Newer
Older
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
#!/usr/bin/env perl
#
# Implemented as a Perl wrapper as we want to support several different
# architectures with single file. We pick up the target based on the
# file name we are asked to generate.
#
# It should be noted though that this perl code is nothing like
# <openssl>/crypto/perlasm/x86*. In this case perl is used pretty much
# as pre-processor to cover for platform differences in name decoration,
# linker tables, 32-/64-bit instruction sets...
#
# As you might know there're several PowerPC ABI in use. Most notably
# Linux and AIX use different 32-bit ABIs. Good news are that these ABIs
# are similar enough to implement leaf(!) functions, which would be ABI
# neutral. And that's what you find here: ABI neutral leaf functions.
# In case you wonder what that is...
#
# AIX performance
#
# MEASUREMENTS WITH cc ON a 200 MhZ PowerPC 604e.
#
# The following is the performance of 32-bit compiler
# generated code:
#
# OpenSSL 0.9.6c 21 dec 2001
# built on: Tue Jun 11 11:06:51 EDT 2002
# options:bn(64,32) ...
#compiler: cc -DTHREADS -DAIX -DB_ENDIAN -DBN_LLONG -O3
# sign verify sign/s verify/s
#rsa 512 bits 0.0098s 0.0009s 102.0 1170.6
#rsa 1024 bits 0.0507s 0.0026s 19.7 387.5
#rsa 2048 bits 0.3036s 0.0085s 3.3 117.1
#rsa 4096 bits 2.0040s 0.0299s 0.5 33.4
#dsa 512 bits 0.0087s 0.0106s 114.3 94.5
#dsa 1024 bits 0.0256s 0.0313s 39.0 32.0
#
# Same bechmark with this assembler code:
#
#rsa 512 bits 0.0056s 0.0005s 178.6 2049.2
#rsa 1024 bits 0.0283s 0.0015s 35.3 674.1
#rsa 2048 bits 0.1744s 0.0050s 5.7 201.2
#rsa 4096 bits 1.1644s 0.0179s 0.9 55.7
#dsa 512 bits 0.0052s 0.0062s 191.6 162.0
#dsa 1024 bits 0.0149s 0.0180s 67.0 55.5
#
# Number of operations increases by at almost 75%
#
# Here are performance numbers for 64-bit compiler
# generated code:
#
# OpenSSL 0.9.6g [engine] 9 Aug 2002
# built on: Fri Apr 18 16:59:20 EDT 2003
# options:bn(64,64) ...
# compiler: cc -DTHREADS -D_REENTRANT -q64 -DB_ENDIAN -O3
# sign verify sign/s verify/s
#rsa 512 bits 0.0028s 0.0003s 357.1 3844.4
#rsa 1024 bits 0.0148s 0.0008s 67.5 1239.7
#rsa 2048 bits 0.0963s 0.0028s 10.4 353.0
#rsa 4096 bits 0.6538s 0.0102s 1.5 98.1
#dsa 512 bits 0.0026s 0.0032s 382.5 313.7
#dsa 1024 bits 0.0081s 0.0099s 122.8 100.6
#
# Same benchmark with this assembler code:
#
#rsa 512 bits 0.0020s 0.0002s 510.4 6273.7
#rsa 1024 bits 0.0088s 0.0005s 114.1 2128.3
#rsa 2048 bits 0.0540s 0.0016s 18.5 622.5
#rsa 4096 bits 0.3700s 0.0058s 2.7 171.0
#dsa 512 bits 0.0016s 0.0020s 610.7 507.1
#dsa 1024 bits 0.0047s 0.0058s 212.5 173.2
#
# Again, performance increases by at about 75%
#
# Mac OS X, Apple G5 1.8GHz (Note this is 32 bit code)
# OpenSSL 0.9.7c 30 Sep 2003
#
# Original code.
#
#rsa 512 bits 0.0011s 0.0001s 906.1 11012.5
#rsa 1024 bits 0.0060s 0.0003s 166.6 3363.1
#rsa 2048 bits 0.0370s 0.0010s 27.1 982.4
#rsa 4096 bits 0.2426s 0.0036s 4.1 280.4
#dsa 512 bits 0.0010s 0.0012s 1038.1 841.5
#dsa 1024 bits 0.0030s 0.0037s 329.6 269.7
#dsa 2048 bits 0.0101s 0.0127s 98.9 78.6
#
# Same benchmark with this assembler code:
#
#rsa 512 bits 0.0007s 0.0001s 1416.2 16645.9
#rsa 1024 bits 0.0036s 0.0002s 274.4 5380.6
#rsa 2048 bits 0.0222s 0.0006s 45.1 1589.5
#rsa 4096 bits 0.1469s 0.0022s 6.8 449.6
#dsa 512 bits 0.0006s 0.0007s 1664.2 1376.2
#dsa 1024 bits 0.0018s 0.0023s 545.0 442.2
#dsa 2048 bits 0.0061s 0.0075s 163.5 132.8
#
# Performance increase of ~60%
#
# If you have comments or suggestions to improve code send
# me a note at schari@us.ibm.com
#
$flavour = shift;
if ($flavour =~ /32/) {
$BITS= 32;
$BNSZ= $BITS/8;
$ISA= "\"ppc\"";
$LD= "lwz"; # load
$LDU= "lwzu"; # load and update
$ST= "stw"; # store
$STU= "stwu"; # store and update
$UMULL= "mullw"; # unsigned multiply low
$UMULH= "mulhwu"; # unsigned multiply high
$UDIV= "divwu"; # unsigned divide
$UCMPI= "cmplwi"; # unsigned compare with immediate
$UCMP= "cmplw"; # unsigned compare
$CNTLZ= "cntlzw"; # count leading zeros
$SHL= "slw"; # shift left
$SHR= "srw"; # unsigned shift right
$SHRI= "srwi"; # unsigned shift right by immediate
$SHLI= "slwi"; # shift left by immediate
$CLRU= "clrlwi"; # clear upper bits
$INSR= "insrwi"; # insert right
$ROTL= "rotlwi"; # rotate left by immediate
$TR= "tw"; # conditional trap
} elsif ($flavour =~ /64/) {
$BITS= 64;
$BNSZ= $BITS/8;
$ISA= "\"ppc64\"";
# same as above, but 64-bit mnemonics...
$LD= "ld"; # load
$LDU= "ldu"; # load and update
$ST= "std"; # store
$STU= "stdu"; # store and update
$UMULL= "mulld"; # unsigned multiply low
$UMULH= "mulhdu"; # unsigned multiply high
$UDIV= "divdu"; # unsigned divide
$UCMPI= "cmpldi"; # unsigned compare with immediate
$UCMP= "cmpld"; # unsigned compare
$CNTLZ= "cntlzd"; # count leading zeros
$SHL= "sld"; # shift left
$SHR= "srd"; # unsigned shift right
$SHRI= "srdi"; # unsigned shift right by immediate
$SHLI= "sldi"; # shift left by immediate
$CLRU= "clrldi"; # clear upper bits
$INSR= "insrdi"; # insert right
$ROTL= "rotldi"; # rotate left by immediate
$TR= "td"; # conditional trap
} else { die "nonsense $flavour"; }
$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or
( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or
die "can't locate ppc-xlate.pl";
open STDOUT,"| $^X $xlate $flavour ".shift || die "can't call $xlate: $!";
$data=<<EOF;
#--------------------------------------------------------------------
#
#
#
#
# File: ppc32.s
#
# Created by: Suresh Chari
# IBM Thomas J. Watson Research Library
# Hawthorne, NY
#
#
# Description: Optimized assembly routines for OpenSSL crypto
# on the 32 bitPowerPC platform.
#
#
# Version History
#
# 2. Fixed bn_add,bn_sub and bn_div_words, added comments,
# cleaned up code. Also made a single version which can
# be used for both the AIX and Linux compilers. See NOTE
# below.
# 12/05/03 Suresh Chari
# (with lots of help from) Andy Polyakov
##
# 1. Initial version 10/20/02 Suresh Chari
#
#
# The following file works for the xlc,cc
# and gcc compilers.
#
# NOTE: To get the file to link correctly with the gcc compiler
# you have to change the names of the routines and remove
# the first .(dot) character. This should automatically
# be done in the build process.
#
# Hand optimized assembly code for the following routines
#
# bn_sqr_comba4
# bn_sqr_comba8
# bn_mul_comba4
# bn_mul_comba8
# bn_sub_words
# bn_add_words
# bn_div_words
# bn_sqr_words
# bn_mul_words
# bn_mul_add_words
#
# NOTE: It is possible to optimize this code more for
# specific PowerPC or Power architectures. On the Northstar
# architecture the optimizations in this file do
# NOT provide much improvement.
#
# If you have comments or suggestions to improve code send
# me a note at schari\@us.ibm.com
#
#--------------------------------------------------------------------------
#
# Defines to be used in the assembly code.
#
#.set r0,0 # we use it as storage for value of 0
#.set SP,1 # preserved
#.set RTOC,2 # preserved
#.set r3,3 # 1st argument/return value
#.set r4,4 # 2nd argument/volatile register
#.set r5,5 # 3rd argument/volatile register
#.set r6,6 # ...
#.set r7,7
#.set r8,8
#.set r9,9
#.set r10,10
#.set r11,11
#.set r12,12
#.set r13,13 # not used, nor any other "below" it...
# Declare function names to be global
# NOTE: For gcc these names MUST be changed to remove
# the first . i.e. for example change ".bn_sqr_comba4"
# to "bn_sqr_comba4". This should be automatically done
# in the build.
.globl .bn_sqr_comba4
.globl .bn_sqr_comba8
.globl .bn_mul_comba4
.globl .bn_mul_comba8
.globl .bn_sub_words
.globl .bn_add_words
.globl .bn_div_words
.globl .bn_sqr_words
.globl .bn_mul_words
.globl .bn_mul_add_words
# .text section
.machine "any"
#
# NOTE: The following label name should be changed to
# "bn_sqr_comba4" i.e. remove the first dot
# for the gcc compiler. This should be automatically
# done in the build
#
.align 4
.bn_sqr_comba4:
#
# Optimized version of bn_sqr_comba4.
#
# void bn_sqr_comba4(BN_ULONG *r, BN_ULONG *a)
# r3 contains r
# r4 contains a
#
# Freely use registers r5,r6,r7,r8,r9,r10,r11 as follows:
#
# r5,r6 are the two BN_ULONGs being multiplied.
# r7,r8 are the results of the 32x32 giving 64 bit multiply.
# r9,r10, r11 are the equivalents of c1,c2, c3.
# Here's the assembly
#
#
xor r0,r0,r0 # set r0 = 0. Used in the addze
# instructions below
#sqr_add_c(a,0,c1,c2,c3)
$LD r5,`0*$BNSZ`(r4)
$UMULL r9,r5,r5
$UMULH r10,r5,r5 #in first iteration. No need
#to add since c1=c2=c3=0.
# Note c3(r11) is NOT set to 0
# but will be.
$ST r9,`0*$BNSZ`(r3) # r[0]=c1;
# sqr_add_c2(a,1,0,c2,c3,c1);
$LD r6,`1*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r7,r7,r7 # compute (r7,r8)=2*(r7,r8)
adde r8,r8,r8
addze r9,r0 # catch carry if any.
# r9= r0(=0) and carry
addc r10,r7,r10 # now add to temp result.
addze r11,r8 # r8 added to r11 which is 0
addze r9,r9
$ST r10,`1*$BNSZ`(r3) #r[1]=c2;
#sqr_add_c(a,1,c3,c1,c2)
$UMULL r7,r6,r6
$UMULH r8,r6,r6
addc r11,r7,r11
adde r9,r8,r9
addze r10,r0
#sqr_add_c2(a,2,0,c3,c1,c2)
$LD r6,`2*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r7,r7,r7
adde r8,r8,r8
addze r10,r10
addc r11,r7,r11
adde r9,r8,r9
addze r10,r10
$ST r11,`2*$BNSZ`(r3) #r[2]=c3
#sqr_add_c2(a,3,0,c1,c2,c3);
$LD r6,`3*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r7,r7,r7
adde r8,r8,r8
addze r11,r0
addc r9,r7,r9
adde r10,r8,r10
addze r11,r11
#sqr_add_c2(a,2,1,c1,c2,c3);
$LD r5,`1*$BNSZ`(r4)
$LD r6,`2*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r7,r7,r7
adde r8,r8,r8
addze r11,r11
addc r9,r7,r9
adde r10,r8,r10
addze r11,r11
$ST r9,`3*$BNSZ`(r3) #r[3]=c1
#sqr_add_c(a,2,c2,c3,c1);
$UMULL r7,r6,r6
$UMULH r8,r6,r6
addc r10,r7,r10
adde r11,r8,r11
addze r9,r0
#sqr_add_c2(a,3,1,c2,c3,c1);
$LD r6,`3*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r7,r7,r7
adde r8,r8,r8
addze r9,r9
addc r10,r7,r10
adde r11,r8,r11
addze r9,r9
$ST r10,`4*$BNSZ`(r3) #r[4]=c2
#sqr_add_c2(a,3,2,c3,c1,c2);
$LD r5,`2*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r7,r7,r7
adde r8,r8,r8
addze r10,r0
addc r11,r7,r11
adde r9,r8,r9
addze r10,r10
$ST r11,`5*$BNSZ`(r3) #r[5] = c3
#sqr_add_c(a,3,c1,c2,c3);
$UMULL r7,r6,r6
$UMULH r8,r6,r6
addc r9,r7,r9
adde r10,r8,r10
$ST r9,`6*$BNSZ`(r3) #r[6]=c1
$ST r10,`7*$BNSZ`(r3) #r[7]=c2
blr
.long 0
.byte 0,12,0x14,0,0,0,2,0
.long 0
#
# NOTE: The following label name should be changed to
# "bn_sqr_comba8" i.e. remove the first dot
# for the gcc compiler. This should be automatically
# done in the build
#
.align 4
.bn_sqr_comba8:
#
# This is an optimized version of the bn_sqr_comba8 routine.
# Tightly uses the adde instruction
#
#
# void bn_sqr_comba8(BN_ULONG *r, BN_ULONG *a)
# r3 contains r
# r4 contains a
#
# Freely use registers r5,r6,r7,r8,r9,r10,r11 as follows:
#
# r5,r6 are the two BN_ULONGs being multiplied.
# r7,r8 are the results of the 32x32 giving 64 bit multiply.
# r9,r10, r11 are the equivalents of c1,c2, c3.
#
# Possible optimization of loading all 8 longs of a into registers
# doesnt provide any speedup
#
xor r0,r0,r0 #set r0 = 0.Used in addze
#instructions below.
#sqr_add_c(a,0,c1,c2,c3);
$LD r5,`0*$BNSZ`(r4)
$UMULL r9,r5,r5 #1st iteration: no carries.
$UMULH r10,r5,r5
$ST r9,`0*$BNSZ`(r3) # r[0]=c1;
#sqr_add_c2(a,1,0,c2,c3,c1);
$LD r6,`1*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r10,r7,r10 #add the two register number
adde r11,r8,r0 # (r8,r7) to the three register
addze r9,r0 # number (r9,r11,r10).NOTE:r0=0
addc r10,r7,r10 #add the two register number
adde r11,r8,r11 # (r8,r7) to the three register
addze r9,r9 # number (r9,r11,r10).
$ST r10,`1*$BNSZ`(r3) # r[1]=c2
#sqr_add_c(a,1,c3,c1,c2);
$UMULL r7,r6,r6
$UMULH r8,r6,r6
addc r11,r7,r11
adde r9,r8,r9
addze r10,r0
#sqr_add_c2(a,2,0,c3,c1,c2);
$LD r6,`2*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r11,r7,r11
adde r9,r8,r9
addze r10,r10
addc r11,r7,r11
adde r9,r8,r9
addze r10,r10
$ST r11,`2*$BNSZ`(r3) #r[2]=c3
#sqr_add_c2(a,3,0,c1,c2,c3);
$LD r6,`3*$BNSZ`(r4) #r6 = a[3]. r5 is already a[0].
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r9,r7,r9
adde r10,r8,r10
addze r11,r0
addc r9,r7,r9
adde r10,r8,r10
addze r11,r11
#sqr_add_c2(a,2,1,c1,c2,c3);
$LD r5,`1*$BNSZ`(r4)
$LD r6,`2*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r9,r7,r9
adde r10,r8,r10
addze r11,r11
addc r9,r7,r9
adde r10,r8,r10
addze r11,r11
$ST r9,`3*$BNSZ`(r3) #r[3]=c1;
#sqr_add_c(a,2,c2,c3,c1);
$UMULL r7,r6,r6
$UMULH r8,r6,r6
addc r10,r7,r10
adde r11,r8,r11
addze r9,r0
#sqr_add_c2(a,3,1,c2,c3,c1);
$LD r6,`3*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r10,r7,r10
adde r11,r8,r11
addze r9,r9
addc r10,r7,r10
adde r11,r8,r11
addze r9,r9
#sqr_add_c2(a,4,0,c2,c3,c1);
$LD r5,`0*$BNSZ`(r4)
$LD r6,`4*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r10,r7,r10
adde r11,r8,r11
addze r9,r9
addc r10,r7,r10
adde r11,r8,r11
addze r9,r9
$ST r10,`4*$BNSZ`(r3) #r[4]=c2;
#sqr_add_c2(a,5,0,c3,c1,c2);
$LD r6,`5*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r11,r7,r11
adde r9,r8,r9
addze r10,r0
addc r11,r7,r11
adde r9,r8,r9
addze r10,r10
#sqr_add_c2(a,4,1,c3,c1,c2);
$LD r5,`1*$BNSZ`(r4)
$LD r6,`4*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r11,r7,r11
adde r9,r8,r9
addze r10,r10
addc r11,r7,r11
adde r9,r8,r9
addze r10,r10
#sqr_add_c2(a,3,2,c3,c1,c2);
$LD r5,`2*$BNSZ`(r4)
$LD r6,`3*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r11,r7,r11
adde r9,r8,r9
addze r10,r10
addc r11,r7,r11
adde r9,r8,r9
addze r10,r10
$ST r11,`5*$BNSZ`(r3) #r[5]=c3;
#sqr_add_c(a,3,c1,c2,c3);
$UMULL r7,r6,r6
$UMULH r8,r6,r6
addc r9,r7,r9
adde r10,r8,r10
addze r11,r0
#sqr_add_c2(a,4,2,c1,c2,c3);
$LD r6,`4*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r9,r7,r9
adde r10,r8,r10
addze r11,r11
addc r9,r7,r9
adde r10,r8,r10
addze r11,r11
#sqr_add_c2(a,5,1,c1,c2,c3);
$LD r5,`1*$BNSZ`(r4)
$LD r6,`5*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r9,r7,r9
adde r10,r8,r10
addze r11,r11
addc r9,r7,r9
adde r10,r8,r10
addze r11,r11
#sqr_add_c2(a,6,0,c1,c2,c3);
$LD r5,`0*$BNSZ`(r4)
$LD r6,`6*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r9,r7,r9
adde r10,r8,r10
addze r11,r11
addc r9,r7,r9
adde r10,r8,r10
addze r11,r11
$ST r9,`6*$BNSZ`(r3) #r[6]=c1;
#sqr_add_c2(a,7,0,c2,c3,c1);
$LD r6,`7*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r10,r7,r10
adde r11,r8,r11
addze r9,r0
addc r10,r7,r10
adde r11,r8,r11
addze r9,r9
#sqr_add_c2(a,6,1,c2,c3,c1);
$LD r5,`1*$BNSZ`(r4)
$LD r6,`6*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r10,r7,r10
adde r11,r8,r11
addze r9,r9
addc r10,r7,r10
adde r11,r8,r11
addze r9,r9
#sqr_add_c2(a,5,2,c2,c3,c1);
$LD r5,`2*$BNSZ`(r4)
$LD r6,`5*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r10,r7,r10
adde r11,r8,r11
addze r9,r9
addc r10,r7,r10
adde r11,r8,r11
addze r9,r9
#sqr_add_c2(a,4,3,c2,c3,c1);
$LD r5,`3*$BNSZ`(r4)
$LD r6,`4*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r10,r7,r10
adde r11,r8,r11
addze r9,r9
addc r10,r7,r10
adde r11,r8,r11
addze r9,r9
$ST r10,`7*$BNSZ`(r3) #r[7]=c2;
#sqr_add_c(a,4,c3,c1,c2);
$UMULL r7,r6,r6
$UMULH r8,r6,r6
addc r11,r7,r11
adde r9,r8,r9
addze r10,r0
#sqr_add_c2(a,5,3,c3,c1,c2);
$LD r6,`5*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r11,r7,r11
adde r9,r8,r9
addze r10,r10
addc r11,r7,r11
adde r9,r8,r9
addze r10,r10
#sqr_add_c2(a,6,2,c3,c1,c2);
$LD r5,`2*$BNSZ`(r4)
$LD r6,`6*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r11,r7,r11
adde r9,r8,r9
addze r10,r10
addc r11,r7,r11
adde r9,r8,r9
addze r10,r10
#sqr_add_c2(a,7,1,c3,c1,c2);
$LD r5,`1*$BNSZ`(r4)
$LD r6,`7*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r11,r7,r11
adde r9,r8,r9
addze r10,r10
addc r11,r7,r11
adde r9,r8,r9
addze r10,r10
$ST r11,`8*$BNSZ`(r3) #r[8]=c3;
#sqr_add_c2(a,7,2,c1,c2,c3);
$LD r5,`2*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r9,r7,r9
adde r10,r8,r10
addze r11,r0
addc r9,r7,r9
adde r10,r8,r10
addze r11,r11
#sqr_add_c2(a,6,3,c1,c2,c3);
$LD r5,`3*$BNSZ`(r4)
$LD r6,`6*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r9,r7,r9
adde r10,r8,r10
addze r11,r11
addc r9,r7,r9
adde r10,r8,r10
addze r11,r11
#sqr_add_c2(a,5,4,c1,c2,c3);
$LD r5,`4*$BNSZ`(r4)
$LD r6,`5*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r9,r7,r9
adde r10,r8,r10
addze r11,r11
addc r9,r7,r9
adde r10,r8,r10
addze r11,r11
$ST r9,`9*$BNSZ`(r3) #r[9]=c1;
#sqr_add_c(a,5,c2,c3,c1);
$UMULL r7,r6,r6
$UMULH r8,r6,r6
addc r10,r7,r10
adde r11,r8,r11
addze r9,r0
#sqr_add_c2(a,6,4,c2,c3,c1);
$LD r6,`6*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r10,r7,r10
adde r11,r8,r11
addze r9,r9
addc r10,r7,r10
adde r11,r8,r11
addze r9,r9
#sqr_add_c2(a,7,3,c2,c3,c1);
$LD r5,`3*$BNSZ`(r4)
$LD r6,`7*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r10,r7,r10
adde r11,r8,r11
addze r9,r9
addc r10,r7,r10
adde r11,r8,r11
addze r9,r9
$ST r10,`10*$BNSZ`(r3) #r[10]=c2;
#sqr_add_c2(a,7,4,c3,c1,c2);
$LD r5,`4*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r11,r7,r11
adde r9,r8,r9
addze r10,r0
addc r11,r7,r11
adde r9,r8,r9
addze r10,r10
#sqr_add_c2(a,6,5,c3,c1,c2);
$LD r5,`5*$BNSZ`(r4)
$LD r6,`6*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r11,r7,r11
adde r9,r8,r9
addze r10,r10
addc r11,r7,r11
adde r9,r8,r9
addze r10,r10
$ST r11,`11*$BNSZ`(r3) #r[11]=c3;
#sqr_add_c(a,6,c1,c2,c3);
$UMULL r7,r6,r6
$UMULH r8,r6,r6
addc r9,r7,r9
adde r10,r8,r10
addze r11,r0
#sqr_add_c2(a,7,5,c1,c2,c3)
$LD r6,`7*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r9,r7,r9
adde r10,r8,r10
addze r11,r11
addc r9,r7,r9
adde r10,r8,r10
addze r11,r11
$ST r9,`12*$BNSZ`(r3) #r[12]=c1;
#sqr_add_c2(a,7,6,c2,c3,c1)
$LD r5,`6*$BNSZ`(r4)
$UMULL r7,r5,r6
$UMULH r8,r5,r6
addc r10,r7,r10
adde r11,r8,r11
addze r9,r0
addc r10,r7,r10
adde r11,r8,r11
addze r9,r9
$ST r10,`13*$BNSZ`(r3) #r[13]=c2;
#sqr_add_c(a,7,c3,c1,c2);
$UMULL r7,r6,r6
$UMULH r8,r6,r6
addc r11,r7,r11
adde r9,r8,r9
$ST r11,`14*$BNSZ`(r3) #r[14]=c3;
$ST r9, `15*$BNSZ`(r3) #r[15]=c1;
blr
.long 0
.byte 0,12,0x14,0,0,0,2,0
.long 0
#
# NOTE: The following label name should be changed to
# "bn_mul_comba4" i.e. remove the first dot
# for the gcc compiler. This should be automatically
# done in the build
#
.align 4
.bn_mul_comba4:
#
# This is an optimized version of the bn_mul_comba4 routine.
#
# void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
# r3 contains r
# r4 contains a
# r5 contains b
# r6, r7 are the 2 BN_ULONGs being multiplied.
# r8, r9 are the results of the 32x32 giving 64 multiply.
# r10, r11, r12 are the equivalents of c1, c2, and c3.
#
xor r0,r0,r0 #r0=0. Used in addze below.
#mul_add_c(a[0],b[0],c1,c2,c3);
$LD r6,`0*$BNSZ`(r4)
$LD r7,`0*$BNSZ`(r5)
$UMULL r10,r6,r7
$UMULH r11,r6,r7
$ST r10,`0*$BNSZ`(r3) #r[0]=c1
#mul_add_c(a[0],b[1],c2,c3,c1);
$LD r7,`1*$BNSZ`(r5)
$UMULL r8,r6,r7
$UMULH r9,r6,r7
addc r11,r8,r11
adde r12,r9,r0
addze r10,r0
#mul_add_c(a[1],b[0],c2,c3,c1);
$LD r6, `1*$BNSZ`(r4)
$LD r7, `0*$BNSZ`(r5)
$UMULL r8,r6,r7
$UMULH r9,r6,r7
addc r11,r8,r11
adde r12,r9,r12
addze r10,r10
$ST r11,`1*$BNSZ`(r3) #r[1]=c2
#mul_add_c(a[2],b[0],c3,c1,c2);
$LD r6,`2*$BNSZ`(r4)
$UMULL r8,r6,r7
$UMULH r9,r6,r7
addc r12,r8,r12
adde r10,r9,r10
addze r11,r0
#mul_add_c(a[1],b[1],c3,c1,c2);
$LD r6,`1*$BNSZ`(r4)
$LD r7,`1*$BNSZ`(r5)
$UMULL r8,r6,r7
$UMULH r9,r6,r7
addc r12,r8,r12
adde r10,r9,r10
addze r11,r11
#mul_add_c(a[0],b[2],c3,c1,c2);
$LD r6,`0*$BNSZ`(r4)
$LD r7,`2*$BNSZ`(r5)
$UMULL r8,r6,r7
$UMULH r9,r6,r7
addc r12,r8,r12
adde r10,r9,r10
addze r11,r11
$ST r12,`2*$BNSZ`(r3) #r[2]=c3
#mul_add_c(a[0],b[3],c1,c2,c3);
$LD r7,`3*$BNSZ`(r5)
$UMULL r8,r6,r7
$UMULH r9,r6,r7
addc r10,r8,r10
adde r11,r9,r11
addze r12,r0
#mul_add_c(a[1],b[2],c1,c2,c3);
$LD r6,`1*$BNSZ`(r4)
$LD r7,`2*$BNSZ`(r5)
$UMULL r8,r6,r7
$UMULH r9,r6,r7
addc r10,r8,r10
adde r11,r9,r11
addze r12,r12
#mul_add_c(a[2],b[1],c1,c2,c3);
$LD r6,`2*$BNSZ`(r4)
$LD r7,`1*$BNSZ`(r5)
$UMULL r8,r6,r7
$UMULH r9,r6,r7
addc r10,r8,r10
adde r11,r9,r11
addze r12,r12
#mul_add_c(a[3],b[0],c1,c2,c3);
$LD r6,`3*$BNSZ`(r4)
$LD r7,`0*$BNSZ`(r5)
$UMULL r8,r6,r7
$UMULH r9,r6,r7
addc r10,r8,r10
adde r11,r9,r11
addze r12,r12
$ST r10,`3*$BNSZ`(r3) #r[3]=c1
#mul_add_c(a[3],b[1],c2,c3,c1);
$LD r7,`1*$BNSZ`(r5)
$UMULL r8,r6,r7
$UMULH r9,r6,r7
addc r11,r8,r11
adde r12,r9,r12
addze r10,r0
#mul_add_c(a[2],b[2],c2,c3,c1);
$LD r6,`2*$BNSZ`(r4)
$LD r7,`2*$BNSZ`(r5)
$UMULL r8,r6,r7
$UMULH r9,r6,r7
addc r11,r8,r11
adde r12,r9,r12
addze r10,r10
#mul_add_c(a[1],b[3],c2,c3,c1);
$LD r6,`1*$BNSZ`(r4)
$LD r7,`3*$BNSZ`(r5)
$UMULL r8,r6,r7
$UMULH r9,r6,r7
addc r11,r8,r11
adde r12,r9,r12
addze r10,r10
$ST r11,`4*$BNSZ`(r3) #r[4]=c2
#mul_add_c(a[2],b[3],c3,c1,c2);
$LD r6,`2*$BNSZ`(r4)
$UMULL r8,r6,r7
$UMULH r9,r6,r7
addc r12,r8,r12
adde r10,r9,r10
addze r11,r0
#mul_add_c(a[3],b[2],c3,c1,c2);
$LD r6,`3*$BNSZ`(r4)
$LD r7,`2*$BNSZ`(r5)
$UMULL r8,r6,r7
$UMULH r9,r6,r7
addc r12,r8,r12
adde r10,r9,r10
addze r11,r11
$ST r12,`5*$BNSZ`(r3) #r[5]=c3
#mul_add_c(a[3],b[3],c1,c2,c3);
$LD r7,`3*$BNSZ`(r5)
$UMULL r8,r6,r7
$UMULH r9,r6,r7
addc r10,r8,r10
adde r11,r9,r11
$ST r10,`6*$BNSZ`(r3) #r[6]=c1
$ST r11,`7*$BNSZ`(r3) #r[7]=c2
blr
.long 0
.byte 0,12,0x14,0,0,0,3,0
.long 0
#
# NOTE: The following label name should be changed to
# "bn_mul_comba8" i.e. remove the first dot
# for the gcc compiler. This should be automatically
# done in the build
#
.align 4
.bn_mul_comba8:
#
# Optimized version of the bn_mul_comba8 routine.
#
# void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
# r3 contains r
# r4 contains a
# r5 contains b
# r6, r7 are the 2 BN_ULONGs being multiplied.
# r8, r9 are the results of the 32x32 giving 64 multiply.
# r10, r11, r12 are the equivalents of c1, c2, and c3.
#
xor r0,r0,r0 #r0=0. Used in addze below.
#mul_add_c(a[0],b[0],c1,c2,c3);
$LD r6,`0*$BNSZ`(r4) #a[0]
$LD r7,`0*$BNSZ`(r5) #b[0]