORD_REG(si)*8+0]	)\
	AS2(	mov		WORD PTR [L_LASTROUND+c], di					)\

	// last round
	LAST(ch, dl, 2)
	LAST(dh, al, 6)
	AS2(	shr		edx, 16)
	LAST(ah, bl, 10)
	AS2(	shr		eax, 16)
	LAST(bh, cl, 14)
	AS2(	shr		ebx, 16)
	LAST(dh, al, 12)
	AS2(	shr		ecx, 16)
	LAST(ah, bl, 0)
	LAST(bh, cl, 4)
	LAST(ch, dl, 8)

	AS2(	mov		WORD_REG(ax), [L_OUTXORBLOCKS])
	AS2(	mov		WORD_REG(bx), [L_OUTBLOCKS])

	AS2(	mov		WORD_REG(cx), [L_LENGTH])
	AS2(	sub		WORD_REG(cx), 16)

	AS2(	movdqu	xmm2, [WORD_REG(ax)])
	AS2(	pxor	xmm2, xmm4)

#if CRYPTOPP_BOOL_X86
	AS2(	movdqa	xmm0, [L_INCREMENTS])
	AS2(	paddd	xmm0, [L_INBLOCKS])
	AS2(	movdqa	[L_INBLOCKS], xmm0)
#else
	AS2(	movdqa	xmm0, [L_INCREMENTS+16])
	AS2(	paddq	xmm0, [L_INBLOCKS+16])
	AS2(	movdqa	[L_INBLOCKS+16], xmm0)
#endif

	AS2(	pxor	xmm2, [L_LASTROUND])
	AS2(	movdqu	[WORD_REG(bx)], xmm2)

	ASJ(	jle,	7, f)
	AS2(	mov		[L_LENGTH], WORD_REG(cx))
	AS2(	test	WORD_REG(cx), 1)
	ASJ(	jnz,	1, b)
#if CRYPTOPP_BOOL_X64
	AS2(	movdqa	xmm0, [L_INCREMENTS])
	AS2(	paddq	xmm0, [L_INBLOCKS])
	AS2(	movdqa	[L_INBLOCKS], xmm0)
#endif
	ASJ(	jmp,	3, b)

	ASL(7)
	// erase keys on stack
	AS2(	xorps	xmm0, xmm0)
	AS2(	lea		WORD_REG(ax), [L_SUBKEYS+7*16])
	AS2(	movaps	[WORD_REG(ax)-7*16], xmm0)
	AS2(	movaps	[WORD_REG(ax)-6*16], xmm0)
	AS2(	movaps	[WORD_REG(ax)-5*16], xmm0)
	AS2(	movaps	[WORD_REG(ax)-4*16], xmm0)
	AS2(	movaps	[WORD_REG(ax)-3*16], xmm0)
	AS2(	movaps	[WORD_REG(ax)-2*16], xmm0)
	AS2(	movaps	[WORD_REG(ax)-1*16], xmm0)
	AS2(	movaps	[WORD_REG(ax)+0*16], xmm0)
	AS2(	movaps	[WORD_REG(ax)+1*16], xmm0)
	AS2(	movaps	[WORD_REG(ax)+2*16], xmm0)
	AS2(	movaps	[WORD_REG(ax)+3*16], xmm0)
	AS2(	movaps	[WORD_REG(ax)+4*16], xmm0)
	AS2(	movaps	[WORD_REG(ax)+5*16], xmm0)
	AS2(	movaps	[WORD_REG(ax)+6*16], xmm0)
#if CRYPTOPP_BOOL_X86
	AS2(	mov		esp, [L_SP])
	AS1(	emms)
#endif
	AS_POP_IF86(bp)
	AS_POP_IF86(bx)
#if defined(_MSC_VER) && CRYPTOPP_BOOL_X86
	AS_POP_IF86(di)
	AS_POP_IF86(si)
	AS1(ret)
#endif
#ifdef CRYPTOPP_GENERATE_X64_MASM
	pop r12
	pop rbx
	pop rdi
	pop rsi
	ret
	Rijndael_Enc_AdvancedProcessBlocks ENDP
#endif
#ifdef __GNUC__
	".att_syntax prefix;"
	: 
	: "c" (locals), "d" (k), "S" (Te), "D" (g_cacheLineSize)
	: "memory", "cc", "%eax"
	#if CRYPTOPP_BOOL_X64
		, "%rbx", "%r8", "%r9", "%r10", "%r11", "%r12"
	#endif
	);
#endif
}

#endif

#ifndef CRYPTOPP_GENERATE_X64_MASM

#ifdef CRYPTOPP_X64_MASM_AVAILABLE
extern "C" {
void Rijndael_Enc_AdvancedProcessBlocks(void *locals, const word32 *k);
}
#endif

#if CRYPTOPP_BOOL_X64 || CRYPTOPP_BOOL_X86

static inline bool AliasedWithTable(const byte *begin, const byte *end)
{
	size_t s0 = size_t(begin)%4096, s1 = size_t(end)%4096;
	size_t t0 = size_t(Te)%4096, t1 = (size_t(Te)+sizeof(Te))%4096;
	if (t1 > t0)
		return (s0 >= t0 && s0 < t1) || (s1 > t0 && s1 <= t1);
	else
		return (s0 < t1 || s1 <= t1) || (s0 >= t0 || s1 > t0);
}

#if CRYPTOPP_BOOL_AESNI_INTRINSICS_AVAILABLE

inline void AESNI_Enc_Block(__m128i &block, const __m128i *subkeys, unsigned int rounds)
{
	block = _mm_xor_si128(block, subkeys[0]);
	for (unsigned int i=1; i<rounds-1; i+=2)
	{
		block = _mm_aesenc_si128(block, subkeys[i]);
		block = _mm_aesenc_si128(block, subkeys[i+1]);
	}
	block = _mm_aesenc_si128(block, subkeys[rounds-1]);
	block = _mm_aesenclast_si128(block, subkeys[rounds]);
}

inline void AESNI_Enc_4_Blocks(__m128i &block0, __m128i &block1, __m128i &block2, __m128i &block3, const __m128i *subkeys, unsigned int rounds)
{
	__m128i rk = subkeys[0];
	block0 = _mm_xor_si128(block0, rk);
	block1 = _mm_xor_si128(block1, rk);
	block2 = _mm_xor_si128(block2, rk);
	block3 = _mm_xor_si128(block3, rk);
	for (unsigned int i=1; i<rounds; i++)
	{
		rk = subkeys[i];
		block0 = _mm_aesenc_si128(block0, rk);
		block1 = _mm_aesenc_si128(block1, rk);
		block2 = _mm_aesenc_si128(block2, rk);
		block3 = _mm_aesenc_si128(block3, rk);
	}
	rk = subkeys[rounds];
	block0 = _mm_aesenclast_si128(block0, rk);
	block1 = _mm_aesenclast_si128(block1, rk);
	block2 = _mm_aesenclast_si128(block2, rk);
	block3 = _mm_aesenclast_si128(block3, rk);
}

inline void AESNI_Dec_Block(__m128i &block, const __m128i *subkeys, unsigned int rounds)
{
	block = _mm_xor_si128(block, subkeys[0]);
	for (unsigned int i=1; i<rounds-1; i+=2)
	{
		block = _mm_aesdec_si128(block, subkeys[i]);
		block = _mm_aesdec_si128(block, subkeys[i+1]);
	}
	block = _mm_aesdec_si128(block, subkeys[rounds-1]);
	block = _mm_aesdeclast_si128(block, subkeys[rounds]);
}

inline void AESNI_Dec_4_Blocks(__m128i &block0, __m128i &block1, __m128i &block2, __m128i &block3, const __m128i *subkeys, unsigned int rounds)
{
	__m128i rk = subkeys[0];
	block0 = _mm_xor_si128(block0, rk);
	block1 = _mm_xor_si128(block1, rk);
	block2 = _mm_xor_si128(block2, rk);
	block3 = _mm_xor_si128(block3, rk);
	for (unsigned int i=1; i<rounds; i++)
	{
		rk = subkeys[i];
		block0 = _mm_aesdec_si128(block0, rk);
		block1 = _mm_aesdec_si128(block1, rk);
		block2 = _mm_aesdec_si128(block2, rk);
		block3 = _mm_aesdec_si128(block3, rk);
	}
	rk = subkeys[rounds];
	block0 = _mm_aesdeclast_si128(block0, rk);
	block1 = _mm_aesdeclast_si128(block1, rk);
	block2 = _mm_aesdeclast_si128(block2, rk);
	block3 = _mm_aesdeclast_si128(block3, rk);
}

static CRYPTOPP_ALIGN_DATA(16) const word32 s_one[] = {0, 0, 0, 1<<24};

template <typename F1, typename F4>
inline size_t AESNI_AdvancedProcessBlocks(F1 func1, F4 func4, const __m128i *subkeys, unsigned int rounds, const byte *inBlocks, const byte *xorBlocks, byte *outBlocks, size_t length, word32 flags)
{
	size_t blockSize = 16;
	size_t inIncrement = (flags & (BlockTransformation::BT_InBlockIsCounter|BlockTransformation::BT_DontIncrementInOutPointers)) ? 0 : blockSize;
	size_t xorIncrement = xorBlocks ? blockSize : 0;
	size_t outIncrement = (flags & BlockTransformation::BT_DontIncrementInOutPointers) ? 0 : blockSize;

	if (flags & BlockTransformation::BT_ReverseDirection)
	{
		assert(length % blockSize == 0);
		inBlocks += length - blockSize;
		xorBlocks += length - blockSize;
		outBlocks += length - blockSize;
		inIncrement = 0-inIncrement;
		xorIncrement = 0-xorIncrement;
		outIncrement = 0-outIncrement;
	}

	if (flags & BlockTransformation::BT_AllowParallel)
	{
		while (length >= 4*blockSize)
		{
			__m128i block0 = _mm_loadu_si128((const __m128i *)inBlocks), block1, block2, block3;
			if (flags & BlockTransformation::BT_InBlockIsCounter)
			{
				const __m128i be1 = *(const __m128i *)s_one;
				block1 = _mm_add_epi32(block0, be1);
				block2 = _mm_add_epi32(block1, be1);
				block3 = _mm_add_epi32(block2, be1);
				_mm_storeu_si128((__m128i *)inBlocks, _mm_add_epi32(block3, be1));
			}
			else
			{
				inBlocks += inIncrement;
				block1 = _mm_loadu_si128((const __m128i *)inBlocks);
				inBlocks += inIncrement;
				block2 = _mm_loadu_si128((const __m128i *)inBlocks);
				inBlocks += inIncrement;
				block3 = _mm_loadu_si128((const __m128i *)inBlocks);
				inBlocks += inIncrement;
			}

			if (flags & BlockTransformation::BT_XorInput)
			{
				block0 = _mm_xor_si128(block0, _mm_loadu_si128((const __m128i *)xorBlocks));
				xorBlocks += xorIncrement;
				block1 = _mm_xor_si128(block1, _mm_loadu_si128((const __m128i *)xorBlocks));
				xorBlocks += xorIncrement;
				block2 = _mm_xor_si128(block2, _mm_loadu_si128((const __m128i *)xorBlocks));
				xorBlocks += xorIncrement;
				block3 = _mm_xor_si128(block3, _mm_loadu_si128((const __m128i *)xorBlocks));
				xorBlocks += xorIncrement;
			}

			func4(block0, block1, block2, block3, subkeys, rounds);

			if (xorBlocks && !(flags & BlockTransformation::BT_XorInput))
			{
				block0 = _mm_xor_si128(block0, _mm_loadu_si128((const __m128i *)xorBlocks));
				xorBlocks += xorIncrement;
				block1 = _mm_xor_si128(block1, _mm_loadu_si128((const __m128i *)xorBlocks));
				xorBlocks += xorIncrement;
				block2 = _mm_xor_si128(block2, _mm_loadu_si128((const __m128i *)xorBlocks));
				xorBlocks += xorIncrement;
				block3 = _mm_xor_si128(block3, _mm_loadu_si128((const __m128i *)xorBlocks));
				xorBlocks += xorIncrement;
			}

			_mm_storeu_si128((__m128i *)outBlocks, block0);
			outBlocks += outIncrement;
			_mm_storeu_si128((__m128i *)outBlocks, block1);
			outBlocks += outIncrement;
			_mm_storeu_si128((__m128i *)outBlocks, block2);
			outBlocks += outIncrement;
			_mm_storeu_si128((__m128i *)outBlocks, block3);
			outBlocks += outIncrement;

			length -= 4*blockSize;
		}
	}

	while (length >= blockSize)
	{
		__m128i block = _mm_loadu_si128((const __m128i *)inBlocks);

		if (flags & BlockTransformation::BT_XorInput)
			block = _mm_xor_si128(block, _mm_loadu_si128((const __m128i *)xorBlocks));

		if (flags & BlockTransformation::BT_InBlockIsCounter)
			const_cast<byte *>(inBlocks)[15]++;

		func1(block, subkeys, rounds);

		if (xorBlocks && !(flags & BlockTransformation::BT_XorInput))
			block = _mm_xor_si128(block, _mm_loadu_si128((const __m128i *)xorBlocks));

		_mm_storeu_si128((__m128i *)outBlocks, block);

		inBlocks += inIncrement;
		outBlocks += outIncrement;
		xorBlocks += xorIncrement;
		length -= blockSize;
	}

	return length;
}
#endif

size_t Rijndael::Enc::AdvancedProcessBlocks(const byte *inBlocks, const byte *xorBlocks, byte *outBlocks, size_t length, word32 flags) const
{
#if CRYPTOPP_BOOL_AESNI_INTRINSICS_AVAILABLE
	if (HasAESNI())
		return AESNI_AdvancedProcessBlocks(AESNI_Enc_Block, AESNI_Enc_4_Blocks, (const __m128i *)m_key.begin(), m_rounds, inBlocks, xorBlocks, outBlocks, length, flags);
#endif

#if CRYPTOPP_BOOL_SSE2_ASM_AVAILABLE || defined(CRYPTOPP_X64_MASM_AVAILABLE)
	if (HasSSE2())
	{
		if (length < BLOCKSIZE)
			return length;

		struct Locals
		{
			word32 subkeys[4*12], workspace[8];
			const byte *inBlocks, *inXorBlocks, *outXorBlocks;
			byte *outBlocks;
			size_t inIncrement, inXorIncrement, outXorIncrement, outIncrement;
			size_t regSpill, lengthAndCounterFlag, keysBegin;
		};

		size_t increment = BLOCKSIZE;
		const byte* zeros = (byte *)(Te+256);
		byte *space;

		do {
			space = (byte *)alloca(255+sizeof(Locals));
			space += (256-(size_t)space%256)%256;
		}
		while (AliasedWithTable(space, space+sizeof(Locals)));

		if (flags & BT_ReverseDirection)
		{
			assert(length % BLOCKSIZE == 0);
			inBlocks += length - BLOCKSIZE;
			xorBlocks += length - BLOCKSIZE;
			outBlocks += length - BLOCKSIZE;
			increment = 0-increment;
		}

		Locals &locals = *(Locals *)space;

		locals.inBlocks = inBlocks;
		locals.inXorBlocks = (flags & BT_XorInput) && xorBlocks ? xorBlocks : zeros;
		locals.outXorBlocks = (flags & BT_XorInput) || !xorBlocks ? zeros : xorBlocks;
		locals.outBlocks = outBlocks;

		locals.inIncrement = (flags & BT_DontIncrementInOutPointers) ? 0 : increment;
		locals.inXorIncrement = (flags & BT_XorInput) && xorBlocks ? increment : 0;
		locals.outXorIncrement = (flags & BT_XorInput) || !xorBlocks ? 0 : increment;
		locals.outIncrement = (flags & BT_DontIncrementInOutPointers) ? 0 : increment;

		locals.lengthAndCounterFlag = length - (length%16) - bool(flags & BT_InBlockIsCounter);
		int keysToCopy = m_rounds - (flags & BT_InBlockIsCounter ? 3 : 2);
		locals.keysBegin = (12-keysToCopy)*16;

		Rijndael_Enc_AdvancedProcessBlocks(&locals, m_key);
		return length % BLOCKSIZE;
	}
#endif

	return BlockTransformation::AdvancedProcessBlocks(inBlocks, xorBlocks, outBlocks, length, flags);
}

#endif

#if CRYPTOPP_BOOL_AESNI_INTRINSICS_AVAILABLE

size_t Rijndael::Dec::AdvancedProcessBlocks(const byte *inBlocks, const byte *xorBlocks, byte *outBlocks, size_t length, word32 flags) const
{
	if (HasAESNI())
		return AESNI_AdvancedProcessBlocks(AESNI_Dec_Block, AESNI_Dec_4_Blocks, (const __m128i *)m_key.begin(), m_rounds, inBlocks, xorBlocks, outBlocks, length, flags);

	return BlockTransformation::AdvancedProcessBlocks(inBlocks, xorBlocks, outBlocks, length, flags);
}

#endif	// #if CRYPTOPP_BOOL_AESNI_INTRINSICS_AVAILABLE

NAMESPACE_END

#endif
#endif