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
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
|
#! /usr/bin/env perl
# Copyright 2010-2016 The OpenSSL Project Authors. All Rights Reserved.
#
# Licensed under the OpenSSL license (the "License"). You may not use
# this file except in compliance with the License. You can obtain a copy
# in the file LICENSE in the source distribution or at
# https://www.openssl.org/source/license.html
#
# ====================================================================
# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
# project. The module is, however, dual licensed under OpenSSL and
# CRYPTOGAMS licenses depending on where you obtain it. For further
# details see http://www.openssl.org/~appro/cryptogams/.
# ====================================================================
#
# March, June 2010
#
# The module implements "4-bit" GCM GHASH function and underlying
# single multiplication operation in GF(2^128). "4-bit" means that
# it uses 256 bytes per-key table [+128 bytes shared table]. GHASH
# function features so called "528B" variant utilizing additional
# 256+16 bytes of per-key storage [+512 bytes shared table].
# Performance results are for this streamed GHASH subroutine and are
# expressed in cycles per processed byte, less is better:
#
# gcc 3.4.x(*) assembler
#
# P4 28.6 14.0 +100%
# Opteron 19.3 7.7 +150%
# Core2 17.8 8.1(**) +120%
# Atom 31.6 16.8 +88%
# VIA Nano 21.8 10.1 +115%
#
# (*) comparison is not completely fair, because C results are
# for vanilla "256B" implementation, while assembler results
# are for "528B";-)
# (**) it's mystery [to me] why Core2 result is not same as for
# Opteron;
# May 2010
#
# Add PCLMULQDQ version performing at 2.02 cycles per processed byte.
# See ghash-x86.pl for background information and details about coding
# techniques.
#
# Special thanks to David Woodhouse for providing access to a
# Westmere-based system on behalf of Intel Open Source Technology Centre.
# December 2012
#
# Overhaul: aggregate Karatsuba post-processing, improve ILP in
# reduction_alg9, increase reduction aggregate factor to 4x. As for
# the latter. ghash-x86.pl discusses that it makes lesser sense to
# increase aggregate factor. Then why increase here? Critical path
# consists of 3 independent pclmulqdq instructions, Karatsuba post-
# processing and reduction. "On top" of this we lay down aggregated
# multiplication operations, triplets of independent pclmulqdq's. As
# issue rate for pclmulqdq is limited, it makes lesser sense to
# aggregate more multiplications than it takes to perform remaining
# non-multiplication operations. 2x is near-optimal coefficient for
# contemporary Intel CPUs (therefore modest improvement coefficient),
# but not for Bulldozer. Latter is because logical SIMD operations
# are twice as slow in comparison to Intel, so that critical path is
# longer. A CPU with higher pclmulqdq issue rate would also benefit
# from higher aggregate factor...
#
# Westmere 1.78(+13%)
# Sandy Bridge 1.80(+8%)
# Ivy Bridge 1.80(+7%)
# Haswell 0.55(+93%) (if system doesn't support AVX)
# Broadwell 0.45(+110%)(if system doesn't support AVX)
# Skylake 0.44(+110%)(if system doesn't support AVX)
# Bulldozer 1.49(+27%)
# Silvermont 2.88(+13%)
# Knights L 2.12(-) (if system doesn't support AVX)
# Goldmont 1.08(+24%)
# March 2013
#
# ... 8x aggregate factor AVX code path is using reduction algorithm
# suggested by Shay Gueron[1]. Even though contemporary AVX-capable
# CPUs such as Sandy and Ivy Bridge can execute it, the code performs
# sub-optimally in comparison to above mentioned version. But thanks
# to Ilya Albrekht and Max Locktyukhin of Intel Corp. we knew that
# it performs in 0.41 cycles per byte on Haswell processor, in
# 0.29 on Broadwell, and in 0.36 on Skylake.
#
# Knights Landing achieves 1.09 cpb.
#
# [1] http://rt.openssl.org/Ticket/Display.html?id=2900&user=guest&pass=guest
# This file was patched in BoringSSL to remove the variable-time 4-bit
# implementation.
$flavour = shift;
$output = shift;
if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or
( $xlate="${dir}../../../perlasm/x86_64-xlate.pl" and -f $xlate) or
die "can't locate x86_64-xlate.pl";
# See the notes about |$avx| in aesni-gcm-x86_64.pl; otherwise tags will be
# computed incorrectly.
#
# In upstream, this is controlled by shelling out to the compiler to check
# versions, but BoringSSL is intended to be used with pre-generated perlasm
# output, so this isn't useful anyway.
$avx = 1;
open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\"";
*STDOUT=*OUT;
$do4xaggr=1;
$code=<<___;
.text
___
######################################################################
# PCLMULQDQ version.
@_4args=$win64? ("%rcx","%rdx","%r8", "%r9") : # Win64 order
("%rdi","%rsi","%rdx","%rcx"); # Unix order
($Xi,$Xhi)=("%xmm0","%xmm1"); $Hkey="%xmm2";
($T1,$T2,$T3)=("%xmm3","%xmm4","%xmm5");
sub clmul64x64_T2 { # minimal register pressure
my ($Xhi,$Xi,$Hkey,$HK)=@_;
if (!defined($HK)) { $HK = $T2;
$code.=<<___;
movdqa $Xi,$Xhi #
pshufd \$0b01001110,$Xi,$T1
pshufd \$0b01001110,$Hkey,$T2
pxor $Xi,$T1 #
pxor $Hkey,$T2
___
} else {
$code.=<<___;
movdqa $Xi,$Xhi #
pshufd \$0b01001110,$Xi,$T1
pxor $Xi,$T1 #
___
}
$code.=<<___;
pclmulqdq \$0x00,$Hkey,$Xi #######
pclmulqdq \$0x11,$Hkey,$Xhi #######
pclmulqdq \$0x00,$HK,$T1 #######
pxor $Xi,$T1 #
pxor $Xhi,$T1 #
movdqa $T1,$T2 #
psrldq \$8,$T1
pslldq \$8,$T2 #
pxor $T1,$Xhi
pxor $T2,$Xi #
___
}
sub reduction_alg9 { # 17/11 times faster than Intel version
my ($Xhi,$Xi) = @_;
$code.=<<___;
# 1st phase
movdqa $Xi,$T2 #
movdqa $Xi,$T1
psllq \$5,$Xi
pxor $Xi,$T1 #
psllq \$1,$Xi
pxor $T1,$Xi #
psllq \$57,$Xi #
movdqa $Xi,$T1 #
pslldq \$8,$Xi
psrldq \$8,$T1 #
pxor $T2,$Xi
pxor $T1,$Xhi #
# 2nd phase
movdqa $Xi,$T2
psrlq \$1,$Xi
pxor $T2,$Xhi #
pxor $Xi,$T2
psrlq \$5,$Xi
pxor $T2,$Xi #
psrlq \$1,$Xi #
pxor $Xhi,$Xi #
___
}
{ my ($Htbl,$Xip)=@_4args;
my $HK="%xmm6";
$code.=<<___;
.globl gcm_init_clmul
.type gcm_init_clmul,\@abi-omnipotent
.align 16
gcm_init_clmul:
.cfi_startproc
.seh_startproc
_CET_ENDBR
.L_init_clmul:
___
$code.=<<___ if ($win64);
sub \$0x18,%rsp
.seh_allocstack 0x18
movaps %xmm6,(%rsp)
.seh_savexmm128 %xmm6, 0
___
$code.=<<___;
movdqu ($Xip),$Hkey
pshufd \$0b01001110,$Hkey,$Hkey # dword swap
# <<1 twist
pshufd \$0b11111111,$Hkey,$T2 # broadcast uppermost dword
movdqa $Hkey,$T1
psllq \$1,$Hkey
pxor $T3,$T3 #
psrlq \$63,$T1
pcmpgtd $T2,$T3 # broadcast carry bit
pslldq \$8,$T1
por $T1,$Hkey # H<<=1
# magic reduction
pand .L0x1c2_polynomial(%rip),$T3
pxor $T3,$Hkey # if(carry) H^=0x1c2_polynomial
# calculate H^2
pshufd \$0b01001110,$Hkey,$HK
movdqa $Hkey,$Xi
pxor $Hkey,$HK
___
&clmul64x64_T2 ($Xhi,$Xi,$Hkey,$HK);
&reduction_alg9 ($Xhi,$Xi);
$code.=<<___;
pshufd \$0b01001110,$Hkey,$T1
pshufd \$0b01001110,$Xi,$T2
pxor $Hkey,$T1 # Karatsuba pre-processing
movdqu $Hkey,0x00($Htbl) # save H
pxor $Xi,$T2 # Karatsuba pre-processing
movdqu $Xi,0x10($Htbl) # save H^2
palignr \$8,$T1,$T2 # low part is H.lo^H.hi...
movdqu $T2,0x20($Htbl) # save Karatsuba "salt"
___
if ($do4xaggr) {
&clmul64x64_T2 ($Xhi,$Xi,$Hkey,$HK); # H^3
&reduction_alg9 ($Xhi,$Xi);
$code.=<<___;
movdqa $Xi,$T3
___
&clmul64x64_T2 ($Xhi,$Xi,$Hkey,$HK); # H^4
&reduction_alg9 ($Xhi,$Xi);
$code.=<<___;
pshufd \$0b01001110,$T3,$T1
pshufd \$0b01001110,$Xi,$T2
pxor $T3,$T1 # Karatsuba pre-processing
movdqu $T3,0x30($Htbl) # save H^3
pxor $Xi,$T2 # Karatsuba pre-processing
movdqu $Xi,0x40($Htbl) # save H^4
palignr \$8,$T1,$T2 # low part is H^3.lo^H^3.hi...
movdqu $T2,0x50($Htbl) # save Karatsuba "salt"
___
}
$code.=<<___ if ($win64);
movaps (%rsp),%xmm6
lea 0x18(%rsp),%rsp
___
$code.=<<___;
ret
.cfi_endproc
.seh_endproc
.size gcm_init_clmul,.-gcm_init_clmul
___
}
{ my ($Xip,$Htbl)=@_4args;
$code.=<<___;
.globl gcm_gmult_clmul
.type gcm_gmult_clmul,\@abi-omnipotent
.align 16
gcm_gmult_clmul:
.cfi_startproc
_CET_ENDBR
.L_gmult_clmul:
movdqu ($Xip),$Xi
movdqa .Lbswap_mask(%rip),$T3
movdqu ($Htbl),$Hkey
movdqu 0x20($Htbl),$T2
pshufb $T3,$Xi
___
&clmul64x64_T2 ($Xhi,$Xi,$Hkey,$T2);
$code.=<<___ if (0 || (&reduction_alg9($Xhi,$Xi)&&0));
# experimental alternative. special thing about is that there
# no dependency between the two multiplications...
mov \$`0xE1<<1`,%eax
mov \$0xA040608020C0E000,%r10 # ((7..0)·0xE0)&0xff
mov \$0x07,%r11d
movq %rax,$T1
movq %r10,$T2
movq %r11,$T3 # borrow $T3
pand $Xi,$T3
pshufb $T3,$T2 # ($Xi&7)·0xE0
movq %rax,$T3
pclmulqdq \$0x00,$Xi,$T1 # ·(0xE1<<1)
pxor $Xi,$T2
pslldq \$15,$T2
paddd $T2,$T2 # <<(64+56+1)
pxor $T2,$Xi
pclmulqdq \$0x01,$T3,$Xi
movdqa .Lbswap_mask(%rip),$T3 # reload $T3
psrldq \$1,$T1
pxor $T1,$Xhi
pslldq \$7,$Xi
pxor $Xhi,$Xi
___
$code.=<<___;
pshufb $T3,$Xi
movdqu $Xi,($Xip)
ret
.cfi_endproc
.size gcm_gmult_clmul,.-gcm_gmult_clmul
___
}
{ my ($Xip,$Htbl,$inp,$len)=@_4args;
my ($Xln,$Xmn,$Xhn,$Hkey2,$HK) = map("%xmm$_",(3..7));
my ($T1,$T2,$T3)=map("%xmm$_",(8..10));
$code.=<<___;
.globl gcm_ghash_clmul
.type gcm_ghash_clmul,\@abi-omnipotent
.align 32
gcm_ghash_clmul:
.cfi_startproc
.seh_startproc
_CET_ENDBR
.L_ghash_clmul:
___
$code.=<<___ if ($win64);
lea -0x88(%rsp),%rax
lea -0x20(%rax),%rsp
.seh_allocstack 0x20+0x88
movaps %xmm6,-0x20(%rax)
.seh_savexmm128 %xmm6, 0x20-0x20
movaps %xmm7,-0x10(%rax)
.seh_savexmm128 %xmm7, 0x20-0x10
movaps %xmm8,0(%rax)
.seh_savexmm128 %xmm8, 0x20+0
movaps %xmm9,0x10(%rax)
.seh_savexmm128 %xmm9, 0x20+0x10
movaps %xmm10,0x20(%rax)
.seh_savexmm128 %xmm10, 0x20+0x20
movaps %xmm11,0x30(%rax)
.seh_savexmm128 %xmm11, 0x20+0x30
movaps %xmm12,0x40(%rax)
.seh_savexmm128 %xmm12, 0x20+0x40
movaps %xmm13,0x50(%rax)
.seh_savexmm128 %xmm13, 0x20+0x50
movaps %xmm14,0x60(%rax)
.seh_savexmm128 %xmm14, 0x20+0x60
movaps %xmm15,0x70(%rax)
.seh_savexmm128 %xmm15, 0x20+0x70
___
$code.=<<___;
movdqa .Lbswap_mask(%rip),$T3
movdqu ($Xip),$Xi
movdqu ($Htbl),$Hkey
movdqu 0x20($Htbl),$HK
pshufb $T3,$Xi
sub \$0x10,$len
jz .Lodd_tail
movdqu 0x10($Htbl),$Hkey2
___
if ($do4xaggr) {
my ($Xl,$Xm,$Xh,$Hkey3,$Hkey4)=map("%xmm$_",(11..15));
$code.=<<___;
cmp \$0x30,$len
jb .Lskip4x
sub \$0x30,$len
mov \$0xA040608020C0E000,%rax # ((7..0)·0xE0)&0xff
movdqu 0x30($Htbl),$Hkey3
movdqu 0x40($Htbl),$Hkey4
#######
# Xi+4 =[(H*Ii+3) + (H^2*Ii+2) + (H^3*Ii+1) + H^4*(Ii+Xi)] mod P
#
movdqu 0x30($inp),$Xln
movdqu 0x20($inp),$Xl
pshufb $T3,$Xln
pshufb $T3,$Xl
movdqa $Xln,$Xhn
pshufd \$0b01001110,$Xln,$Xmn
pxor $Xln,$Xmn
pclmulqdq \$0x00,$Hkey,$Xln
pclmulqdq \$0x11,$Hkey,$Xhn
pclmulqdq \$0x00,$HK,$Xmn
movdqa $Xl,$Xh
pshufd \$0b01001110,$Xl,$Xm
pxor $Xl,$Xm
pclmulqdq \$0x00,$Hkey2,$Xl
pclmulqdq \$0x11,$Hkey2,$Xh
pclmulqdq \$0x10,$HK,$Xm
xorps $Xl,$Xln
xorps $Xh,$Xhn
movups 0x50($Htbl),$HK
xorps $Xm,$Xmn
movdqu 0x10($inp),$Xl
movdqu 0($inp),$T1
pshufb $T3,$Xl
pshufb $T3,$T1
movdqa $Xl,$Xh
pshufd \$0b01001110,$Xl,$Xm
pxor $T1,$Xi
pxor $Xl,$Xm
pclmulqdq \$0x00,$Hkey3,$Xl
movdqa $Xi,$Xhi
pshufd \$0b01001110,$Xi,$T1
pxor $Xi,$T1
pclmulqdq \$0x11,$Hkey3,$Xh
pclmulqdq \$0x00,$HK,$Xm
xorps $Xl,$Xln
xorps $Xh,$Xhn
lea 0x40($inp),$inp
sub \$0x40,$len
jc .Ltail4x
jmp .Lmod4_loop
.align 32
.Lmod4_loop:
pclmulqdq \$0x00,$Hkey4,$Xi
xorps $Xm,$Xmn
movdqu 0x30($inp),$Xl
pshufb $T3,$Xl
pclmulqdq \$0x11,$Hkey4,$Xhi
xorps $Xln,$Xi
movdqu 0x20($inp),$Xln
movdqa $Xl,$Xh
pclmulqdq \$0x10,$HK,$T1
pshufd \$0b01001110,$Xl,$Xm
xorps $Xhn,$Xhi
pxor $Xl,$Xm
pshufb $T3,$Xln
movups 0x20($Htbl),$HK
xorps $Xmn,$T1
pclmulqdq \$0x00,$Hkey,$Xl
pshufd \$0b01001110,$Xln,$Xmn
pxor $Xi,$T1 # aggregated Karatsuba post-processing
movdqa $Xln,$Xhn
pxor $Xhi,$T1 #
pxor $Xln,$Xmn
movdqa $T1,$T2 #
pclmulqdq \$0x11,$Hkey,$Xh
pslldq \$8,$T1
psrldq \$8,$T2 #
pxor $T1,$Xi
movdqa .L7_mask(%rip),$T1
pxor $T2,$Xhi #
movq %rax,$T2
pand $Xi,$T1 # 1st phase
pshufb $T1,$T2 #
pxor $Xi,$T2 #
pclmulqdq \$0x00,$HK,$Xm
psllq \$57,$T2 #
movdqa $T2,$T1 #
pslldq \$8,$T2
pclmulqdq \$0x00,$Hkey2,$Xln
psrldq \$8,$T1 #
pxor $T2,$Xi
pxor $T1,$Xhi #
movdqu 0($inp),$T1
movdqa $Xi,$T2 # 2nd phase
psrlq \$1,$Xi
pclmulqdq \$0x11,$Hkey2,$Xhn
xorps $Xl,$Xln
movdqu 0x10($inp),$Xl
pshufb $T3,$Xl
pclmulqdq \$0x10,$HK,$Xmn
xorps $Xh,$Xhn
movups 0x50($Htbl),$HK
pshufb $T3,$T1
pxor $T2,$Xhi #
pxor $Xi,$T2
psrlq \$5,$Xi
movdqa $Xl,$Xh
pxor $Xm,$Xmn
pshufd \$0b01001110,$Xl,$Xm
pxor $T2,$Xi #
pxor $T1,$Xhi
pxor $Xl,$Xm
pclmulqdq \$0x00,$Hkey3,$Xl
psrlq \$1,$Xi #
pxor $Xhi,$Xi #
movdqa $Xi,$Xhi
pclmulqdq \$0x11,$Hkey3,$Xh
xorps $Xl,$Xln
pshufd \$0b01001110,$Xi,$T1
pxor $Xi,$T1
pclmulqdq \$0x00,$HK,$Xm
xorps $Xh,$Xhn
lea 0x40($inp),$inp
sub \$0x40,$len
jnc .Lmod4_loop
.Ltail4x:
pclmulqdq \$0x00,$Hkey4,$Xi
pclmulqdq \$0x11,$Hkey4,$Xhi
pclmulqdq \$0x10,$HK,$T1
xorps $Xm,$Xmn
xorps $Xln,$Xi
xorps $Xhn,$Xhi
pxor $Xi,$Xhi # aggregated Karatsuba post-processing
pxor $Xmn,$T1
pxor $Xhi,$T1 #
pxor $Xi,$Xhi
movdqa $T1,$T2 #
psrldq \$8,$T1
pslldq \$8,$T2 #
pxor $T1,$Xhi
pxor $T2,$Xi #
___
&reduction_alg9($Xhi,$Xi);
$code.=<<___;
add \$0x40,$len
jz .Ldone
movdqu 0x20($Htbl),$HK
sub \$0x10,$len
jz .Lodd_tail
.Lskip4x:
___
}
$code.=<<___;
#######
# Xi+2 =[H*(Ii+1 + Xi+1)] mod P =
# [(H*Ii+1) + (H*Xi+1)] mod P =
# [(H*Ii+1) + H^2*(Ii+Xi)] mod P
#
movdqu ($inp),$T1 # Ii
movdqu 16($inp),$Xln # Ii+1
pshufb $T3,$T1
pshufb $T3,$Xln
pxor $T1,$Xi # Ii+Xi
movdqa $Xln,$Xhn
pshufd \$0b01001110,$Xln,$Xmn
pxor $Xln,$Xmn
pclmulqdq \$0x00,$Hkey,$Xln
pclmulqdq \$0x11,$Hkey,$Xhn
pclmulqdq \$0x00,$HK,$Xmn
lea 32($inp),$inp # i+=2
nop
sub \$0x20,$len
jbe .Leven_tail
nop
jmp .Lmod_loop
.align 32
.Lmod_loop:
movdqa $Xi,$Xhi
movdqa $Xmn,$T1
pshufd \$0b01001110,$Xi,$Xmn #
pxor $Xi,$Xmn #
pclmulqdq \$0x00,$Hkey2,$Xi
pclmulqdq \$0x11,$Hkey2,$Xhi
pclmulqdq \$0x10,$HK,$Xmn
pxor $Xln,$Xi # (H*Ii+1) + H^2*(Ii+Xi)
pxor $Xhn,$Xhi
movdqu ($inp),$T2 # Ii
pxor $Xi,$T1 # aggregated Karatsuba post-processing
pshufb $T3,$T2
movdqu 16($inp),$Xln # Ii+1
pxor $Xhi,$T1
pxor $T2,$Xhi # "Ii+Xi", consume early
pxor $T1,$Xmn
pshufb $T3,$Xln
movdqa $Xmn,$T1 #
psrldq \$8,$T1
pslldq \$8,$Xmn #
pxor $T1,$Xhi
pxor $Xmn,$Xi #
movdqa $Xln,$Xhn #
movdqa $Xi,$T2 # 1st phase
movdqa $Xi,$T1
psllq \$5,$Xi
pxor $Xi,$T1 #
pclmulqdq \$0x00,$Hkey,$Xln #######
psllq \$1,$Xi
pxor $T1,$Xi #
psllq \$57,$Xi #
movdqa $Xi,$T1 #
pslldq \$8,$Xi
psrldq \$8,$T1 #
pxor $T2,$Xi
pshufd \$0b01001110,$Xhn,$Xmn
pxor $T1,$Xhi #
pxor $Xhn,$Xmn #
movdqa $Xi,$T2 # 2nd phase
psrlq \$1,$Xi
pclmulqdq \$0x11,$Hkey,$Xhn #######
pxor $T2,$Xhi #
pxor $Xi,$T2
psrlq \$5,$Xi
pxor $T2,$Xi #
lea 32($inp),$inp
psrlq \$1,$Xi #
pclmulqdq \$0x00,$HK,$Xmn #######
pxor $Xhi,$Xi #
sub \$0x20,$len
ja .Lmod_loop
.Leven_tail:
movdqa $Xi,$Xhi
movdqa $Xmn,$T1
pshufd \$0b01001110,$Xi,$Xmn #
pxor $Xi,$Xmn #
pclmulqdq \$0x00,$Hkey2,$Xi
pclmulqdq \$0x11,$Hkey2,$Xhi
pclmulqdq \$0x10,$HK,$Xmn
pxor $Xln,$Xi # (H*Ii+1) + H^2*(Ii+Xi)
pxor $Xhn,$Xhi
pxor $Xi,$T1
pxor $Xhi,$T1
pxor $T1,$Xmn
movdqa $Xmn,$T1 #
psrldq \$8,$T1
pslldq \$8,$Xmn #
pxor $T1,$Xhi
pxor $Xmn,$Xi #
___
&reduction_alg9 ($Xhi,$Xi);
$code.=<<___;
test $len,$len
jnz .Ldone
.Lodd_tail:
movdqu ($inp),$T1 # Ii
pshufb $T3,$T1
pxor $T1,$Xi # Ii+Xi
___
&clmul64x64_T2 ($Xhi,$Xi,$Hkey,$HK); # H*(Ii+Xi)
&reduction_alg9 ($Xhi,$Xi);
$code.=<<___;
.Ldone:
pshufb $T3,$Xi
movdqu $Xi,($Xip)
___
$code.=<<___ if ($win64);
movaps (%rsp),%xmm6
movaps 0x10(%rsp),%xmm7
movaps 0x20(%rsp),%xmm8
movaps 0x30(%rsp),%xmm9
movaps 0x40(%rsp),%xmm10
movaps 0x50(%rsp),%xmm11
movaps 0x60(%rsp),%xmm12
movaps 0x70(%rsp),%xmm13
movaps 0x80(%rsp),%xmm14
movaps 0x90(%rsp),%xmm15
lea 0xa8(%rsp),%rsp
___
$code.=<<___;
ret
.cfi_endproc
.seh_endproc
.size gcm_ghash_clmul,.-gcm_ghash_clmul
___
}
$code.=<<___;
.globl gcm_init_avx
.type gcm_init_avx,\@abi-omnipotent
.align 32
gcm_init_avx:
.cfi_startproc
_CET_ENDBR
___
if ($avx) {
my ($Htbl,$Xip)=@_4args;
my $HK="%xmm6";
$code.=<<___ if ($win64);
.seh_startproc
sub \$0x18,%rsp
.seh_allocstack 0x18
movaps %xmm6,(%rsp)
.seh_savexmm128 %xmm6, 0
___
$code.=<<___;
vzeroupper
vmovdqu ($Xip),$Hkey
vpshufd \$0b01001110,$Hkey,$Hkey # dword swap
# <<1 twist
vpshufd \$0b11111111,$Hkey,$T2 # broadcast uppermost dword
vpsrlq \$63,$Hkey,$T1
vpsllq \$1,$Hkey,$Hkey
vpxor $T3,$T3,$T3 #
vpcmpgtd $T2,$T3,$T3 # broadcast carry bit
vpslldq \$8,$T1,$T1
vpor $T1,$Hkey,$Hkey # H<<=1
# magic reduction
vpand .L0x1c2_polynomial(%rip),$T3,$T3
vpxor $T3,$Hkey,$Hkey # if(carry) H^=0x1c2_polynomial
vpunpckhqdq $Hkey,$Hkey,$HK
vmovdqa $Hkey,$Xi
vpxor $Hkey,$HK,$HK
mov \$4,%r10 # up to H^8
jmp .Linit_start_avx
___
sub clmul64x64_avx {
my ($Xhi,$Xi,$Hkey,$HK)=@_;
if (!defined($HK)) { $HK = $T2;
$code.=<<___;
vpunpckhqdq $Xi,$Xi,$T1
vpunpckhqdq $Hkey,$Hkey,$T2
vpxor $Xi,$T1,$T1 #
vpxor $Hkey,$T2,$T2
___
} else {
$code.=<<___;
vpunpckhqdq $Xi,$Xi,$T1
vpxor $Xi,$T1,$T1 #
___
}
$code.=<<___;
vpclmulqdq \$0x11,$Hkey,$Xi,$Xhi #######
vpclmulqdq \$0x00,$Hkey,$Xi,$Xi #######
vpclmulqdq \$0x00,$HK,$T1,$T1 #######
vpxor $Xi,$Xhi,$T2 #
vpxor $T2,$T1,$T1 #
vpslldq \$8,$T1,$T2 #
vpsrldq \$8,$T1,$T1
vpxor $T2,$Xi,$Xi #
vpxor $T1,$Xhi,$Xhi
___
}
sub reduction_avx {
my ($Xhi,$Xi) = @_;
$code.=<<___;
vpsllq \$57,$Xi,$T1 # 1st phase
vpsllq \$62,$Xi,$T2
vpxor $T1,$T2,$T2 #
vpsllq \$63,$Xi,$T1
vpxor $T1,$T2,$T2 #
vpslldq \$8,$T2,$T1 #
vpsrldq \$8,$T2,$T2
vpxor $T1,$Xi,$Xi #
vpxor $T2,$Xhi,$Xhi
vpsrlq \$1,$Xi,$T2 # 2nd phase
vpxor $Xi,$Xhi,$Xhi
vpxor $T2,$Xi,$Xi #
vpsrlq \$5,$T2,$T2
vpxor $T2,$Xi,$Xi #
vpsrlq \$1,$Xi,$Xi #
vpxor $Xhi,$Xi,$Xi #
___
}
$code.=<<___;
.align 32
.Linit_loop_avx:
vpalignr \$8,$T1,$T2,$T3 # low part is H.lo^H.hi...
vmovdqu $T3,-0x10($Htbl) # save Karatsuba "salt"
___
&clmul64x64_avx ($Xhi,$Xi,$Hkey,$HK); # calculate H^3,5,7
&reduction_avx ($Xhi,$Xi);
$code.=<<___;
.Linit_start_avx:
vmovdqa $Xi,$T3
___
&clmul64x64_avx ($Xhi,$Xi,$Hkey,$HK); # calculate H^2,4,6,8
&reduction_avx ($Xhi,$Xi);
$code.=<<___;
vpshufd \$0b01001110,$T3,$T1
vpshufd \$0b01001110,$Xi,$T2
vpxor $T3,$T1,$T1 # Karatsuba pre-processing
vmovdqu $T3,0x00($Htbl) # save H^1,3,5,7
vpxor $Xi,$T2,$T2 # Karatsuba pre-processing
vmovdqu $Xi,0x10($Htbl) # save H^2,4,6,8
lea 0x30($Htbl),$Htbl
sub \$1,%r10
jnz .Linit_loop_avx
vpalignr \$8,$T2,$T1,$T3 # last "salt" is flipped
vmovdqu $T3,-0x10($Htbl)
vzeroupper
___
$code.=<<___ if ($win64);
movaps (%rsp),%xmm6
lea 0x18(%rsp),%rsp
___
$code.=<<___;
ret
.seh_endproc
.cfi_endproc
.size gcm_init_avx,.-gcm_init_avx
___
} else {
$code.=<<___;
jmp .L_init_clmul
.size gcm_init_avx,.-gcm_init_avx
___
}
$code.=<<___;
.globl gcm_gmult_avx
.type gcm_gmult_avx,\@abi-omnipotent
.align 32
gcm_gmult_avx:
.cfi_startproc
_CET_ENDBR
jmp .L_gmult_clmul
.cfi_endproc
.size gcm_gmult_avx,.-gcm_gmult_avx
___
$code.=<<___;
.globl gcm_ghash_avx
.type gcm_ghash_avx,\@abi-omnipotent
.align 32
gcm_ghash_avx:
.cfi_startproc
_CET_ENDBR
___
if ($avx) {
my ($Xip,$Htbl,$inp,$len)=@_4args;
my ($Xlo,$Xhi,$Xmi,
$Zlo,$Zhi,$Zmi,
$Hkey,$HK,$T1,$T2,
$Xi,$Xo,$Tred,$bswap,$Ii,$Ij) = map("%xmm$_",(0..15));
$code.=<<___ if ($win64);
.seh_startproc
lea -0x88(%rsp),%rax
lea -0x20(%rax),%rsp
.seh_allocstack 0x20+0x88
movaps %xmm6,-0x20(%rax)
.seh_savexmm128 %xmm6, 0x20-0x20
movaps %xmm7,-0x10(%rax)
.seh_savexmm128 %xmm7, 0x20-0x10
movaps %xmm8,0(%rax)
.seh_savexmm128 %xmm8, 0x20+0
movaps %xmm9,0x10(%rax)
.seh_savexmm128 %xmm9, 0x20+0x10
movaps %xmm10,0x20(%rax)
.seh_savexmm128 %xmm10, 0x20+0x20
movaps %xmm11,0x30(%rax)
.seh_savexmm128 %xmm11, 0x20+0x30
movaps %xmm12,0x40(%rax)
.seh_savexmm128 %xmm12, 0x20+0x40
movaps %xmm13,0x50(%rax)
.seh_savexmm128 %xmm13, 0x20+0x50
movaps %xmm14,0x60(%rax)
.seh_savexmm128 %xmm14, 0x20+0x60
movaps %xmm15,0x70(%rax)
.seh_savexmm128 %xmm15, 0x20+0x70
___
$code.=<<___;
vzeroupper
vmovdqu ($Xip),$Xi # load $Xi
lea .L0x1c2_polynomial(%rip),%r10
lea 0x40($Htbl),$Htbl # size optimization
vmovdqu .Lbswap_mask(%rip),$bswap
vpshufb $bswap,$Xi,$Xi
cmp \$0x80,$len
jb .Lshort_avx
sub \$0x80,$len
vmovdqu 0x70($inp),$Ii # I[7]
vmovdqu 0x00-0x40($Htbl),$Hkey # $Hkey^1
vpshufb $bswap,$Ii,$Ii
vmovdqu 0x20-0x40($Htbl),$HK
vpunpckhqdq $Ii,$Ii,$T2
vmovdqu 0x60($inp),$Ij # I[6]
vpclmulqdq \$0x00,$Hkey,$Ii,$Xlo
vpxor $Ii,$T2,$T2
vpshufb $bswap,$Ij,$Ij
vpclmulqdq \$0x11,$Hkey,$Ii,$Xhi
vmovdqu 0x10-0x40($Htbl),$Hkey # $Hkey^2
vpunpckhqdq $Ij,$Ij,$T1
vmovdqu 0x50($inp),$Ii # I[5]
vpclmulqdq \$0x00,$HK,$T2,$Xmi
vpxor $Ij,$T1,$T1
vpshufb $bswap,$Ii,$Ii
vpclmulqdq \$0x00,$Hkey,$Ij,$Zlo
vpunpckhqdq $Ii,$Ii,$T2
vpclmulqdq \$0x11,$Hkey,$Ij,$Zhi
vmovdqu 0x30-0x40($Htbl),$Hkey # $Hkey^3
vpxor $Ii,$T2,$T2
vmovdqu 0x40($inp),$Ij # I[4]
vpclmulqdq \$0x10,$HK,$T1,$Zmi
vmovdqu 0x50-0x40($Htbl),$HK
vpshufb $bswap,$Ij,$Ij
vpxor $Xlo,$Zlo,$Zlo
vpclmulqdq \$0x00,$Hkey,$Ii,$Xlo
vpxor $Xhi,$Zhi,$Zhi
vpunpckhqdq $Ij,$Ij,$T1
vpclmulqdq \$0x11,$Hkey,$Ii,$Xhi
vmovdqu 0x40-0x40($Htbl),$Hkey # $Hkey^4
vpxor $Xmi,$Zmi,$Zmi
vpclmulqdq \$0x00,$HK,$T2,$Xmi
vpxor $Ij,$T1,$T1
vmovdqu 0x30($inp),$Ii # I[3]
vpxor $Zlo,$Xlo,$Xlo
vpclmulqdq \$0x00,$Hkey,$Ij,$Zlo
vpxor $Zhi,$Xhi,$Xhi
vpshufb $bswap,$Ii,$Ii
vpclmulqdq \$0x11,$Hkey,$Ij,$Zhi
vmovdqu 0x60-0x40($Htbl),$Hkey # $Hkey^5
vpxor $Zmi,$Xmi,$Xmi
vpunpckhqdq $Ii,$Ii,$T2
vpclmulqdq \$0x10,$HK,$T1,$Zmi
vmovdqu 0x80-0x40($Htbl),$HK
vpxor $Ii,$T2,$T2
vmovdqu 0x20($inp),$Ij # I[2]
vpxor $Xlo,$Zlo,$Zlo
vpclmulqdq \$0x00,$Hkey,$Ii,$Xlo
vpxor $Xhi,$Zhi,$Zhi
vpshufb $bswap,$Ij,$Ij
vpclmulqdq \$0x11,$Hkey,$Ii,$Xhi
vmovdqu 0x70-0x40($Htbl),$Hkey # $Hkey^6
vpxor $Xmi,$Zmi,$Zmi
vpunpckhqdq $Ij,$Ij,$T1
vpclmulqdq \$0x00,$HK,$T2,$Xmi
vpxor $Ij,$T1,$T1
vmovdqu 0x10($inp),$Ii # I[1]
vpxor $Zlo,$Xlo,$Xlo
vpclmulqdq \$0x00,$Hkey,$Ij,$Zlo
vpxor $Zhi,$Xhi,$Xhi
vpshufb $bswap,$Ii,$Ii
vpclmulqdq \$0x11,$Hkey,$Ij,$Zhi
vmovdqu 0x90-0x40($Htbl),$Hkey # $Hkey^7
vpxor $Zmi,$Xmi,$Xmi
vpunpckhqdq $Ii,$Ii,$T2
vpclmulqdq \$0x10,$HK,$T1,$Zmi
vmovdqu 0xb0-0x40($Htbl),$HK
vpxor $Ii,$T2,$T2
vmovdqu ($inp),$Ij # I[0]
vpxor $Xlo,$Zlo,$Zlo
vpclmulqdq \$0x00,$Hkey,$Ii,$Xlo
vpxor $Xhi,$Zhi,$Zhi
vpshufb $bswap,$Ij,$Ij
vpclmulqdq \$0x11,$Hkey,$Ii,$Xhi
vmovdqu 0xa0-0x40($Htbl),$Hkey # $Hkey^8
vpxor $Xmi,$Zmi,$Zmi
vpclmulqdq \$0x10,$HK,$T2,$Xmi
lea 0x80($inp),$inp
cmp \$0x80,$len
jb .Ltail_avx
vpxor $Xi,$Ij,$Ij # accumulate $Xi
sub \$0x80,$len
jmp .Loop8x_avx
.align 32
.Loop8x_avx:
vpunpckhqdq $Ij,$Ij,$T1
vmovdqu 0x70($inp),$Ii # I[7]
vpxor $Xlo,$Zlo,$Zlo
vpxor $Ij,$T1,$T1
vpclmulqdq \$0x00,$Hkey,$Ij,$Xi
vpshufb $bswap,$Ii,$Ii
vpxor $Xhi,$Zhi,$Zhi
vpclmulqdq \$0x11,$Hkey,$Ij,$Xo
vmovdqu 0x00-0x40($Htbl),$Hkey # $Hkey^1
vpunpckhqdq $Ii,$Ii,$T2
vpxor $Xmi,$Zmi,$Zmi
vpclmulqdq \$0x00,$HK,$T1,$Tred
vmovdqu 0x20-0x40($Htbl),$HK
vpxor $Ii,$T2,$T2
vmovdqu 0x60($inp),$Ij # I[6]
vpclmulqdq \$0x00,$Hkey,$Ii,$Xlo
vpxor $Zlo,$Xi,$Xi # collect result
vpshufb $bswap,$Ij,$Ij
vpclmulqdq \$0x11,$Hkey,$Ii,$Xhi
vxorps $Zhi,$Xo,$Xo
vmovdqu 0x10-0x40($Htbl),$Hkey # $Hkey^2
vpunpckhqdq $Ij,$Ij,$T1
vpclmulqdq \$0x00,$HK, $T2,$Xmi
vpxor $Zmi,$Tred,$Tred
vxorps $Ij,$T1,$T1
vmovdqu 0x50($inp),$Ii # I[5]
vpxor $Xi,$Tred,$Tred # aggregated Karatsuba post-processing
vpclmulqdq \$0x00,$Hkey,$Ij,$Zlo
vpxor $Xo,$Tred,$Tred
vpslldq \$8,$Tred,$T2
vpxor $Xlo,$Zlo,$Zlo
vpclmulqdq \$0x11,$Hkey,$Ij,$Zhi
vpsrldq \$8,$Tred,$Tred
vpxor $T2, $Xi, $Xi
vmovdqu 0x30-0x40($Htbl),$Hkey # $Hkey^3
vpshufb $bswap,$Ii,$Ii
vxorps $Tred,$Xo, $Xo
vpxor $Xhi,$Zhi,$Zhi
vpunpckhqdq $Ii,$Ii,$T2
vpclmulqdq \$0x10,$HK, $T1,$Zmi
vmovdqu 0x50-0x40($Htbl),$HK
vpxor $Ii,$T2,$T2
vpxor $Xmi,$Zmi,$Zmi
vmovdqu 0x40($inp),$Ij # I[4]
vpalignr \$8,$Xi,$Xi,$Tred # 1st phase
vpclmulqdq \$0x00,$Hkey,$Ii,$Xlo
vpshufb $bswap,$Ij,$Ij
vpxor $Zlo,$Xlo,$Xlo
vpclmulqdq \$0x11,$Hkey,$Ii,$Xhi
vmovdqu 0x40-0x40($Htbl),$Hkey # $Hkey^4
vpunpckhqdq $Ij,$Ij,$T1
vpxor $Zhi,$Xhi,$Xhi
vpclmulqdq \$0x00,$HK, $T2,$Xmi
vxorps $Ij,$T1,$T1
vpxor $Zmi,$Xmi,$Xmi
vmovdqu 0x30($inp),$Ii # I[3]
vpclmulqdq \$0x10,(%r10),$Xi,$Xi
vpclmulqdq \$0x00,$Hkey,$Ij,$Zlo
vpshufb $bswap,$Ii,$Ii
vpxor $Xlo,$Zlo,$Zlo
vpclmulqdq \$0x11,$Hkey,$Ij,$Zhi
vmovdqu 0x60-0x40($Htbl),$Hkey # $Hkey^5
vpunpckhqdq $Ii,$Ii,$T2
vpxor $Xhi,$Zhi,$Zhi
vpclmulqdq \$0x10,$HK, $T1,$Zmi
vmovdqu 0x80-0x40($Htbl),$HK
vpxor $Ii,$T2,$T2
vpxor $Xmi,$Zmi,$Zmi
vmovdqu 0x20($inp),$Ij # I[2]
vpclmulqdq \$0x00,$Hkey,$Ii,$Xlo
vpshufb $bswap,$Ij,$Ij
vpxor $Zlo,$Xlo,$Xlo
vpclmulqdq \$0x11,$Hkey,$Ii,$Xhi
vmovdqu 0x70-0x40($Htbl),$Hkey # $Hkey^6
vpunpckhqdq $Ij,$Ij,$T1
vpxor $Zhi,$Xhi,$Xhi
vpclmulqdq \$0x00,$HK, $T2,$Xmi
vpxor $Ij,$T1,$T1
vpxor $Zmi,$Xmi,$Xmi
vxorps $Tred,$Xi,$Xi
vmovdqu 0x10($inp),$Ii # I[1]
vpalignr \$8,$Xi,$Xi,$Tred # 2nd phase
vpclmulqdq \$0x00,$Hkey,$Ij,$Zlo
vpshufb $bswap,$Ii,$Ii
vpxor $Xlo,$Zlo,$Zlo
vpclmulqdq \$0x11,$Hkey,$Ij,$Zhi
vmovdqu 0x90-0x40($Htbl),$Hkey # $Hkey^7
vpclmulqdq \$0x10,(%r10),$Xi,$Xi
vxorps $Xo,$Tred,$Tred
vpunpckhqdq $Ii,$Ii,$T2
vpxor $Xhi,$Zhi,$Zhi
vpclmulqdq \$0x10,$HK, $T1,$Zmi
vmovdqu 0xb0-0x40($Htbl),$HK
vpxor $Ii,$T2,$T2
vpxor $Xmi,$Zmi,$Zmi
vmovdqu ($inp),$Ij # I[0]
vpclmulqdq \$0x00,$Hkey,$Ii,$Xlo
vpshufb $bswap,$Ij,$Ij
vpclmulqdq \$0x11,$Hkey,$Ii,$Xhi
vmovdqu 0xa0-0x40($Htbl),$Hkey # $Hkey^8
vpxor $Tred,$Ij,$Ij
vpclmulqdq \$0x10,$HK, $T2,$Xmi
vpxor $Xi,$Ij,$Ij # accumulate $Xi
lea 0x80($inp),$inp
sub \$0x80,$len
jnc .Loop8x_avx
add \$0x80,$len
jmp .Ltail_no_xor_avx
.align 32
.Lshort_avx:
vmovdqu -0x10($inp,$len),$Ii # very last word
lea ($inp,$len),$inp
vmovdqu 0x00-0x40($Htbl),$Hkey # $Hkey^1
vmovdqu 0x20-0x40($Htbl),$HK
vpshufb $bswap,$Ii,$Ij
vmovdqa $Xlo,$Zlo # subtle way to zero $Zlo,
vmovdqa $Xhi,$Zhi # $Zhi and
vmovdqa $Xmi,$Zmi # $Zmi
sub \$0x10,$len
jz .Ltail_avx
vpunpckhqdq $Ij,$Ij,$T1
vpxor $Xlo,$Zlo,$Zlo
vpclmulqdq \$0x00,$Hkey,$Ij,$Xlo
vpxor $Ij,$T1,$T1
vmovdqu -0x20($inp),$Ii
vpxor $Xhi,$Zhi,$Zhi
vpclmulqdq \$0x11,$Hkey,$Ij,$Xhi
vmovdqu 0x10-0x40($Htbl),$Hkey # $Hkey^2
vpshufb $bswap,$Ii,$Ij
vpxor $Xmi,$Zmi,$Zmi
vpclmulqdq \$0x00,$HK,$T1,$Xmi
vpsrldq \$8,$HK,$HK
sub \$0x10,$len
jz .Ltail_avx
vpunpckhqdq $Ij,$Ij,$T1
vpxor $Xlo,$Zlo,$Zlo
vpclmulqdq \$0x00,$Hkey,$Ij,$Xlo
vpxor $Ij,$T1,$T1
vmovdqu -0x30($inp),$Ii
vpxor $Xhi,$Zhi,$Zhi
vpclmulqdq \$0x11,$Hkey,$Ij,$Xhi
vmovdqu 0x30-0x40($Htbl),$Hkey # $Hkey^3
vpshufb $bswap,$Ii,$Ij
vpxor $Xmi,$Zmi,$Zmi
vpclmulqdq \$0x00,$HK,$T1,$Xmi
vmovdqu 0x50-0x40($Htbl),$HK
sub \$0x10,$len
jz .Ltail_avx
vpunpckhqdq $Ij,$Ij,$T1
vpxor $Xlo,$Zlo,$Zlo
vpclmulqdq \$0x00,$Hkey,$Ij,$Xlo
vpxor $Ij,$T1,$T1
vmovdqu -0x40($inp),$Ii
vpxor $Xhi,$Zhi,$Zhi
vpclmulqdq \$0x11,$Hkey,$Ij,$Xhi
vmovdqu 0x40-0x40($Htbl),$Hkey # $Hkey^4
vpshufb $bswap,$Ii,$Ij
vpxor $Xmi,$Zmi,$Zmi
vpclmulqdq \$0x00,$HK,$T1,$Xmi
vpsrldq \$8,$HK,$HK
sub \$0x10,$len
jz .Ltail_avx
vpunpckhqdq $Ij,$Ij,$T1
vpxor $Xlo,$Zlo,$Zlo
vpclmulqdq \$0x00,$Hkey,$Ij,$Xlo
vpxor $Ij,$T1,$T1
vmovdqu -0x50($inp),$Ii
vpxor $Xhi,$Zhi,$Zhi
vpclmulqdq \$0x11,$Hkey,$Ij,$Xhi
vmovdqu 0x60-0x40($Htbl),$Hkey # $Hkey^5
vpshufb $bswap,$Ii,$Ij
vpxor $Xmi,$Zmi,$Zmi
vpclmulqdq \$0x00,$HK,$T1,$Xmi
vmovdqu 0x80-0x40($Htbl),$HK
sub \$0x10,$len
jz .Ltail_avx
vpunpckhqdq $Ij,$Ij,$T1
vpxor $Xlo,$Zlo,$Zlo
vpclmulqdq \$0x00,$Hkey,$Ij,$Xlo
vpxor $Ij,$T1,$T1
vmovdqu -0x60($inp),$Ii
vpxor $Xhi,$Zhi,$Zhi
vpclmulqdq \$0x11,$Hkey,$Ij,$Xhi
vmovdqu 0x70-0x40($Htbl),$Hkey # $Hkey^6
vpshufb $bswap,$Ii,$Ij
vpxor $Xmi,$Zmi,$Zmi
vpclmulqdq \$0x00,$HK,$T1,$Xmi
vpsrldq \$8,$HK,$HK
sub \$0x10,$len
jz .Ltail_avx
vpunpckhqdq $Ij,$Ij,$T1
vpxor $Xlo,$Zlo,$Zlo
vpclmulqdq \$0x00,$Hkey,$Ij,$Xlo
vpxor $Ij,$T1,$T1
vmovdqu -0x70($inp),$Ii
vpxor $Xhi,$Zhi,$Zhi
vpclmulqdq \$0x11,$Hkey,$Ij,$Xhi
vmovdqu 0x90-0x40($Htbl),$Hkey # $Hkey^7
vpshufb $bswap,$Ii,$Ij
vpxor $Xmi,$Zmi,$Zmi
vpclmulqdq \$0x00,$HK,$T1,$Xmi
vmovq 0xb8-0x40($Htbl),$HK
sub \$0x10,$len
jmp .Ltail_avx
.align 32
.Ltail_avx:
vpxor $Xi,$Ij,$Ij # accumulate $Xi
.Ltail_no_xor_avx:
vpunpckhqdq $Ij,$Ij,$T1
vpxor $Xlo,$Zlo,$Zlo
vpclmulqdq \$0x00,$Hkey,$Ij,$Xlo
vpxor $Ij,$T1,$T1
vpxor $Xhi,$Zhi,$Zhi
vpclmulqdq \$0x11,$Hkey,$Ij,$Xhi
vpxor $Xmi,$Zmi,$Zmi
vpclmulqdq \$0x00,$HK,$T1,$Xmi
vmovdqu (%r10),$Tred
vpxor $Xlo,$Zlo,$Xi
vpxor $Xhi,$Zhi,$Xo
vpxor $Xmi,$Zmi,$Zmi
vpxor $Xi, $Zmi,$Zmi # aggregated Karatsuba post-processing
vpxor $Xo, $Zmi,$Zmi
vpslldq \$8, $Zmi,$T2
vpsrldq \$8, $Zmi,$Zmi
vpxor $T2, $Xi, $Xi
vpxor $Zmi,$Xo, $Xo
vpclmulqdq \$0x10,$Tred,$Xi,$T2 # 1st phase
vpalignr \$8,$Xi,$Xi,$Xi
vpxor $T2,$Xi,$Xi
vpclmulqdq \$0x10,$Tred,$Xi,$T2 # 2nd phase
vpalignr \$8,$Xi,$Xi,$Xi
vpxor $Xo,$Xi,$Xi
vpxor $T2,$Xi,$Xi
cmp \$0,$len
jne .Lshort_avx
vpshufb $bswap,$Xi,$Xi
vmovdqu $Xi,($Xip)
vzeroupper
___
$code.=<<___ if ($win64);
movaps (%rsp),%xmm6
movaps 0x10(%rsp),%xmm7
movaps 0x20(%rsp),%xmm8
movaps 0x30(%rsp),%xmm9
movaps 0x40(%rsp),%xmm10
movaps 0x50(%rsp),%xmm11
movaps 0x60(%rsp),%xmm12
movaps 0x70(%rsp),%xmm13
movaps 0x80(%rsp),%xmm14
movaps 0x90(%rsp),%xmm15
lea 0xa8(%rsp),%rsp
___
$code.=<<___;
ret
.cfi_endproc
.seh_endproc
.size gcm_ghash_avx,.-gcm_ghash_avx
___
} else {
$code.=<<___;
jmp .L_ghash_clmul
.size gcm_ghash_avx,.-gcm_ghash_avx
___
}
$code.=<<___;
.section .rodata
.align 64
.Lbswap_mask:
.byte 15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0
.L0x1c2_polynomial:
.byte 1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0xc2
.L7_mask:
.long 7,0,7,0
.align 64
.asciz "GHASH for x86_64, CRYPTOGAMS by <appro\@openssl.org>"
.align 64
.text
___
$code =~ s/\`([^\`]*)\`/eval($1)/gem;
print $code;
close STDOUT or die "error closing STDOUT: $!";
|