forked from JayDDee/cpuminer-opt
/
avxdefs.h
1241 lines (1043 loc) · 45.2 KB
/
avxdefs.h
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
#ifndef AVXDEFS_H__
#define AVXDEFS_H__
// Some tools to help using AVX and AVX2.
// At this time SSE2 is sufficient for all 128 bit code in this file
// but could change without notice.
// 256 bit requires AVX2.
// AVX512 has more powerful 256 bit instructions but with AVX512 available
// there is little reason to use them.
// Proper alignment of data is required, 16 bytes for 128 bit vectors and
// 32 bytes for 256 bit vectors. 64 byte alignment is recommended for
// best cache alignment.
//
// There exist dupplicates of some functions. In general the first defined
// is preferred as it is more efficient but also more restrictive and may
// not be applicable. The less efficient versions are more flexible.
#include <inttypes.h>
#include <immintrin.h>
#include <memory.h>
#include <stdbool.h>
//
// 128 bit utilities and shortcuts
//
// Pseudo constants, there are no real vector constants.
// These can't be used for compile time initialization.
// Constant zero
#define mm_zero _mm_setzero_si128()
// Constant 1
#define mm_one_128 _mm_set_epi64x( 0ULL, 1ULL )
#define mm_one_64 _mm_set1_epi64x( 1ULL )
#define mm_one_32 _mm_set1_epi32( 1UL )
#define mm_one_16 _mm_set1_epi16( 1U )
#define mm_one_8 _mm_set1_epi8( 1U )
// Constant minus 1
#define mm_neg1 _mm_set1_epi64x( 0xFFFFFFFFFFFFFFFFULL )
// Lane index, useful for byte rotate using shuffle
#define mm_lanex_64 _mm_set_epi64( 1ULL, 0ULL );
#define mm_lanex_32 _mm_set_epi32( 3UL, 2UL, 1UL, 0UL );
#define mm_lanex_16 _mm_set_epi16( 7U, 6U, 5U, 4U, 3U, 2U, 1U, 0U );
#define mm_lanex_8 _mm_set_epi8( 15U, 14U, 13U, 12U, 11U, 10U , 9U, 8U, \
7U, 6U, 5U, 4U, 3U, 2U, 1U, 0U );
//
// Basic operations without equivalent SIMD intrinsic
// Bitwise not (~x)
#define mm_not( x ) _mm_xor_si128( (x), mm_neg1 )
// Unary negation (-a)
#define mm_negate_64( a ) _mm_sub_epi64( mm_zero, a )
#define mm_negate_32( a ) _mm_sub_epi32( mm_zero, a )
#define mm_negate_16( a ) _mm_sub_epi16( mm_zero, a )
//
// Bit operations
// Return bit n in position, all other bits zeroed.
#define mm_bitextract_64 ( x, n ) \
_mm_and_si128( _mm_slli_epi64( mm_one_64, n ), x )
#define mm_bitextract_32 ( x, n ) \
_mm_and_si128( _mm_slli_epi32( mm_one_32, n ), x )
#define mm_bitextract_16 ( x, n ) \
_mm_and_si128( _mm_slli_epi16( mm_one_16, n ), x )
// Return bit n as bool
#define mm_bittest_64( x, n ) \
_mm_and_si256( mm_one_64, _mm_srli_epi64( x, n ) )
#define mm_bittest_32( x, n ) \
_mm_and_si256( mm_one_32, _mm_srli_epi32( x, n ) )
#define mm_bittest_16( x, n ) \
_mm_and_si256( mm_one_16, _mm_srli_epi16( x, n ) )
// Return x with bit n set/cleared in all elements
#define mm_bitset_64( x, n ) \
_mm_or_si128( _mm_slli_epi64( mm_one_64, n ), x )
#define mm_bitclr_64( x, n ) \
_mm_andnot_si128( _mm_slli_epi64( mm_one_64, n ), x )
#define mm_bitset_32( x, n ) \
_mm_or_si128( _mm_slli_epi32( mm_one_32, n ), x )
#define mm_bitclr_32( x, n ) \
_mm_andnot_si128( _mm_slli_epi32( mm_one_32, n ), x )
#define mm_bitset_16( x, n ) \
_mm_or_si128( _mm_slli_epi16( mm_one_16, n ), x )
#define mm_bitclr_16( x, n ) \
_mm_andnot_si128( _mm_slli_epi16( mm_one_16, n ), x )
// Return x with bit n toggled
#define mm_bitflip_64( x, n ) \
_mm_xor_si128( _mm_slli_epi64( mm_one_64, n ), x )
#define mm_bitflip_32( x, n ) \
_mm_xor_si128( _mm_slli_epi32( mm_one_32, n ), x )
#define mm_bitflip_16( x, n ) \
_mm_xor_si128( _mm_slli_epi16( mm_one_16, n ), x )
//
// Memory functions
// n = number of __m128i, bytes/16
inline void memset_zero_128( __m128i *dst, int n )
{
for ( int i = 0; i < n; i++ ) dst[i] = mm_zero;
}
inline void memset_128( __m128i *dst, const __m128i a, int n )
{
for ( int i = 0; i < n; i++ ) dst[i] = a;
}
inline void memcpy_128( __m128i *dst, const __m128i *src, int n )
{
for ( int i = 0; i < n; i ++ ) dst[i] = src[i];
}
// Compare data in memory, return true if different
inline bool memcmp_128( __m128i src1, __m128i src2, int n )
{
for ( int i = 0; i < n; i++ )
if ( src1[i] != src2[i] ) return true;
return false;
}
// A couple of 64 bit scalar functions
// n = bytes/8
inline void memcpy_64( uint64_t *dst, const uint64_t *src, int n )
{
for ( int i = 0; i < n; i++ ) dst[i] = src[i];
}
inline void memset_zero_64( uint64_t *src, int n )
{
for ( int i = 0; i < n; i++ ) src[i] = 0;
}
inline void memset_64( uint64_t *dst, uint64_t a, int n )
{
for ( int i = 0; i < n; i++ ) dst[i] = a;
}
//
// Pointer cast
// p = any aligned pointer
// returns p as pointer to vector type
#define castp_m128i(p) ((__m128i*)(p))
// p = any aligned pointer
// returns *p, watch your pointer arithmetic
#define cast_m128i(p) (*((__m128i*)(p)))
// p = any aligned pointer, i = scaled array index
// returns p[i]
#define casti_m128i(p,i) (((__m128i*)(p))[(i)])
//
// Bit rotations
// XOP is an obsolete AMD feature that has native rotation.
// _mm_roti_epi64( w, c)
// Never implemented by Intel and since removed from Zen by AMD.
// Rotate bits in vector elements
#define mm_rotr_64( w, c ) _mm_or_si128( _mm_srli_epi64( w, c ), \
_mm_slli_epi64( w, 64-(c) ) )
#define mm_rotl_64( w, c ) _mm_or_si128( _mm_slli_epi64( w, c ), \
_mm_srli_epi64( w, 64-(c) ) )
#define mm_rotr_32( w, c ) _mm_or_si128( _mm_srli_epi32( w, c ), \
_mm_slli_epi32( w, 32-(c) ) )
#define mm_rotl_32( w, c ) _mm_or_si128( _mm_slli_epi32( w, c ), \
_mm_srli_epi32( w, 32-(c) ) )
#define mm_rotr_16( w, c ) _mm_or_si128( _mm_srli_epi16( w, c ), \
_mm_slli_epi16( w, 16-(c) ) )
#define mm_rotl_16( w, c ) _mm_or_si128( _mm_slli_epi16( w, c ), \
_mm_srli_epi16( w, 16-(c) ) )
//
// Rotate elements in vector
// Optimized shuffle
// Swap hi/lo 64 bits in 128 bit vector
#define mm_swap_64( w ) _mm_shuffle_epi32( w, 0x4e )
// rotate 128 bit vector by 32 bits
#define mm_rotr_1x32( w ) _mm_shuffle_epi32( w, 0x39 )
#define mm_rotl_1x32( w ) _mm_shuffle_epi32( w, 0x93 )
// Swap hi/lo 32 bits in each 64 bit element
#define mm_swap64_32( x ) _mm_shuffle_epi32( x, 0xb1 )
// Less efficient but more versatile. Use only for odd number rotations.
// Use shuffle above when possible.
// Rotate vector by n bytes.
#define mm_rotr128_x8( w, n ) \
_mm_or_si128( _mm_srli_si128( w, n ), _mm_slli_si128( w, 16-(n) ) )
#define mm_rotl128_x8( w, n ) \
_mm_or_si128( _mm_slli_si128( w, n ), _mm_srli_si128( w, 16-(n) ) )
// Rotate vector by c elements, use only for odd number rotations
#define mm_rotr128_x32( w, c ) mm_rotr128_x8( w, (c)>>2 )
#define mm_rotl128_x32( w, c ) mm_rotl128_x8( w, (c)>>2 )
#define mm_rotr128_x16( w, c ) mm_rotr128_x8( w, (c)>>1 )
#define mm_rotl128_x16( w, c ) mm_rotl128_x8( w, (c)>>1 )
//
// Rotate elements across two 128 bit vectors as one 256 bit vector {hi,lo}
// Swap 128 bit source vectors in place, aka rotate 256 bits by 128 bits.
// void mm128_swap128( __m128i, __m128i )
#define mm_swap_128(hi, lo) \
{ \
hi = _mm_xor_si128(hi, lo); \
lo = _mm_xor_si128(hi, lo); \
hi = _mm_xor_si128(hi, lo); \
}
// Rotate two 128 bit vectors in place as one 256 vector by 1 element
#define mm_rotl256_1x64( hi, lo ) \
do { \
__m128i t; \
hi = mm_swap_64( hi ); \
lo = mm_swap_64( lo ); \
t = _mm_blendv_epi8( hi, lo, _mm_set_epi64x( 0xffffffffffffffffull, 0ull )); \
lo = _mm_blendv_epi8( hi, lo, _mm_set_epi64x( 0ull, 0xffffffffffffffffull )); \
hi = t; \
} while(0)
#define mm_rotr256_1x64( hi, lo ) \
do { \
__m128i t; \
hi = mm_swap_64( hi ); \
lo = mm_swap_64( lo ); \
t = _mm_blendv_epi8( hi, lo, _mm_set_epi64x( 0ull, 0xffffffffffffffffull )); \
lo = _mm_blendv_epi8( hi, lo, _mm_set_epi64x( 0xffffffffffffffffull, 0ull )); \
hi = t; \
} while(0)
#define mm_rotl256_1x32( hi, lo ) \
do { \
__m128i t; \
hi = mm_swap_64( hi ); \
lo = mm_swap_64( lo ); \
t = _mm_blendv_epi8( hi, lo, _mm_set_epi32( \
0xfffffffful, 0xfffffffful, 0xfffffffful, 0ul )); \
lo = _mm_blendv_epi8( hi, lo, _mm_set_epi32( \
0ul, 0ul, 0ul, 0xfffffffful )); \
hi = t; \
} while(0)
#define mm_rotr256_1x32( hi, lo ) \
do { \
__m128i t; \
hi = mm_swap_64( hi ); \
lo = mm_swap_64( lo ); \
t = _mm_blendv_epi8( hi, lo, _mm_set_epi32( \
0ul, 0ul, 0ul, 0xfffffffful )); \
lo = _mm_blendv_epi8( hi, lo, _mm_set_epi32( \
0xfffffffful, 0xfffffffful, 0xfffffffful, 0ul )); \
hi = t; \
} while(0)
// Return hi 128 bits with elements shifted one lane with vacated lane filled
// with data rotated from lo.
// Partially rotate elements in two 128 bit vectors as one 256 bit vector
// and return the rotated high 128 bits.
// Similar to mm_rotr256_1x32 but only a partial rotation as lo is not
// completed. It's faster than a full rotation.
inline __m128i mm_rotr256hi_1x32( __m128i hi, __m128i lo, int n )
{
return _mm_or_si128( _mm_srli_si128( hi, n<<2 ),
_mm_slli_si128( lo, 16 - (n<<2) ) );
}
inline __m128i mm_rotl256hi_1x32( __m128i hi, __m128i lo, int n )
{
return _mm_or_si128( _mm_slli_si128( hi, n<<2 ),
_mm_srli_si128( lo, 16 - (n<<2) ) );
}
//
// Swap bytes in vector elements
inline __m128i mm_byteswap_64( __m128i x )
{
return _mm_shuffle_epi8( x, _mm_set_epi8(
0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f,
0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07 ) );
}
inline __m128i mm_byteswap_32( __m128i x )
{
return _mm_shuffle_epi8( x, _mm_set_epi8(
0x0c, 0x0d, 0x0e, 0x0f, 0x08, 0x09, 0x0a, 0x0b,
0x04, 0x05, 0x06, 0x07, 0x00, 0x01, 0x02, 0x03 ) );
}
inline __m128i mm_byteswap_16( __m128i x )
{
return _mm_shuffle_epi8( x, _mm_set_epi8(
0x0e, 0x0f, 0x0c, 0x0d, 0x0a, 0x0b, 0x08, 0x09,
0x06, 0x07, 0x04, 0x05, 0x02, 0x03, 0x00, 0x01 ) );
}
/////////////////////////////////////////////////////////////////////
#if defined (__AVX2__)
//
// 256 bit utilities and Shortcuts
//
// Pseudo constants, there are no real vector constants.
// These can't be used for compile time initialization
// Constant zero
#define mm256_zero _mm256_setzero_si256()
// Constant 1
#define mm256_one_128 _mm256_set_epi64x( 0ULL, 1ULL, 0ULL, 1ULL )
#define mm256_one_64 _mm256_set1_epi64x( 1ULL )
#define mm256_one_32 _mm256_set1_epi32( 1UL )
#define mm256_one_16 _mm256_set1_epi16( 1U )
// Constant minus 1
#define mm256_neg1 _mm256_set1_epi64x( 0xFFFFFFFFFFFFFFFFULL )
// Lane index, useful for rotate using permutevar
#define mm256_lane_64 _mm_set_epi64x( 3ULL, 2ULL, 1ULL, 0ULL );
#define mm256_lane_32 _mm_set_epi32( 7UL, 6UL, 5UL, 4UL, 3UL, 2UL, 1UL, 0UL );
#define mm256_lane_16 _mm_set_epi16( 15U, 14U, 13U, 12U, 11U, 10U , 9U, 8U, \
7U, 6U, 5U, 4U, 3U, 2U, 1U, 0U );
#define mm256_lane_8 _mm_set_epi8( 31U, 30U, 29U, 28U, 27U, 26U, 25U, 24U, \
23U, 22U, 21U, 20U, 19U, 18U, 17U, 16U, \
15U, 14U, 13U, 12U, 11U, 10U , 9U, 8U, \
7U, 6U, 5U, 4U, 3U, 2U, 1U, 0U );
//
// Basic operations without SIMD equivalent
// Bitwise not ( ~x )
#define mm256_not( x ) _mm256_xor_si256( (x), mm256_neg1 ) \
// Unary negation ( -a )
#define mm256_negate_64( a ) _mm256_sub_epi64( mm256_zero, a )
#define mm256_negate_32( a ) _mm256_sub_epi32( mm256_zero, a )
#define mm256_negate_16( a ) _mm256_sub_epi16( mm256_zero, a )
//
// Bit operations
// return bit n in position, all othr bits cleared
#define mm256_bitextract_64 ( x, n ) \
_mm256_and_si128( _mm256_slli_epi64( mm256_one_64, n ), x )
#define mm256_bitextract_32 ( x, n ) \
_mm256_and_si128( _mm256_slli_epi32( mm256_one_32, n ), x )
#define mm256_bitextract_16 ( x, n ) \
_mm256_and_si128( _mm256_slli_epi16( mm256_one_16, n ), x )
// Return bit n as bool (bit 0)
#define mm256_bittest_64( x, n ) \
_mm256_and_si256( mm256_one_64, _mm256_srli_epi64( x, n ) )
#define mm256_bittest_32( x, n ) \
_mm256_and_si256( mm256_one_32, _mm256_srli_epi32( x, n ) )
#define mm256_bittest_16( x, n ) \
_mm256_and_si256( mm256_one_16, _mm256_srli_epi16( x, n ) )
// Return x with bit n set/cleared in all elements
#define mm256_bitset_64( x, n ) \
_mm256_or_si256( _mm256_slli_epi64( mm256_one_64, n ), x )
#define mm256_bitclr_64( x, n ) \
_mm256_andnot_si256( _mm256_slli_epi64( mm256_one_64, n ), x )
#define mm256_bitset_32( x, n ) \
_mm256_or_si256( _mm256_slli_epi32( mm256_one_32, n ), x )
#define mm256_bitclr_32( x, n ) \
_mm256_andnot_si256( _mm256_slli_epi32( mm256_one_32, n ), x )
#define mm256_bitset_16( x, n ) \
_mm256_or_si256( _mm256_slli_epi16( mm256_one_16, n ), x )
#define mm256_bitclr_16( x, n ) \
_mm256_andnot_si256( _mm256_slli_epi16( mm256_one_16, n ), x )
// Return x with bit n toggled
#define mm256_bitflip_64( x, n ) \
_mm256_xor_si128( _mm256_slli_epi64( mm256_one_64, n ), x )
#define mm256_bitflip_32( x, n ) \
_mm256_xor_si128( _mm256_slli_epi32( mm256_one_32, n ), x )
#define mm256_bitflip_16( x, n ) \
_mm256_xor_si128( _mm256_slli_epi16( mm256_one_16, n ), x )
//
// Memory functions
// n = number of 256 bit (32 byte) vectors
inline void memset_zero_256( __m256i *dst, int n )
{
for ( int i = 0; i < n; i++ ) dst[i] = mm256_zero;
}
inline void memset_256( __m256i *dst, const __m256i a, int n )
{
for ( int i = 0; i < n; i++ ) dst[i] = a;
}
inline void memcpy_256( __m256i *dst, const __m256i *src, int n )
{
for ( int i = 0; i < n; i ++ ) dst[i] = src[i];
}
// Compare data in memory, return true if different
inline bool memcmp_256( __m256i src1, __m256i src2, int n )
{
for ( int i = 0; i < n; i++ )
if ( src1[i] != src2[i] ) return true;
return false;
}
//
// Pointer casting
// p = any aligned pointer
// returns p as pointer to vector type, not very useful
#define castp_m256i(p) ((__m256i*)(p))
// p = any aligned pointer
// returns *p, watch your pointer arithmetic
#define cast_m256i(p) (*((__m256i*)(p)))
// p = any aligned pointer, i = scaled array index
// returns p[i]
#define casti_m256i(p,i) (((__m256i*)(p))[(i)])
//
// Bit rotations
//
// Rotate bits in vector elements
// w = packed data, c = number of bits to rotate
#define mm256_rotr_64( w, c ) \
_mm256_or_si256( _mm256_srli_epi64(w, c), _mm256_slli_epi64(w, 64-(c)) )
#define mm256_rotl_64( w, c ) \
_mm256_or_si256( _mm256_slli_epi64(w, c), _mm256_srli_epi64(w, 64-(c)) )
#define mm256_rotr_32( w, c ) \
_mm256_or_si256( _mm256_srli_epi32(w, c), _mm256_slli_epi32(w, 32-(c)) )
#define mm256_rotl_32( w, c ) \
_mm256_or_si256( _mm256_slli_epi32(w, c), _mm256_srli_epi32(w, 32-(c)) )
#define mm256_rotr_16( w, c ) \
_mm256_or_si256( _mm256_srli_epi16(w, c), _mm256_slli_epi16(w, 32-(c)) )
#define mm256_rotl_16( w, c ) \
_mm256_or_si256( _mm256_slli_epi16(w, c), _mm256_srli_epi16(w, 32-(c)) )
//
// Rotate elements in vector
// There is no full vector permute for elements less than 64 bits or 256 bit
// shift, a little more work is needed.
// Optimized 64 bit permutations
// Swap 128 bit elements in 256 bit vector
#define mm256_swap_128( w ) _mm256_permute4x64_epi64( w, 0x4e )
// Rotate 256 bit vector by one 64 bit element
#define mm256_rotl256_1x64( w ) _mm256_permute4x64_epi64( w, 0x93 )
#define mm256_rotr256_1x64( w ) _mm256_permute4x64_epi64( w, 0x39 )
// Swap 64 bits in each 128 bit element of 256 bit vector
#define mm256_swap128_64( x ) _mm256_shuffle_epi32( x, 0x4e )
// Rotate 128 bit elements in 256 bit vector by 32 bits
#define mm256_rotr128_1x32( x ) _mm256_shuffle_epi32( x, 0x39 )
#define mm256_rotl128_1x32( x ) _mm256_shuffle_epi32( x, 0x93 )
// Swap 32 bits in each 64 bit element olf 256 bit vector
#define mm256_swap64_32( x ) _mm256_shuffle_epi32( x, 0xb1 )
// Less efficient but more versatile. Use only for rotations that are not
// integrals of 64 bits. Use permutations above when possible.
// Rotate 256 bit vector by c bytes.
#define mm256_rotr256_x8( w, c ) \
_mm256_or_si256( _mm256_srli_si256( w, c ), \
mm256_swap_128( _mm256i_slli_si256( w, 32-(c) ) ) )
#define mm256_rotl256_x8( w, c ) \
_mm256_or_si256( _mm256_slli_si256( w, c ), \
mm256_swap_128( _mm256i_srli_si256( w, 32-(c) ) ) )
// Rotate 256 bit vector by c elements, use only for odd value rotations
#define mm256_rotr256_x32( w, c ) mm256_rotr256_x8( w, (c)>>2 )
#define mm256_rotl256_x32( w, c ) mm256_rotl256_x8( w, (c)>>2 )
#define mm256_rotr256_x16( w, c ) mm256_rotr256_x8( w, (c)>>1 )
#define mm256_rotl256_x16( w, c ) mm256_rotl256_x8( w, (c)>>1 )
//
// Rotate two 256 bit vectors as one 512 bit vector
// Fast but limited to 128 bit granularity
#define mm256_swap512_256(a, b) _mm256_permute2x128_si256( a, b, 0x4e )
#define mm256_rotr512_1x128(a, b) _mm256_permute2x128_si256( a, b, 0x39 )
#define mm256_rotl512_1x128(a, b) _mm256_permute2x128_si256( a, b, 0x93 )
// Much slower, for 64 and 32 bit granularity
#define mm256_rotr512_1x64(a, b) \
do { \
__m256i t; \
t = _mm256_or_si256( _mm256_srli_si256(a,8), _mm256_slli_si256(b,24) ); \
b = _mm256_or_si256( _mm256_srli_si256(b,8), _mm256_slli_si256(a,24) ); \
a = t; \
while (0);
#define mm256_rotl512_1x64(a, b) \
do { \
__m256i t; \
t = _mm256_or_si256( _mm256_slli_si256(a,8), _mm256_srli_si256(b,24) ); \
b = _mm256_or_si256( _mm256_slli_si256(b,8), _mm256_srli_si256(a,24) ); \
a = t; \
while (0);
#define mm256_rotr512_1x32(a, b) \
do { \
__m256i t; \
t = _mm256_or_si256( _mm256_srli_si256(a,4), _mm256_slli_si256(b,28) ); \
b = _mm256_or_si256( _mm256_srli_si256(b,4), _mm256_slli_si256(a,28) ); \
a = t; \
while (0);
#define mm256_rotl512_1x32(a, b) \
do { \
__m256i t; \
t = _mm256_or_si256( _mm256_slli_si256(a,4), _mm256_srli_si256(b,28) ); \
b = _mm256_or_si256( _mm256_slli_si256(b,4), _mm256_srli_si256(a,28) ); \
a = t; \
while (0);
// Byte granularity but even a bit slower
#define mm256_rotr512_x8( a, b, n ) \
do { \
__m256i t; \
t = _mm256_or_si256( _mm256_srli_epi64( a, n ), \
_mm256_slli_epi64( b, ( 32 - (n) ) ) ); \
b = _mm256_or_si256( _mm256_srli_epi64( b, n ), \
_mm256_slli_epi64( a, ( 32 - (n) ) ) ); \
a = t; \
while (0);
#define mm256_rotl512_x8( a, b, n ) \
do { \
__m256i t; \
t = _mm256_or_si256( _mm256_slli_epi64( a, n ), \
_mm256_srli_epi64( b, ( 32 - (n) ) ) ); \
b = _mm256_or_si256( _mm256_slli_epi64( b, n ), \
_mm256_srli_epi64( a, ( 32 - (n) ) ) ); \
a = t; \
while (0);
//
// Swap bytes in vector elements
inline __m256i mm256_byteswap_64( __m256i x )
{
return _mm256_shuffle_epi8( x, _mm256_set_epi8(
0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f,
0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f,
0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07 ) );
}
inline __m256i mm256_byteswap_32( __m256i x )
{
return _mm256_shuffle_epi8( x, _mm256_set_epi8(
0x0c, 0x0d, 0x0e, 0x0f, 0x08, 0x09, 0x0a, 0x0b,
0x04, 0x05, 0x06, 0x07, 0x00, 0x01, 0x02, 0x03,
0x0c, 0x0d, 0x0e, 0x0f, 0x08, 0x09, 0x0a, 0x0b,
0x04, 0x05, 0x06, 0x07, 0x00, 0x01, 0x02, 0x03 ) );
}
inline __m256i mm256_byteswap_16( __m256i x )
{
return _mm256_shuffle_epi8( x, _mm256_set_epi8(
0x0e, 0x0f, 0x0c, 0x0d, 0x0a, 0x0b, 0x08, 0x09,
0x06, 0x07, 0x04, 0x05, 0x02, 0x03, 0x00, 0x01,
0x0e, 0x0f, 0x0c, 0x0d, 0x0a, 0x0b, 0x08, 0x09,
0x06, 0x07, 0x04, 0x05, 0x02, 0x03, 0x00, 0x01 ) );
}
// Pack/Unpack two 128 bit vectors into/from one 256 bit vector
// usefulness tbd
// __m128i hi, __m128i lo, returns __m256i
#define mm256_pack_2x128( hi, lo ) \
_mm256_inserti128_si256( _mm256_castsi128_si256( lo ), hi, 1 ) \
// __m128i hi, __m128i lo, __m256i src
#define mm256_unpack_2x128( hi, lo, src ) \
lo = _mm256_castsi256_si128( src ); \
hi = _mm256_castsi256_si128( mm256_swap_128( src ) );
// hi = _mm256_extracti128_si256( src, 1 );
// Pseudo parallel AES
// Probably noticeably slower than using pure 128 bit vectors
inline __m256i mm256_aesenc_2x128( __m256i x, __m256i k )
{
__m128i hi, lo, khi, klo;
mm256_unpack_2x128( hi, lo, x );
mm256_unpack_2x128( khi, klo, k );
lo = _mm_aesenc_si128( lo, klo );
hi = _mm_aesenc_si128( hi, khi );
return mm256_pack_2x128( hi, lo );
}
inline __m256i mm256_aesenc_nokey_2x128( __m256i x )
{
__m128i hi, lo;
mm256_unpack_2x128( hi, lo, x );
lo = _mm_aesenc_si128( lo, mm_zero );
hi = _mm_aesenc_si128( hi, mm_zero );
return mm256_pack_2x128( hi, lo );
}
#endif // AVX2
// Paired functions for interleaving and deinterleaving data for vector
// processing.
// Size is specfied in bits regardless of vector size to avoid pointer
// arithmetic confusion with different size vectors and be consistent with
// the function's name.
//
// Each function has 2 implementations, an optimized version that uses
// vector indexing and a slower version that uses pointers. The optimized
// version can only be used with 64 bit elements and only supports sizes
// of 256, 512 or 640 bits, 32, 64, and 80 bytes respectively.
//
// NOTE: Contrary to GCC documentation accessing vector elements using array
// indexes only works with 64 bit elements.
// Interleaving and deinterleaving of vectors of 32 bit elements
// must use the slower implementations that don't use vector indexing.
//
// All data must be aligned to 256 bits for AVX2, or 128 bits for AVX.
// Interleave source args and deinterleave destination args are not required
// to be contiguous in memory but it's more efficient if they are.
// Interleave source agrs may be the same actual arg repeated.
// 640 bit deinterleaving 4x64 using 256 bit AVX2 requires the
// destination buffers be defined with padding up to 768 bits for overrun
// space. Although overrun space use is non destructive it should not overlay
// useful data and should be ignored by the caller.
// SSE2 AVX
// interleave 4 arrays of 32 bit elements for 128 bit processing
// bit_len must be 256, 512 or 640 bits.
inline void mm_interleave_4x32( void *dst, const void *src0, const void *src1,
const void *src2, const void *src3, int bit_len )
{
uint32_t *s0 = (uint32_t*)src0;
uint32_t *s1 = (uint32_t*)src1;
uint32_t *s2 = (uint32_t*)src2;
uint32_t *s3 = (uint32_t*)src3;
__m128i* d = (__m128i*)dst;
d[0] = _mm_set_epi32( s3[ 0], s2[ 0], s1[ 0], s0[ 0] );
d[1] = _mm_set_epi32( s3[ 1], s2[ 1], s1[ 1], s0[ 1] );
d[2] = _mm_set_epi32( s3[ 2], s2[ 2], s1[ 2], s0[ 2] );
d[3] = _mm_set_epi32( s3[ 3], s2[ 3], s1[ 3], s0[ 3] );
d[4] = _mm_set_epi32( s3[ 4], s2[ 4], s1[ 4], s0[ 4] );
d[5] = _mm_set_epi32( s3[ 5], s2[ 5], s1[ 5], s0[ 5] );
d[6] = _mm_set_epi32( s3[ 6], s2[ 6], s1[ 6], s0[ 6] );
d[7] = _mm_set_epi32( s3[ 7], s2[ 7], s1[ 7], s0[ 7] );
if ( bit_len <= 256 ) return;
d[ 8] = _mm_set_epi32( s3[ 8], s2[ 8], s1[ 8], s0[ 8] );
d[ 9] = _mm_set_epi32( s3[ 9], s2[ 9], s1[ 9], s0[ 9] );
d[10] = _mm_set_epi32( s3[10], s2[10], s1[10], s0[10] );
d[11] = _mm_set_epi32( s3[11], s2[11], s1[11], s0[11] );
d[12] = _mm_set_epi32( s3[12], s2[12], s1[12], s0[12] );
d[13] = _mm_set_epi32( s3[13], s2[13], s1[13], s0[13] );
d[14] = _mm_set_epi32( s3[14], s2[14], s1[14], s0[14] );
d[15] = _mm_set_epi32( s3[15], s2[15], s1[15], s0[15] );
if ( bit_len <= 512 ) return;
d[16] = _mm_set_epi32( s3[16], s2[16], s1[16], s0[16] );
d[17] = _mm_set_epi32( s3[17], s2[17], s1[17], s0[17] );
d[18] = _mm_set_epi32( s3[18], s2[18], s1[18], s0[18] );
d[19] = _mm_set_epi32( s3[19], s2[19], s1[19], s0[19] );
if ( bit_len <= 640 ) return;
d[20] = _mm_set_epi32( s3[20], s2[20], s1[20], s0[20] );
d[21] = _mm_set_epi32( s3[21], s2[21], s1[21], s0[21] );
d[22] = _mm_set_epi32( s3[22], s2[22], s1[22], s0[22] );
d[23] = _mm_set_epi32( s3[23], s2[23], s1[23], s0[23] );
d[24] = _mm_set_epi32( s3[24], s2[24], s1[24], s0[24] );
d[25] = _mm_set_epi32( s3[25], s2[25], s1[25], s0[25] );
d[26] = _mm_set_epi32( s3[26], s2[26], s1[26], s0[26] );
d[27] = _mm_set_epi32( s3[27], s2[27], s1[27], s0[27] );
d[28] = _mm_set_epi32( s3[28], s2[28], s1[28], s0[28] );
d[29] = _mm_set_epi32( s3[29], s2[29], s1[29], s0[29] );
d[30] = _mm_set_epi32( s3[30], s2[30], s1[30], s0[30] );
d[31] = _mm_set_epi32( s3[31], s2[31], s1[31], s0[31] );
// bit_len == 1024
}
// bit_len must be multiple of 32
inline void mm_interleave_4x32x( void *dst, void *src0, void *src1,
void *src2, void *src3, int bit_len )
{
uint32_t *d = (uint32_t*)dst;
uint32_t *s0 = (uint32_t*)src0;
uint32_t *s1 = (uint32_t*)src1;
uint32_t *s2 = (uint32_t*)src2;
uint32_t *s3 = (uint32_t*)src3;
for ( int i = 0; i < bit_len >> 5; i++, d += 4 )
{
*d = *(s0+i);
*(d+1) = *(s1+i);
*(d+2) = *(s2+i);
*(d+3) = *(s3+i);
}
}
inline void mm_deinterleave_4x32( void *dst0, void *dst1, void *dst2,
void *dst3, const void *src, int bit_len )
{
uint32_t *s = (uint32_t*)src;
__m128i* d0 = (__m128i*)dst0;
__m128i* d1 = (__m128i*)dst1;
__m128i* d2 = (__m128i*)dst2;
__m128i* d3 = (__m128i*)dst3;
d0[0] = _mm_set_epi32( s[12], s[ 8], s[ 4], s[ 0] );
d1[0] = _mm_set_epi32( s[13], s[ 9], s[ 5], s[ 1] );
d2[0] = _mm_set_epi32( s[14], s[10], s[ 6], s[ 2] );
d3[0] = _mm_set_epi32( s[15], s[11], s[ 7], s[ 3] );
d0[1] = _mm_set_epi32( s[28], s[24], s[20], s[16] );
d1[1] = _mm_set_epi32( s[29], s[25], s[21], s[17] );
d2[1] = _mm_set_epi32( s[30], s[26], s[22], s[18] );
d3[1] = _mm_set_epi32( s[31], s[27], s[23], s[19] );
if ( bit_len <= 256 ) return;
d0[2] = _mm_set_epi32( s[44], s[40], s[36], s[32] );
d1[2] = _mm_set_epi32( s[45], s[41], s[37], s[33] );
d2[2] = _mm_set_epi32( s[46], s[42], s[38], s[34] );
d3[2] = _mm_set_epi32( s[47], s[43], s[39], s[35] );
d0[3] = _mm_set_epi32( s[60], s[56], s[52], s[48] );
d1[3] = _mm_set_epi32( s[61], s[57], s[53], s[49] );
d2[3] = _mm_set_epi32( s[62], s[58], s[54], s[50] );
d3[3] = _mm_set_epi32( s[63], s[59], s[55], s[51] );
if ( bit_len <= 512 ) return;
d0[4] = _mm_set_epi32( s[76], s[72], s[68], s[64] );
d1[4] = _mm_set_epi32( s[77], s[73], s[69], s[65] );
d2[4] = _mm_set_epi32( s[78], s[74], s[70], s[66] );
d3[4] = _mm_set_epi32( s[79], s[75], s[71], s[67] );
if ( bit_len <= 640 ) return;
d0[5] = _mm_set_epi32( s[92], s[88], s[84], s[80] );
d1[5] = _mm_set_epi32( s[93], s[89], s[85], s[81] );
d2[5] = _mm_set_epi32( s[94], s[90], s[86], s[82] );
d3[5] = _mm_set_epi32( s[95], s[91], s[87], s[83] );
d0[6] = _mm_set_epi32( s[108], s[104], s[100], s[ 96] );
d1[6] = _mm_set_epi32( s[109], s[105], s[101], s[ 97] );
d2[6] = _mm_set_epi32( s[110], s[106], s[102], s[ 98] );
d3[6] = _mm_set_epi32( s[111], s[107], s[103], s[ 99] );
d0[7] = _mm_set_epi32( s[124], s[120], s[116], s[112] );
d1[7] = _mm_set_epi32( s[125], s[121], s[117], s[113] );
d2[7] = _mm_set_epi32( s[126], s[122], s[118], s[114] );
d3[7] = _mm_set_epi32( s[127], s[123], s[119], s[115] );
// bit_len == 1024
}
// deinterleave 4 arrays into individual buffers for scalarm processing
// bit_len must be multiple of 32
inline void mm_deinterleave_4x32x( void *dst0, void *dst1, void *dst2,
void *dst3, const void *src, int bit_len )
{
uint32_t *s = (uint32_t*)src;
uint32_t *d0 = (uint32_t*)dst0;
uint32_t *d1 = (uint32_t*)dst1;
uint32_t *d2 = (uint32_t*)dst2;
uint32_t *d3 = (uint32_t*)dst3;
for ( int i = 0; i < bit_len >> 5; i++, s += 4 )
{
*(d0+i) = *s;
*(d1+i) = *(s+1);
*(d2+i) = *(s+2);
*(d3+i) = *(s+3);
}
}
#if defined (__AVX2__)
// Interleave 4 source buffers containing 64 bit data into the destination
// buffer. Only bit_len 256, 512, 640 & 1024 are supported.
inline void mm256_interleave_4x64( void *dst, const void *src0,
const void *src1, const void *src2, const void *src3, int bit_len )
{
__m256i* d = (__m256i*)dst;
uint64_t *s0 = (uint64_t*)src0;
uint64_t *s1 = (uint64_t*)src1;
uint64_t *s2 = (uint64_t*)src2;
uint64_t *s3 = (uint64_t*)src3;
d[0] = _mm256_set_epi64x( s3[0], s2[0], s1[0], s0[0] );
d[1] = _mm256_set_epi64x( s3[1], s2[1], s1[1], s0[1] );
d[2] = _mm256_set_epi64x( s3[2], s2[2], s1[2], s0[2] );
d[3] = _mm256_set_epi64x( s3[3], s2[3], s1[3], s0[3] );
if ( bit_len <= 256 ) return;
d[4] = _mm256_set_epi64x( s3[4], s2[4], s1[4], s0[4] );
d[5] = _mm256_set_epi64x( s3[5], s2[5], s1[5], s0[5] );
d[6] = _mm256_set_epi64x( s3[6], s2[6], s1[6], s0[6] );
d[7] = _mm256_set_epi64x( s3[7], s2[7], s1[7], s0[7] );
if ( bit_len <= 512 ) return;
d[8] = _mm256_set_epi64x( s3[8], s2[8], s1[8], s0[8] );
d[9] = _mm256_set_epi64x( s3[9], s2[9], s1[9], s0[9] );
if ( bit_len <= 640 ) return;
d[10] = _mm256_set_epi64x( s3[10], s2[10], s1[10], s0[10] );
d[11] = _mm256_set_epi64x( s3[11], s2[11], s1[11], s0[11] );
d[12] = _mm256_set_epi64x( s3[12], s2[12], s1[12], s0[12] );
d[13] = _mm256_set_epi64x( s3[13], s2[13], s1[13], s0[13] );
d[14] = _mm256_set_epi64x( s3[14], s2[14], s1[14], s0[14] );
d[15] = _mm256_set_epi64x( s3[15], s2[15], s1[15], s0[15] );
// bit_len == 1024
}
// Slower version
// bit_len must be multiple of 64
inline void mm256_interleave_4x64x( void *dst, void *src0, void *src1,
void *src2, void *src3, int bit_len )
{
uint64_t *d = (uint64_t*)dst;
uint64_t *s0 = (uint64_t*)src0;
uint64_t *s1 = (uint64_t*)src1;
uint64_t *s2 = (uint64_t*)src2;
uint64_t *s3 = (uint64_t*)src3;
for ( int i = 0; i < bit_len>>6; i++, d += 4 )
{
*d = *(s0+i);
*(d+1) = *(s1+i);
*(d+2) = *(s2+i);
*(d+3) = *(s3+i);
}
}
// Deinterleave 4 buffers of 64 bit data from the source buffer.
// bit_len must be 256, 512, 640 or 1024 bits.
// Requires overrun padding for 640 bit len.
inline void mm256_deinterleave_4x64( void *dst0, void *dst1, void *dst2,
void *dst3, const void *src, int bit_len )
{
__m256i* d0 = (__m256i*)dst0;
__m256i* d1 = (__m256i*)dst1;
__m256i* d2 = (__m256i*)dst2;
__m256i* d3 = (__m256i*)dst3;
uint64_t* s = (uint64_t*)src;
d0[0] = _mm256_set_epi64x( s[12], s[ 8], s[ 4], s[ 0] );
d1[0] = _mm256_set_epi64x( s[13], s[ 9], s[ 5], s[ 1] );
d2[0] = _mm256_set_epi64x( s[14], s[10], s[ 6], s[ 2] );
d3[0] = _mm256_set_epi64x( s[15], s[11], s[ 7], s[ 3] );
if ( bit_len <= 256 ) return;
d0[1] = _mm256_set_epi64x( s[28], s[24], s[20], s[16] );
d1[1] = _mm256_set_epi64x( s[29], s[25], s[21], s[17] );
d2[1] = _mm256_set_epi64x( s[30], s[26], s[22], s[18] );
d3[1] = _mm256_set_epi64x( s[31], s[27], s[23], s[19] );
if ( bit_len <= 512 ) return;
if ( bit_len <= 640 )
{
// null change to overrun area
d0[2] = _mm256_set_epi64x( d0[2][3], d0[2][2], s[36], s[32] );
d1[2] = _mm256_set_epi64x( d1[2][3], d1[2][2], s[37], s[33] );
d2[2] = _mm256_set_epi64x( d2[2][3], d2[2][2], s[38], s[34] );
d3[2] = _mm256_set_epi64x( d3[2][3], d3[2][2], s[39], s[35] );
return;
}
d0[2] = _mm256_set_epi64x( s[44], s[40], s[36], s[32] );
d1[2] = _mm256_set_epi64x( s[45], s[41], s[37], s[33] );
d2[2] = _mm256_set_epi64x( s[46], s[42], s[38], s[34] );
d3[2] = _mm256_set_epi64x( s[47], s[43], s[39], s[35] );
d0[3] = _mm256_set_epi64x( s[60], s[56], s[52], s[48] );
d1[3] = _mm256_set_epi64x( s[61], s[57], s[53], s[49] );
d2[3] = _mm256_set_epi64x( s[62], s[58], s[54], s[50] );
d3[3] = _mm256_set_epi64x( s[63], s[59], s[55], s[51] );
// bit_len == 1024
}
// Slower version
// bit_len must be multiple 0f 64
inline void mm256_deinterleave_4x64x( void *dst0, void *dst1, void *dst2,
void *dst3, void *src, int bit_len )
{
uint64_t *s = (uint64_t*)src;
uint64_t *d0 = (uint64_t*)dst0;
uint64_t *d1 = (uint64_t*)dst1;
uint64_t *d2 = (uint64_t*)dst2;
uint64_t *d3 = (uint64_t*)dst3;
for ( int i = 0; i < bit_len>>6; i++, s += 4 )
{
*(d0+i) = *s;
*(d1+i) = *(s+1);
*(d2+i) = *(s+2);
*(d3+i) = *(s+3);
}
}
// Interleave 8 source buffers containing 32 bit data into the destination
// vector
inline void mm256_interleave_8x32( void *dst, const void *src0,
const void *src1, const void *src2, const void *src3, const void *src4,
const void *src5, const void *src6, const void *src7, int bit_len )
{
uint32_t *s0 = (uint32_t*)src0;
uint32_t *s1 = (uint32_t*)src1;
uint32_t *s2 = (uint32_t*)src2;
uint32_t *s3 = (uint32_t*)src3;
uint32_t *s4 = (uint32_t*)src4;
uint32_t *s5 = (uint32_t*)src5;
uint32_t *s6 = (uint32_t*)src6;
uint32_t *s7 = (uint32_t*)src7;
__m256i *d = (__m256i*)dst;
d[ 0] = _mm256_set_epi32( s7[0], s6[0], s5[0], s4[0],
s3[0], s2[0], s1[0], s0[0] );
d[ 1] = _mm256_set_epi32( s7[1], s6[1], s5[1], s4[1],
s3[1], s2[1], s1[1], s0[1] );
d[ 2] = _mm256_set_epi32( s7[2], s6[2], s5[2], s4[2],
s3[2], s2[2], s1[2], s0[2] );
d[ 3] = _mm256_set_epi32( s7[3], s6[3], s5[3], s4[3],
s3[3], s2[3], s1[3], s0[3] );
d[ 4] = _mm256_set_epi32( s7[4], s6[4], s5[4], s4[4],
s3[4], s2[4], s1[4], s0[4] );
d[ 5] = _mm256_set_epi32( s7[5], s6[5], s5[5], s4[5],
s3[5], s2[5], s1[5], s0[5] );
d[ 6] = _mm256_set_epi32( s7[6], s6[6], s5[6], s4[6],
s3[6], s2[6], s1[6], s0[6] );
d[ 7] = _mm256_set_epi32( s7[7], s6[7], s5[7], s4[7],
s3[7], s2[7], s1[7], s0[7] );
if ( bit_len <= 256 ) return;
d[ 8] = _mm256_set_epi32( s7[ 8], s6[ 8], s5[ 8], s4[ 8],
s3[ 8], s2[ 8], s1[ 8], s0[ 8] );
d[ 9] = _mm256_set_epi32( s7[ 9], s6[ 9], s5[ 9], s4[ 9],
s3[ 9], s2[ 9], s1[ 9], s0[ 9] );
d[10] = _mm256_set_epi32( s7[10], s6[10], s5[10], s4[10],
s3[10], s2[10], s1[10], s0[10] );
d[11] = _mm256_set_epi32( s7[11], s6[11], s5[11], s4[11],
s3[11], s2[11], s1[11], s0[11] );
d[12] = _mm256_set_epi32( s7[12], s6[12], s5[12], s4[12],
s3[12], s2[12], s1[12], s0[12] );
d[13] = _mm256_set_epi32( s7[13], s6[13], s5[13], s4[13],
s3[13], s2[13], s1[13], s0[13] );
d[14] = _mm256_set_epi32( s7[14], s6[14], s5[14], s4[14],
s3[14], s2[14], s1[14], s0[14] );
d[15] = _mm256_set_epi32( s7[15], s6[15], s5[15], s4[15],
s3[15], s2[15], s1[15], s0[15] );
if ( bit_len <= 512 ) return;
d[16] = _mm256_set_epi32( s7[16], s6[16], s5[16], s4[16],
s3[16], s2[16], s1[16], s0[16] );
d[17] = _mm256_set_epi32( s7[17], s6[17], s5[17], s4[17],
s3[17], s2[17], s1[17], s0[17] );
d[18] = _mm256_set_epi32( s7[18], s6[18], s5[18], s4[18],