Repository navigation
Expand file tree
/
Copy pathAesOpt.c
More file actions
957 lines (804 loc) · 23.2 KB
/
Copy pathAesOpt.c
File metadata and controls
957 lines (804 loc) · 23.2 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
/* AesOpt.c -- AES optimized code for x86 AES hardware instructions
2024-03-01 : Igor Pavlov : Public domain */
#include "Precomp.h"
#include "Aes.h"
#include "CpuArch.h"
#ifdef MY_CPU_X86_OR_AMD64
#if defined(__INTEL_COMPILER)
#if (__INTEL_COMPILER >= 1110)
#define USE_INTEL_AES
#if (__INTEL_COMPILER >= 1900)
#define USE_INTEL_VAES
#endif
#endif
#elif defined(Z7_CLANG_VERSION) && (Z7_CLANG_VERSION >= 30800) \
|| defined(Z7_GCC_VERSION) && (Z7_GCC_VERSION >= 40400)
#define USE_INTEL_AES
#if !defined(__AES__)
#define ATTRIB_AES __attribute__((__target__("aes")))
#endif
#if defined(__clang__) && (__clang_major__ >= 8) \
|| defined(__GNUC__) && (__GNUC__ >= 8)
#define USE_INTEL_VAES
#if !defined(__AES__) || !defined(__VAES__) || !defined(__AVX__) || !defined(__AVX2__)
#define ATTRIB_VAES __attribute__((__target__("aes,vaes,avx,avx2")))
#endif
#endif
#elif defined(_MSC_VER)
#if (_MSC_VER > 1500) || (_MSC_FULL_VER >= 150030729)
#define USE_INTEL_AES
#if (_MSC_VER >= 1910)
#define USE_INTEL_VAES
#endif
#endif
#ifndef USE_INTEL_AES
#define Z7_USE_AES_HW_STUB
#endif
#ifndef USE_INTEL_VAES
#define Z7_USE_VAES_HW_STUB
#endif
#endif
#ifndef USE_INTEL_AES
// #define Z7_USE_AES_HW_STUB // for debug
#endif
#ifndef USE_INTEL_VAES
// #define Z7_USE_VAES_HW_STUB // for debug
#endif
#ifdef USE_INTEL_AES
#include <wmmintrin.h>
#if !defined(USE_INTEL_VAES) && defined(Z7_USE_VAES_HW_STUB)
#define AES_TYPE_keys UInt32
#define AES_TYPE_data Byte
// #define AES_TYPE_keys __m128i
// #define AES_TYPE_data __m128i
#endif
#ifndef ATTRIB_AES
#define ATTRIB_AES
#endif
#define AES_FUNC_START(name) \
void Z7_FASTCALL name(UInt32 *ivAes, Byte *data8, size_t numBlocks)
// void Z7_FASTCALL name(__m128i *p, __m128i *data, size_t numBlocks)
#define AES_FUNC_START2(name) \
AES_FUNC_START (name); \
ATTRIB_AES \
AES_FUNC_START (name)
#define MM_OP(op, dest, src) dest = op(dest, src);
#define MM_OP_m(op, src) MM_OP(op, m, src)
#define MM_XOR( dest, src) MM_OP(_mm_xor_si128, dest, src)
AES_FUNC_START2 (AesCbc_Encode_HW)
{
__m128i *p = (__m128i *)(void *)ivAes;
__m128i *data = (__m128i *)(void *)data8;
__m128i m = *p;
const __m128i k0 = p[2];
const __m128i k1 = p[3];
const UInt32 numRounds2 = *(const UInt32 *)(p + 1) - 1;
for (; numBlocks != 0; numBlocks--, data++)
{
UInt32 r = numRounds2;
const __m128i *w = p + 4;
__m128i temp = *data;
MM_XOR (temp, k0)
MM_XOR (m, temp)
MM_OP_m (_mm_aesenc_si128, k1)
do
{
MM_OP_m (_mm_aesenc_si128, w[0])
MM_OP_m (_mm_aesenc_si128, w[1])
w += 2;
}
while (--r);
MM_OP_m (_mm_aesenclast_si128, w[0])
*data = m;
}
*p = m;
}
#define WOP_1(op)
#define WOP_2(op) WOP_1 (op) op (m1, 1)
#define WOP_3(op) WOP_2 (op) op (m2, 2)
#define WOP_4(op) WOP_3 (op) op (m3, 3)
#ifdef MY_CPU_AMD64
#define WOP_5(op) WOP_4 (op) op (m4, 4)
#define WOP_6(op) WOP_5 (op) op (m5, 5)
#define WOP_7(op) WOP_6 (op) op (m6, 6)
#define WOP_8(op) WOP_7 (op) op (m7, 7)
#endif
/*
#define WOP_9(op) WOP_8 (op) op (m8, 8);
#define WOP_10(op) WOP_9 (op) op (m9, 9);
#define WOP_11(op) WOP_10(op) op (m10, 10);
#define WOP_12(op) WOP_11(op) op (m11, 11);
#define WOP_13(op) WOP_12(op) op (m12, 12);
#define WOP_14(op) WOP_13(op) op (m13, 13);
*/
#ifdef MY_CPU_AMD64
#define NUM_WAYS 8
#define WOP_M1 WOP_8
#else
#define NUM_WAYS 4
#define WOP_M1 WOP_4
#endif
#define WOP(op) op (m0, 0) WOP_M1(op)
#define DECLARE_VAR(reg, ii) __m128i reg;
#define LOAD_data( reg, ii) reg = data[ii];
#define STORE_data( reg, ii) data[ii] = reg;
#if (NUM_WAYS > 1)
#define XOR_data_M1(reg, ii) MM_XOR (reg, data[ii- 1])
#endif
#define MM_OP_key(op, reg) MM_OP(op, reg, key);
#define AES_DEC( reg, ii) MM_OP_key (_mm_aesdec_si128, reg)
#define AES_DEC_LAST( reg, ii) MM_OP_key (_mm_aesdeclast_si128, reg)
#define AES_ENC( reg, ii) MM_OP_key (_mm_aesenc_si128, reg)
#define AES_ENC_LAST( reg, ii) MM_OP_key (_mm_aesenclast_si128, reg)
#define AES_XOR( reg, ii) MM_OP_key (_mm_xor_si128, reg)
#define CTR_START(reg, ii) MM_OP (_mm_add_epi64, ctr, one) reg = ctr;
#define CTR_END( reg, ii) MM_XOR (data[ii], reg)
#define WOP_KEY(op, n) { \
const __m128i key = w[n]; \
WOP(op); }
#define WIDE_LOOP_START \
dataEnd = data + numBlocks; \
if (numBlocks >= NUM_WAYS) \
{ dataEnd -= NUM_WAYS; do { \
#define WIDE_LOOP_END \
data += NUM_WAYS; \
} while (data <= dataEnd); \
dataEnd += NUM_WAYS; } \
#define SINGLE_LOOP \
for (; data < dataEnd; data++)
#ifdef USE_INTEL_VAES
#define AVX_XOR(dest, src) MM_OP(_mm256_xor_si256, dest, src)
#define AVX_DECLARE_VAR(reg, ii) __m256i reg;
#define AVX_LOAD_data( reg, ii) reg = ((const __m256i *)(const void *)data)[ii];
#define AVX_STORE_data( reg, ii) ((__m256i *)(void *)data)[ii] = reg;
/*
AVX_XOR_data_M1() needs unaligned memory load
if (we don't use _mm256_loadu_si256() here)
{
Most compilers with enabled optimizations generate fused AVX (LOAD + OP)
instruction that can load unaligned data.
But GCC and CLANG without -O2 or -O1 optimizations can generate separated
LOAD-ALIGNED (vmovdqa) instruction that will fail on execution.
}
Note: some compilers generate more instructions, if we use _mm256_loadu_si256() here.
v23.02: we use _mm256_loadu_si256() here, because we need compatibility with any compiler.
*/
#define AVX_XOR_data_M1(reg, ii) AVX_XOR (reg, _mm256_loadu_si256(&(((const __m256i *)(const void *)(data - 1))[ii])))
// for debug only: the following code will fail on execution, if compiled by some compilers:
// #define AVX_XOR_data_M1(reg, ii) AVX_XOR (reg, (((const __m256i *)(const void *)(data - 1))[ii]))
#define AVX_AES_DEC( reg, ii) MM_OP_key (_mm256_aesdec_epi128, reg)
#define AVX_AES_DEC_LAST( reg, ii) MM_OP_key (_mm256_aesdeclast_epi128, reg)
#define AVX_AES_ENC( reg, ii) MM_OP_key (_mm256_aesenc_epi128, reg)
#define AVX_AES_ENC_LAST( reg, ii) MM_OP_key (_mm256_aesenclast_epi128, reg)
#define AVX_AES_XOR( reg, ii) MM_OP_key (_mm256_xor_si256, reg)
#define AVX_CTR_START(reg, ii) MM_OP (_mm256_add_epi64, ctr2, two) reg = _mm256_xor_si256(ctr2, key);
#define AVX_CTR_END( reg, ii) AVX_XOR (((__m256i *)(void *)data)[ii], reg)
#define AVX_WOP_KEY(op, n) { \
const __m256i key = w[n]; \
WOP(op); }
#define NUM_AES_KEYS_MAX 15
#define WIDE_LOOP_START_AVX(OP) \
dataEnd = data + numBlocks; \
if (numBlocks >= NUM_WAYS * 2) \
{ __m256i keys[NUM_AES_KEYS_MAX]; \
UInt32 ii; \
OP \
for (ii = 0; ii < numRounds; ii++) \
keys[ii] = _mm256_broadcastsi128_si256(p[ii]); \
dataEnd -= NUM_WAYS * 2; do { \
#define WIDE_LOOP_END_AVX(OP) \
data += NUM_WAYS * 2; \
} while (data <= dataEnd); \
dataEnd += NUM_WAYS * 2; \
OP \
_mm256_zeroupper(); \
} \
/* MSVC for x86: If we don't call _mm256_zeroupper(), and -arch:IA32 is not specified,
MSVC still can insert vzeroupper instruction. */
#endif
AES_FUNC_START2 (AesCbc_Decode_HW)
{
__m128i *p = (__m128i *)(void *)ivAes;
__m128i *data = (__m128i *)(void *)data8;
__m128i iv = *p;
const __m128i *wStart = p + *(const UInt32 *)(p + 1) * 2 + 2 - 1;
const __m128i *dataEnd;
p += 2;
WIDE_LOOP_START
{
const __m128i *w = wStart;
WOP (DECLARE_VAR)
WOP (LOAD_data)
WOP_KEY (AES_XOR, 1)
do
{
WOP_KEY (AES_DEC, 0)
w--;
}
while (w != p);
WOP_KEY (AES_DEC_LAST, 0)
MM_XOR (m0, iv)
WOP_M1 (XOR_data_M1)
iv = data[NUM_WAYS - 1];
WOP (STORE_data)
}
WIDE_LOOP_END
SINGLE_LOOP
{
const __m128i *w = wStart - 1;
__m128i m = _mm_xor_si128 (w[2], *data);
do
{
MM_OP_m (_mm_aesdec_si128, w[1])
MM_OP_m (_mm_aesdec_si128, w[0])
w -= 2;
}
while (w != p);
MM_OP_m (_mm_aesdec_si128, w[1])
MM_OP_m (_mm_aesdeclast_si128, w[0])
MM_XOR (m, iv)
iv = *data;
*data = m;
}
p[-2] = iv;
}
AES_FUNC_START2 (AesCtr_Code_HW)
{
__m128i *p = (__m128i *)(void *)ivAes;
__m128i *data = (__m128i *)(void *)data8;
__m128i ctr = *p;
UInt32 numRoundsMinus2 = *(const UInt32 *)(p + 1) * 2 - 1;
const __m128i *dataEnd;
__m128i >
p += 2;
WIDE_LOOP_START
{
const __m128i *w = p;
UInt32 r = numRoundsMinus2;
WOP (DECLARE_VAR)
WOP (CTR_START)
WOP_KEY (AES_XOR, 0)
w += 1;
do
{
WOP_KEY (AES_ENC, 0)
w += 1;
}
while (--r);
WOP_KEY (AES_ENC_LAST, 0)
WOP (CTR_END)
}
WIDE_LOOP_END
SINGLE_LOOP
{
UInt32 numRounds2 = *(const UInt32 *)(p - 2 + 1) - 1;
const __m128i *w = p;
__m128i m;
MM_OP (_mm_add_epi64, ctr, one)
m = _mm_xor_si128 (ctr, p[0]);
w += 1;
do
{
MM_OP_m (_mm_aesenc_si128, w[0])
MM_OP_m (_mm_aesenc_si128, w[1])
w += 2;
}
while (--numRounds2);
MM_OP_m (_mm_aesenc_si128, w[0])
MM_OP_m (_mm_aesenclast_si128, w[1])
MM_XOR (*data, m)
}
p[-2] = ctr;
}
#ifdef USE_INTEL_VAES
/*
GCC before 2013-Jun:
<immintrin.h>:
#ifdef __AVX__
#include <avxintrin.h>
#endif
GCC after 2013-Jun:
<immintrin.h>:
#include <avxintrin.h>
CLANG 3.8+:
{
<immintrin.h>:
#if !defined(_MSC_VER) || defined(__AVX__)
#include <avxintrin.h>
#endif
if (the compiler is clang for Windows and if global arch is not set for __AVX__)
[ if (defined(_MSC_VER) && !defined(__AVX__)) ]
{
<immintrin.h> doesn't include <avxintrin.h>
and we have 2 ways to fix it:
1) we can define required __AVX__ before <immintrin.h>
or
2) we can include <avxintrin.h> after <immintrin.h>
}
}
If we include <avxintrin.h> manually for GCC/CLANG, it's
required that <immintrin.h> must be included before <avxintrin.h>.
*/
/*
#if defined(__clang__) && defined(_MSC_VER)
#define __AVX__
#define __AVX2__
#define __VAES__
#endif
*/
#include <immintrin.h>
#if defined(__clang__) && defined(_MSC_VER)
#if !defined(__AVX__)
#include <avxintrin.h>
#endif
#if !defined(__AVX2__)
#include <avx2intrin.h>
#endif
#if !defined(__VAES__)
#include <vaesintrin.h>
#endif
#endif // __clang__ && _MSC_VER
#ifndef ATTRIB_VAES
#define ATTRIB_VAES
#endif
#define VAES_FUNC_START2(name) \
AES_FUNC_START (name); \
ATTRIB_VAES \
AES_FUNC_START (name)
VAES_FUNC_START2 (AesCbc_Decode_HW_256)
{
__m128i *p = (__m128i *)(void *)ivAes;
__m128i *data = (__m128i *)(void *)data8;
__m128i iv = *p;
const __m128i *dataEnd;
UInt32 numRounds = *(const UInt32 *)(p + 1) * 2 + 1;
p += 2;
WIDE_LOOP_START_AVX(;)
{
const __m256i *w = keys + numRounds - 2;
WOP (AVX_DECLARE_VAR)
WOP (AVX_LOAD_data)
AVX_WOP_KEY (AVX_AES_XOR, 1)
do
{
AVX_WOP_KEY (AVX_AES_DEC, 0)
w--;
}
while (w != keys);
AVX_WOP_KEY (AVX_AES_DEC_LAST, 0)
AVX_XOR (m0, _mm256_setr_m128i(iv, data[0]))
WOP_M1 (AVX_XOR_data_M1)
iv = data[NUM_WAYS * 2 - 1];
WOP (AVX_STORE_data)
}
WIDE_LOOP_END_AVX(;)
SINGLE_LOOP
{
const __m128i *w = p + *(const UInt32 *)(p + 1 - 2) * 2 + 1 - 3;
__m128i m = _mm_xor_si128 (w[2], *data);
do
{
MM_OP_m (_mm_aesdec_si128, w[1])
MM_OP_m (_mm_aesdec_si128, w[0])
w -= 2;
}
while (w != p);
MM_OP_m (_mm_aesdec_si128, w[1])
MM_OP_m (_mm_aesdeclast_si128, w[0])
MM_XOR (m, iv)
iv = *data;
*data = m;
}
p[-2] = iv;
}
/*
SSE2: _mm_cvtsi32_si128 : movd
AVX: _mm256_setr_m128i : vinsertf128
AVX2: _mm256_add_epi64 : vpaddq ymm, ymm, ymm
_mm256_extracti128_si256 : vextracti128
_mm256_broadcastsi128_si256 : vbroadcasti128
*/
#define AVX_CTR_LOOP_START \
ctr2 = _mm256_setr_m128i(_mm_sub_epi64(ctr, one), ctr); \
two = _mm256_setr_m128i(one, one); \
two = _mm256_add_epi64(two, two); \
// two = _mm256_setr_epi64x(2, 0, 2, 0);
#define AVX_CTR_LOOP_ENC \
ctr = _mm256_extracti128_si256 (ctr2, 1); \
VAES_FUNC_START2 (AesCtr_Code_HW_256)
{
__m128i *p = (__m128i *)(void *)ivAes;
__m128i *data = (__m128i *)(void *)data8;
__m128i ctr = *p;
UInt32 numRounds = *(const UInt32 *)(p + 1) * 2 + 1;
const __m128i *dataEnd;
__m128i >
__m256i ctr2, two;
p += 2;
WIDE_LOOP_START_AVX (AVX_CTR_LOOP_START)
{
const __m256i *w = keys;
UInt32 r = numRounds - 2;
WOP (AVX_DECLARE_VAR)
AVX_WOP_KEY (AVX_CTR_START, 0)
w += 1;
do
{
AVX_WOP_KEY (AVX_AES_ENC, 0)
w += 1;
}
while (--r);
AVX_WOP_KEY (AVX_AES_ENC_LAST, 0)
WOP (AVX_CTR_END)
}
WIDE_LOOP_END_AVX (AVX_CTR_LOOP_ENC)
SINGLE_LOOP
{
UInt32 numRounds2 = *(const UInt32 *)(p - 2 + 1) - 1;
const __m128i *w = p;
__m128i m;
MM_OP (_mm_add_epi64, ctr, one)
m = _mm_xor_si128 (ctr, p[0]);
w += 1;
do
{
MM_OP_m (_mm_aesenc_si128, w[0])
MM_OP_m (_mm_aesenc_si128, w[1])
w += 2;
}
while (--numRounds2);
MM_OP_m (_mm_aesenc_si128, w[0])
MM_OP_m (_mm_aesenclast_si128, w[1])
MM_XOR (*data, m)
}
p[-2] = ctr;
}
#endif // USE_INTEL_VAES
#else // USE_INTEL_AES
/* no USE_INTEL_AES */
#if defined(Z7_USE_AES_HW_STUB)
// We can compile this file with another C compiler,
// or we can compile asm version.
// So we can generate real code instead of this stub function.
// #if defined(_MSC_VER)
#pragma message("AES HW_SW stub was used")
// #endif
#if !defined(USE_INTEL_VAES) && defined(Z7_USE_VAES_HW_STUB)
#define AES_TYPE_keys UInt32
#define AES_TYPE_data Byte
#endif
#define AES_FUNC_START(name) \
void Z7_FASTCALL name(UInt32 *p, Byte *data, size_t numBlocks) \
#define AES_COMPAT_STUB(name) \
AES_FUNC_START(name); \
AES_FUNC_START(name ## _HW) \
{ name(p, data, numBlocks); }
AES_COMPAT_STUB (AesCbc_Encode)
AES_COMPAT_STUB (AesCbc_Decode)
AES_COMPAT_STUB (AesCtr_Code)
#endif // Z7_USE_AES_HW_STUB
#endif // USE_INTEL_AES
#ifndef USE_INTEL_VAES
#if defined(Z7_USE_VAES_HW_STUB)
// #if defined(_MSC_VER)
#pragma message("VAES HW_SW stub was used")
// #endif
#define VAES_COMPAT_STUB(name) \
void Z7_FASTCALL name ## _256(UInt32 *p, Byte *data, size_t numBlocks); \
void Z7_FASTCALL name ## _256(UInt32 *p, Byte *data, size_t numBlocks) \
{ name((AES_TYPE_keys *)(void *)p, (AES_TYPE_data *)(void *)data, numBlocks); }
VAES_COMPAT_STUB (AesCbc_Decode_HW)
VAES_COMPAT_STUB (AesCtr_Code_HW)
#endif
#endif // ! USE_INTEL_VAES
#elif defined(MY_CPU_ARM_OR_ARM64) && defined(MY_CPU_LE)
#if defined(__ARM_FEATURE_AES) \
|| defined(__ARM_FEATURE_CRYPTO)
#define USE_HW_AES
#else
#if defined(MY_CPU_ARM64) \
|| defined(__ARM_ARCH) && (__ARM_ARCH >= 4) \
|| defined(Z7_MSC_VER_ORIGINAL)
#if defined(__ARM_FP) && \
( defined(Z7_CLANG_VERSION) && (Z7_CLANG_VERSION >= 30800) \
|| defined(__GNUC__) && (__GNUC__ >= 6) \
) \
|| defined(Z7_MSC_VER_ORIGINAL) && (_MSC_VER >= 1910)
#if defined(MY_CPU_ARM64) \
|| !defined(Z7_CLANG_VERSION) \
|| defined(__ARM_NEON) && \
(Z7_CLANG_VERSION < 170000 || \
Z7_CLANG_VERSION > 170001)
#define USE_HW_AES
#endif
#endif
#endif
#endif
#ifdef USE_HW_AES
// #pragma message("=== AES HW === ")
// __ARM_FEATURE_CRYPTO macro is deprecated in favor of the finer grained feature macro __ARM_FEATURE_AES
#if defined(__clang__) || defined(__GNUC__)
#if !defined(__ARM_FEATURE_AES) && \
!defined(__ARM_FEATURE_CRYPTO)
#ifdef MY_CPU_ARM64
#if defined(__clang__)
#define ATTRIB_AES __attribute__((__target__("crypto")))
#else
#define ATTRIB_AES __attribute__((__target__("+crypto")))
#endif
#else
#if defined(__clang__)
#define ATTRIB_AES __attribute__((__target__("armv8-a,aes")))
#else
#define ATTRIB_AES __attribute__((__target__("fpu=crypto-neon-fp-armv8")))
#endif
#endif
#endif
#else
// _MSC_VER
// for arm32
#define _ARM_USE_NEW_NEON_INTRINSICS
#endif
#ifndef ATTRIB_AES
#define ATTRIB_AES
#endif
#if defined(Z7_MSC_VER_ORIGINAL) && defined(MY_CPU_ARM64)
#include <arm64_neon.h>
#else
/*
clang-17.0.1: error : Cannot select: intrinsic %llvm.arm.neon.aese
clang
3.8.1 : __ARM_NEON : defined(__ARM_FEATURE_CRYPTO)
7.0.1 : __ARM_NEON : __ARM_ARCH >= 8 && defined(__ARM_FEATURE_CRYPTO)
11.?.0 : __ARM_NEON && __ARM_FP : __ARM_ARCH >= 8 && defined(__ARM_FEATURE_CRYPTO)
13.0.1 : __ARM_NEON && __ARM_FP : __ARM_ARCH >= 8 && defined(__ARM_FEATURE_AES)
16 : __ARM_NEON && __ARM_FP : __ARM_ARCH >= 8
*/
#if defined(__clang__) && __clang_major__ < 16
#if !defined(__ARM_FEATURE_AES) && \
!defined(__ARM_FEATURE_CRYPTO)
// #pragma message("=== we set __ARM_FEATURE_CRYPTO 1 === ")
Z7_DIAGNOSTIC_IGNORE_BEGIN_RESERVED_MACRO_IDENTIFIER
#define Z7_ARM_FEATURE_CRYPTO_WAS_SET 1
// #if defined(__clang__) && __clang_major__ < 13
#define __ARM_FEATURE_CRYPTO 1
// #else
#define __ARM_FEATURE_AES 1
// #endif
Z7_DIAGNOSTIC_IGNORE_END_RESERVED_MACRO_IDENTIFIER
#endif
#endif // clang
#if defined(__clang__)
#if defined(__ARM_ARCH) && __ARM_ARCH < 8
Z7_DIAGNOSTIC_IGNORE_BEGIN_RESERVED_MACRO_IDENTIFIER
// #pragma message("#define __ARM_ARCH 8")
#undef __ARM_ARCH
#define __ARM_ARCH 8
Z7_DIAGNOSTIC_IGNORE_END_RESERVED_MACRO_IDENTIFIER
#endif
#endif // clang
#include <arm_neon.h>
#if defined(Z7_ARM_FEATURE_CRYPTO_WAS_SET) && \
defined(__ARM_FEATURE_CRYPTO) && \
defined(__ARM_FEATURE_AES)
Z7_DIAGNOSTIC_IGNORE_BEGIN_RESERVED_MACRO_IDENTIFIER
#undef __ARM_FEATURE_CRYPTO
#undef __ARM_FEATURE_AES
#undef Z7_ARM_FEATURE_CRYPTO_WAS_SET
Z7_DIAGNOSTIC_IGNORE_END_RESERVED_MACRO_IDENTIFIER
// #pragma message("=== we undefine __ARM_FEATURE_CRYPTO === ")
#endif
#endif // Z7_MSC_VER_ORIGINAL
typedef uint8x16_t v128;
#define AES_FUNC_START(name) \
void Z7_FASTCALL name(UInt32 *ivAes, Byte *data8, size_t numBlocks)
// void Z7_FASTCALL name(v128 *p, v128 *data, size_t numBlocks)
#define AES_FUNC_START2(name) \
AES_FUNC_START (name); \
ATTRIB_AES \
AES_FUNC_START (name)
#define MM_OP(op, dest, src) dest = op(dest, src);
#define MM_OP_m(op, src) MM_OP(op, m, src)
#define MM_OP1_m(op) m = op(m);
#define MM_XOR( dest, src) MM_OP(veorq_u8, dest, src)
#define MM_XOR_m( src) MM_XOR(m, src)
#define AES_E_m(k) MM_OP_m (vaeseq_u8, k)
#define AES_E_MC_m(k) AES_E_m (k) MM_OP1_m(vaesmcq_u8)
AES_FUNC_START2 (AesCbc_Encode_HW)
{
v128 * const p = (v128*)(void*)ivAes;
v128 *data = (v128*)(void*)data8;
v128 m = *p;
const v128 k0 = p[2];
const v128 k1 = p[3];
const v128 k2 = p[4];
const v128 k3 = p[5];
const v128 k4 = p[6];
const v128 k5 = p[7];
const v128 k6 = p[8];
const v128 k7 = p[9];
const v128 k8 = p[10];
const v128 k9 = p[11];
const UInt32 numRounds2 = *(const UInt32 *)(p + 1);
const v128 *w = p + ((size_t)numRounds2 * 2);
const v128 k_z1 = w[1];
const v128 k_z0 = w[2];
for (; numBlocks != 0; numBlocks--, data++)
{
MM_XOR_m (*data)
AES_E_MC_m (k0)
AES_E_MC_m (k1)
AES_E_MC_m (k2)
AES_E_MC_m (k3)
AES_E_MC_m (k4)
AES_E_MC_m (k5)
AES_E_MC_m (k6)
AES_E_MC_m (k7)
AES_E_MC_m (k8)
if (numRounds2 >= 6)
{
AES_E_MC_m (k9)
AES_E_MC_m (p[12])
if (numRounds2 != 6)
{
AES_E_MC_m (p[13])
AES_E_MC_m (p[14])
}
}
AES_E_m (k_z1)
MM_XOR_m (k_z0)
*data = m;
}
*p = m;
}
#define WOP_1(op)
#define WOP_2(op) WOP_1 (op) op (m1, 1)
#define WOP_3(op) WOP_2 (op) op (m2, 2)
#define WOP_4(op) WOP_3 (op) op (m3, 3)
#define WOP_5(op) WOP_4 (op) op (m4, 4)
#define WOP_6(op) WOP_5 (op) op (m5, 5)
#define WOP_7(op) WOP_6 (op) op (m6, 6)
#define WOP_8(op) WOP_7 (op) op (m7, 7)
#define NUM_WAYS 8
#define WOP_M1 WOP_8
#define WOP(op) op (m0, 0) WOP_M1(op)
#define DECLARE_VAR(reg, ii) v128 reg;
#define LOAD_data( reg, ii) reg = data[ii];
#define STORE_data( reg, ii) data[ii] = reg;
#if (NUM_WAYS > 1)
#define XOR_data_M1(reg, ii) MM_XOR (reg, data[ii- 1])
#endif
#define MM_OP_key(op, reg) MM_OP (op, reg, key)
#define AES_D_m(k) MM_OP_m (vaesdq_u8, k)
#define AES_D_IMC_m(k) AES_D_m (k) MM_OP1_m (vaesimcq_u8)
#define AES_XOR( reg, ii) MM_OP_key (veorq_u8, reg)
#define AES_D( reg, ii) MM_OP_key (vaesdq_u8, reg)
#define AES_E( reg, ii) MM_OP_key (vaeseq_u8, reg)
#define AES_D_IMC( reg, ii) AES_D (reg, ii) reg = vaesimcq_u8(reg);
#define AES_E_MC( reg, ii) AES_E (reg, ii) reg = vaesmcq_u8(reg);
#define CTR_START(reg, ii) MM_OP (vaddq_u64, ctr, one) reg = vreinterpretq_u8_u64(ctr);
#define CTR_END( reg, ii) MM_XOR (data[ii], reg)
#define WOP_KEY(op, n) { \
const v128 key = w[n]; \
WOP(op) }
#define WIDE_LOOP_START \
dataEnd = data + numBlocks; \
if (numBlocks >= NUM_WAYS) \
{ dataEnd -= NUM_WAYS; do { \
#define WIDE_LOOP_END \
data += NUM_WAYS; \
} while (data <= dataEnd); \
dataEnd += NUM_WAYS; } \
#define SINGLE_LOOP \
for (; data < dataEnd; data++)
AES_FUNC_START2 (AesCbc_Decode_HW)
{
v128 *p = (v128*)(void*)ivAes;
v128 *data = (v128*)(void*)data8;
v128 iv = *p;
const v128 *wStart = p + ((size_t)*(const UInt32 *)(p + 1)) * 2;
const v128 *dataEnd;
p += 2;
WIDE_LOOP_START
{
const v128 *w = wStart;
WOP (DECLARE_VAR)
WOP (LOAD_data)
WOP_KEY (AES_D_IMC, 2)
do
{
WOP_KEY (AES_D_IMC, 1)
WOP_KEY (AES_D_IMC, 0)
w -= 2;
}
while (w != p);
WOP_KEY (AES_D, 1)
WOP_KEY (AES_XOR, 0)
MM_XOR (m0, iv)
WOP_M1 (XOR_data_M1)
iv = data[NUM_WAYS - 1];
WOP (STORE_data)
}
WIDE_LOOP_END
SINGLE_LOOP
{
const v128 *w = wStart;
v128 m = *data;
AES_D_IMC_m (w[2])
do
{
AES_D_IMC_m (w[1])
AES_D_IMC_m (w[0])
w -= 2;
}
while (w != p);
AES_D_m (w[1])
MM_XOR_m (w[0])
MM_XOR_m (iv)
iv = *data;
*data = m;
}
p[-2] = iv;
}
AES_FUNC_START2 (AesCtr_Code_HW)
{
v128 *p = (v128*)(void*)ivAes;
v128 *data = (v128*)(void*)data8;
uint64x2_t ctr = vreinterpretq_u64_u8(*p);
const v128 *wEnd = p + ((size_t)*(const UInt32 *)(p + 1)) * 2;
const v128 *dataEnd;
uint64x2_t >
// the bug in clang:
// __builtin_neon_vsetq_lane_i64(__s0, (int8x16_t)__s1, __p2);
#if defined(__clang__) && (__clang_major__ <= 9)
#pragma GCC diagnostic ignored "-Wvector-conversion"
#endif
one, 0);
p += 2;
WIDE_LOOP_START
{
const v128 *w = p;
WOP (DECLARE_VAR)
WOP (CTR_START)
do
{
WOP_KEY (AES_E_MC, 0)
WOP_KEY (AES_E_MC, 1)
w += 2;
}
while (w != wEnd);
WOP_KEY (AES_E_MC, 0)
WOP_KEY (AES_E, 1)
WOP_KEY (AES_XOR, 2)
WOP (CTR_END)
}
WIDE_LOOP_END
SINGLE_LOOP
{
const v128 *w = p;
v128 m;
CTR_START (m, 0)
do
{
AES_E_MC_m (w[0])
AES_E_MC_m (w[1])
w += 2;
}
while (w != wEnd);
AES_E_MC_m (w[0])
AES_E_m (w[1])
MM_XOR_m (w[2])
CTR_END (m, 0)
}
p[-2] = vreinterpretq_u8_u64(ctr);
}
#endif // USE_HW_AES
#endif // MY_CPU_ARM_OR_ARM64
#undef NUM_WAYS
#undef WOP_M1
#undef WOP
#undef DECLARE_VAR
#undef LOAD_data
#undef STORE_data
#undef USE_INTEL_AES
#undef USE_HW_AES
You can’t perform that action at this time.