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
|
#! /usr/bin/env perl
# Copyright 2010-2018 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/.
# ====================================================================
# January 2010
#
# "Teaser" Montgomery multiplication module for IA-64. There are
# several possibilities for improvement:
#
# - modulo-scheduling outer loop would eliminate quite a number of
# stalls after ldf8, xma and getf.sig outside inner loop and
# improve shorter key performance;
# - shorter vector support [with input vectors being fetched only
# once] should be added;
# - 2x unroll with help of n0[1] would make the code scalable on
# "wider" IA-64, "wider" than Itanium 2 that is, which is not of
# acute interest, because upcoming Tukwila's individual cores are
# reportedly based on Itanium 2 design;
# - dedicated squaring procedure(?);
#
# January 2010
#
# Shorter vector support is implemented by zero-padding ap and np
# vectors up to 8 elements, or 512 bits. This means that 256-bit
# inputs will be processed only 2 times faster than 512-bit inputs,
# not 4 [as one would expect, because algorithm complexity is n^2].
# The reason for padding is that inputs shorter than 512 bits won't
# be processed faster anyway, because minimal critical path of the
# core loop happens to match 512-bit timing. Either way, it resulted
# in >100% improvement of 512-bit RSA sign benchmark and 50% - of
# 1024-bit one [in comparison to original version of *this* module].
#
# So far 'openssl speed rsa dsa' output on 900MHz Itanium 2 *with*
# this module is:
# sign verify sign/s verify/s
# rsa 512 bits 0.000290s 0.000024s 3452.8 42031.4
# rsa 1024 bits 0.000793s 0.000058s 1261.7 17172.0
# rsa 2048 bits 0.005908s 0.000148s 169.3 6754.0
# rsa 4096 bits 0.033456s 0.000469s 29.9 2133.6
# dsa 512 bits 0.000253s 0.000198s 3949.9 5057.0
# dsa 1024 bits 0.000585s 0.000607s 1708.4 1647.4
# dsa 2048 bits 0.001453s 0.001703s 688.1 587.4
#
# ... and *without* (but still with ia64.S):
#
# rsa 512 bits 0.000670s 0.000041s 1491.8 24145.5
# rsa 1024 bits 0.001988s 0.000080s 502.9 12499.3
# rsa 2048 bits 0.008702s 0.000189s 114.9 5293.9
# rsa 4096 bits 0.043860s 0.000533s 22.8 1875.9
# dsa 512 bits 0.000441s 0.000427s 2265.3 2340.6
# dsa 1024 bits 0.000823s 0.000867s 1215.6 1153.2
# dsa 2048 bits 0.001894s 0.002179s 528.1 458.9
#
# As it can be seen, RSA sign performance improves by 130-30%,
# hereafter less for longer keys, while verify - by 74-13%.
# DSA performance improves by 115-30%.
$output=pop;
if ($^O eq "hpux") {
$ADDP="addp4";
for (@ARGV) { $ADDP="add" if (/[\+DD|\-mlp]64/); }
} else { $ADDP="add"; }
$code=<<___;
.explicit
.text
// int bn_mul_mont (BN_ULONG *rp,const BN_ULONG *ap,
// const BN_ULONG *bp,const BN_ULONG *np,
// const BN_ULONG *n0p,int num);
.align 64
.global bn_mul_mont#
.proc bn_mul_mont#
bn_mul_mont:
.prologue
.body
{ .mmi; cmp4.le p6,p7=2,r37;;
(p6) cmp4.lt.unc p8,p9=8,r37
mov ret0=r0 };;
{ .bbb;
(p9) br.cond.dptk.many bn_mul_mont_8
(p8) br.cond.dpnt.many bn_mul_mont_general
(p7) br.ret.spnt.many b0 };;
.endp bn_mul_mont#
prevfs=r2; prevpr=r3; prevlc=r10; prevsp=r11;
rptr=r8; aptr=r9; bptr=r14; nptr=r15;
tptr=r16; // &tp[0]
tp_1=r17; // &tp[-1]
num=r18; len=r19; lc=r20;
topbit=r21; // carry bit from tmp[num]
n0=f6;
m0=f7;
bi=f8;
.align 64
.local bn_mul_mont_general#
.proc bn_mul_mont_general#
bn_mul_mont_general:
.prologue
{ .mmi; .save ar.pfs,prevfs
alloc prevfs=ar.pfs,6,2,0,8
$ADDP aptr=0,in1
.save ar.lc,prevlc
mov prevlc=ar.lc }
{ .mmi; .vframe prevsp
mov prevsp=sp
$ADDP bptr=0,in2
.save pr,prevpr
mov prevpr=pr };;
.body
.rotf alo[6],nlo[4],ahi[8],nhi[6]
.rotr a[3],n[3],t[2]
{ .mmi; ldf8 bi=[bptr],8 // (*bp++)
ldf8 alo[4]=[aptr],16 // ap[0]
$ADDP r30=8,in1 };;
{ .mmi; ldf8 alo[3]=[r30],16 // ap[1]
ldf8 alo[2]=[aptr],16 // ap[2]
$ADDP in4=0,in4 };;
{ .mmi; ldf8 alo[1]=[r30] // ap[3]
ldf8 n0=[in4] // n0
$ADDP rptr=0,in0 }
{ .mmi; $ADDP nptr=0,in3
mov r31=16
zxt4 num=in5 };;
{ .mmi; ldf8 nlo[2]=[nptr],8 // np[0]
shladd len=num,3,r0
shladd r31=num,3,r31 };;
{ .mmi; ldf8 nlo[1]=[nptr],8 // np[1]
add lc=-5,num
sub r31=sp,r31 };;
{ .mfb; and sp=-16,r31 // alloca
xmpy.hu ahi[2]=alo[4],bi // ap[0]*bp[0]
nop.b 0 }
{ .mfb; nop.m 0
xmpy.lu alo[4]=alo[4],bi
brp.loop.imp .L1st_ctop,.L1st_cend-16
};;
{ .mfi; nop.m 0
xma.hu ahi[1]=alo[3],bi,ahi[2] // ap[1]*bp[0]
add tp_1=8,sp }
{ .mfi; nop.m 0
xma.lu alo[3]=alo[3],bi,ahi[2]
mov pr.rot=0x20001f<<16
// ------^----- (p40) at first (p23)
// ----------^^ p[16:20]=1
};;
{ .mfi; nop.m 0
xmpy.lu m0=alo[4],n0 // (ap[0]*bp[0])*n0
mov ar.lc=lc }
{ .mfi; nop.m 0
fcvt.fxu.s1 nhi[1]=f0
mov ar.ec=8 };;
.align 32
.L1st_ctop:
.pred.rel "mutex",p40,p42
{ .mfi; (p16) ldf8 alo[0]=[aptr],8 // *(aptr++)
(p18) xma.hu ahi[0]=alo[2],bi,ahi[1]
(p40) add n[2]=n[2],a[2] } // (p23) }
{ .mfi; (p18) ldf8 nlo[0]=[nptr],8 // *(nptr++)(p16)
(p18) xma.lu alo[2]=alo[2],bi,ahi[1]
(p42) add n[2]=n[2],a[2],1 };; // (p23)
{ .mfi; (p21) getf.sig a[0]=alo[5]
(p20) xma.hu nhi[0]=nlo[2],m0,nhi[1]
(p42) cmp.leu p41,p39=n[2],a[2] } // (p23)
{ .mfi; (p23) st8 [tp_1]=n[2],8
(p20) xma.lu nlo[2]=nlo[2],m0,nhi[1]
(p40) cmp.ltu p41,p39=n[2],a[2] } // (p23)
{ .mmb; (p21) getf.sig n[0]=nlo[3]
(p16) nop.m 0
br.ctop.sptk .L1st_ctop };;
.L1st_cend:
{ .mmi; getf.sig a[0]=ahi[6] // (p24)
getf.sig n[0]=nhi[4]
add num=-1,num };; // num--
{ .mmi; .pred.rel "mutex",p40,p42
(p40) add n[0]=n[0],a[0]
(p42) add n[0]=n[0],a[0],1
sub aptr=aptr,len };; // rewind
{ .mmi; .pred.rel "mutex",p40,p42
(p40) cmp.ltu p41,p39=n[0],a[0]
(p42) cmp.leu p41,p39=n[0],a[0]
sub nptr=nptr,len };;
{ .mmi; .pred.rel "mutex",p39,p41
(p39) add topbit=r0,r0
(p41) add topbit=r0,r0,1
nop.i 0 }
{ .mmi; st8 [tp_1]=n[0]
add tptr=16,sp
add tp_1=8,sp };;
.Louter:
{ .mmi; ldf8 bi=[bptr],8 // (*bp++)
ldf8 ahi[3]=[tptr] // tp[0]
add r30=8,aptr };;
{ .mmi; ldf8 alo[4]=[aptr],16 // ap[0]
ldf8 alo[3]=[r30],16 // ap[1]
add r31=8,nptr };;
{ .mfb; ldf8 alo[2]=[aptr],16 // ap[2]
xma.hu ahi[2]=alo[4],bi,ahi[3] // ap[0]*bp[i]+tp[0]
brp.loop.imp .Linner_ctop,.Linner_cend-16
}
{ .mfb; ldf8 alo[1]=[r30] // ap[3]
xma.lu alo[4]=alo[4],bi,ahi[3]
clrrrb.pr };;
{ .mfi; ldf8 nlo[2]=[nptr],16 // np[0]
xma.hu ahi[1]=alo[3],bi,ahi[2] // ap[1]*bp[i]
nop.i 0 }
{ .mfi; ldf8 nlo[1]=[r31] // np[1]
xma.lu alo[3]=alo[3],bi,ahi[2]
mov pr.rot=0x20101f<<16
// ------^----- (p40) at first (p23)
// --------^--- (p30) at first (p22)
// ----------^^ p[16:20]=1
};;
{ .mfi; st8 [tptr]=r0 // tp[0] is already accounted
xmpy.lu m0=alo[4],n0 // (ap[0]*bp[i]+tp[0])*n0
mov ar.lc=lc }
{ .mfi;
fcvt.fxu.s1 nhi[1]=f0
mov ar.ec=8 };;
// This loop spins in 4*(n+7) ticks on Itanium 2 and should spin in
// 7*(n+7) ticks on Itanium (the one codenamed Merced). Factor of 7
// in latter case accounts for two-tick pipeline stall, which means
// that its performance would be ~20% lower than optimal one. No
// attempt was made to address this, because original Itanium is
// hardly represented out in the wild...
.align 32
.Linner_ctop:
.pred.rel "mutex",p40,p42
.pred.rel "mutex",p30,p32
{ .mfi; (p16) ldf8 alo[0]=[aptr],8 // *(aptr++)
(p18) xma.hu ahi[0]=alo[2],bi,ahi[1]
(p40) add n[2]=n[2],a[2] } // (p23)
{ .mfi; (p16) nop.m 0
(p18) xma.lu alo[2]=alo[2],bi,ahi[1]
(p42) add n[2]=n[2],a[2],1 };; // (p23)
{ .mfi; (p21) getf.sig a[0]=alo[5]
(p16) nop.f 0
(p40) cmp.ltu p41,p39=n[2],a[2] } // (p23)
{ .mfi; (p21) ld8 t[0]=[tptr],8
(p16) nop.f 0
(p42) cmp.leu p41,p39=n[2],a[2] };; // (p23)
{ .mfi; (p18) ldf8 nlo[0]=[nptr],8 // *(nptr++)
(p20) xma.hu nhi[0]=nlo[2],m0,nhi[1]
(p30) add a[1]=a[1],t[1] } // (p22)
{ .mfi; (p16) nop.m 0
(p20) xma.lu nlo[2]=nlo[2],m0,nhi[1]
(p32) add a[1]=a[1],t[1],1 };; // (p22)
{ .mmi; (p21) getf.sig n[0]=nlo[3]
(p16) nop.m 0
(p30) cmp.ltu p31,p29=a[1],t[1] } // (p22)
{ .mmb; (p23) st8 [tp_1]=n[2],8
(p32) cmp.leu p31,p29=a[1],t[1] // (p22)
br.ctop.sptk .Linner_ctop };;
.Linner_cend:
{ .mmi; getf.sig a[0]=ahi[6] // (p24)
getf.sig n[0]=nhi[4]
nop.i 0 };;
{ .mmi; .pred.rel "mutex",p31,p33
(p31) add a[0]=a[0],topbit
(p33) add a[0]=a[0],topbit,1
mov topbit=r0 };;
{ .mfi; .pred.rel "mutex",p31,p33
(p31) cmp.ltu p32,p30=a[0],topbit
(p33) cmp.leu p32,p30=a[0],topbit
}
{ .mfi; .pred.rel "mutex",p40,p42
(p40) add n[0]=n[0],a[0]
(p42) add n[0]=n[0],a[0],1
};;
{ .mmi; .pred.rel "mutex",p44,p46
(p40) cmp.ltu p41,p39=n[0],a[0]
(p42) cmp.leu p41,p39=n[0],a[0]
(p32) add topbit=r0,r0,1 }
{ .mmi; st8 [tp_1]=n[0],8
cmp4.ne p6,p0=1,num
sub aptr=aptr,len };; // rewind
{ .mmi; sub nptr=nptr,len
(p41) add topbit=r0,r0,1
add tptr=16,sp }
{ .mmb; add tp_1=8,sp
add num=-1,num // num--
(p6) br.cond.sptk.many .Louter };;
{ .mbb; add lc=4,lc
brp.loop.imp .Lsub_ctop,.Lsub_cend-16
clrrrb.pr };;
{ .mii; nop.m 0
mov pr.rot=0x10001<<16
// ------^---- (p33) at first (p17)
mov ar.lc=lc }
{ .mii; nop.m 0
mov ar.ec=3
nop.i 0 };;
.Lsub_ctop:
.pred.rel "mutex",p33,p35
{ .mfi; (p16) ld8 t[0]=[tptr],8 // t=*(tp++)
(p16) nop.f 0
(p33) sub n[1]=t[1],n[1] } // (p17)
{ .mfi; (p16) ld8 n[0]=[nptr],8 // n=*(np++)
(p16) nop.f 0
(p35) sub n[1]=t[1],n[1],1 };; // (p17)
{ .mib; (p18) st8 [rptr]=n[2],8 // *(rp++)=r
(p33) cmp.gtu p34,p32=n[1],t[1] // (p17)
(p18) nop.b 0 }
{ .mib; (p18) nop.m 0
(p35) cmp.geu p34,p32=n[1],t[1] // (p17)
br.ctop.sptk .Lsub_ctop };;
.Lsub_cend:
{ .mmb; .pred.rel "mutex",p34,p36
(p34) sub topbit=topbit,r0 // (p19)
(p36) sub topbit=topbit,r0,1
brp.loop.imp .Lcopy_ctop,.Lcopy_cend-16
}
{ .mmb; sub rptr=rptr,len // rewind
sub tptr=tptr,len
clrrrb.pr };;
{ .mmi; mov aptr=rptr
mov bptr=tptr
mov pr.rot=1<<16 };;
{ .mii; cmp.eq p0,p6=topbit,r0
mov ar.lc=lc
mov ar.ec=2 };;
.Lcopy_ctop:
{ .mmi; (p16) ld8 a[0]=[aptr],8
(p16) ld8 t[0]=[bptr],8
(p6) mov a[1]=t[1] };; // (p17)
{ .mmb; (p17) st8 [rptr]=a[1],8
(p17) st8 [tptr]=r0,8
br.ctop.sptk .Lcopy_ctop };;
.Lcopy_cend:
{ .mmi; mov ret0=1 // signal "handled"
rum 1<<5 // clear um.mfh
mov ar.lc=prevlc }
{ .mib; .restore sp
mov sp=prevsp
mov pr=prevpr,0x1ffff
br.ret.sptk.many b0 };;
.endp bn_mul_mont_general#
a1=r16; a2=r17; a3=r18; a4=r19; a5=r20; a6=r21; a7=r22; a8=r23;
n1=r24; n2=r25; n3=r26; n4=r27; n5=r28; n6=r29; n7=r30; n8=r31;
t0=r15;
ai0=f8; ai1=f9; ai2=f10; ai3=f11; ai4=f12; ai5=f13; ai6=f14; ai7=f15;
ni0=f16; ni1=f17; ni2=f18; ni3=f19; ni4=f20; ni5=f21; ni6=f22; ni7=f23;
.align 64
.skip 48 // aligns loop body
.local bn_mul_mont_8#
.proc bn_mul_mont_8#
bn_mul_mont_8:
.prologue
{ .mmi; .save ar.pfs,prevfs
alloc prevfs=ar.pfs,6,2,0,8
.vframe prevsp
mov prevsp=sp
.save ar.lc,prevlc
mov prevlc=ar.lc }
{ .mmi; add r17=-6*16,sp
add sp=-7*16,sp
.save pr,prevpr
mov prevpr=pr };;
{ .mmi; .save.gf 0,0x10
stf.spill [sp]=f16,-16
.save.gf 0,0x20
stf.spill [r17]=f17,32
add r16=-5*16,prevsp};;
{ .mmi; .save.gf 0,0x40
stf.spill [r16]=f18,32
.save.gf 0,0x80
stf.spill [r17]=f19,32
$ADDP aptr=0,in1 };;
{ .mmi; .save.gf 0,0x100
stf.spill [r16]=f20,32
.save.gf 0,0x200
stf.spill [r17]=f21,32
$ADDP r29=8,in1 };;
{ .mmi; .save.gf 0,0x400
stf.spill [r16]=f22
.save.gf 0,0x800
stf.spill [r17]=f23
$ADDP rptr=0,in0 };;
.body
.rotf bj[8],mj[2],tf[2],alo[10],ahi[10],nlo[10],nhi[10]
.rotr t[8]
// load input vectors padding them to 8 elements
{ .mmi; ldf8 ai0=[aptr],16 // ap[0]
ldf8 ai1=[r29],16 // ap[1]
$ADDP bptr=0,in2 }
{ .mmi; $ADDP r30=8,in2
$ADDP nptr=0,in3
$ADDP r31=8,in3 };;
{ .mmi; ldf8 bj[7]=[bptr],16 // bp[0]
ldf8 bj[6]=[r30],16 // bp[1]
cmp4.le p4,p5=3,in5 }
{ .mmi; ldf8 ni0=[nptr],16 // np[0]
ldf8 ni1=[r31],16 // np[1]
cmp4.le p6,p7=4,in5 };;
{ .mfi; (p4)ldf8 ai2=[aptr],16 // ap[2]
(p5)fcvt.fxu ai2=f0
cmp4.le p8,p9=5,in5 }
{ .mfi; (p6)ldf8 ai3=[r29],16 // ap[3]
(p7)fcvt.fxu ai3=f0
cmp4.le p10,p11=6,in5 }
{ .mfi; (p4)ldf8 bj[5]=[bptr],16 // bp[2]
(p5)fcvt.fxu bj[5]=f0
cmp4.le p12,p13=7,in5 }
{ .mfi; (p6)ldf8 bj[4]=[r30],16 // bp[3]
(p7)fcvt.fxu bj[4]=f0
cmp4.le p14,p15=8,in5 }
{ .mfi; (p4)ldf8 ni2=[nptr],16 // np[2]
(p5)fcvt.fxu ni2=f0
addp4 r28=-1,in5 }
{ .mfi; (p6)ldf8 ni3=[r31],16 // np[3]
(p7)fcvt.fxu ni3=f0
$ADDP in4=0,in4 };;
{ .mfi; ldf8 n0=[in4]
fcvt.fxu tf[1]=f0
nop.i 0 }
{ .mfi; (p8)ldf8 ai4=[aptr],16 // ap[4]
(p9)fcvt.fxu ai4=f0
mov t[0]=r0 }
{ .mfi; (p10)ldf8 ai5=[r29],16 // ap[5]
(p11)fcvt.fxu ai5=f0
mov t[1]=r0 }
{ .mfi; (p8)ldf8 bj[3]=[bptr],16 // bp[4]
(p9)fcvt.fxu bj[3]=f0
mov t[2]=r0 }
{ .mfi; (p10)ldf8 bj[2]=[r30],16 // bp[5]
(p11)fcvt.fxu bj[2]=f0
mov t[3]=r0 }
{ .mfi; (p8)ldf8 ni4=[nptr],16 // np[4]
(p9)fcvt.fxu ni4=f0
mov t[4]=r0 }
{ .mfi; (p10)ldf8 ni5=[r31],16 // np[5]
(p11)fcvt.fxu ni5=f0
mov t[5]=r0 };;
{ .mfi; (p12)ldf8 ai6=[aptr],16 // ap[6]
(p13)fcvt.fxu ai6=f0
mov t[6]=r0 }
{ .mfi; (p14)ldf8 ai7=[r29],16 // ap[7]
(p15)fcvt.fxu ai7=f0
mov t[7]=r0 }
|