2 # Copyright 2016 The OpenSSL Project Authors. All Rights Reserved.
4 # Licensed under the OpenSSL license (the "License"). You may not use
5 # this file except in compliance with the License. You can obtain a copy
6 # in the file LICENSE in the source distribution or at
7 # https://www.openssl.org/source/license.html
10 # ====================================================================
11 # Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
12 # project. The module is, however, dual licensed under OpenSSL and
13 # CRYPTOGAMS licenses depending on where you obtain it. For further
14 # details see http://www.openssl.org/~appro/cryptogams/.
15 # ====================================================================
17 # This module implements Poly1305 hash for ARMv8.
21 # Numbers are cycles per processed byte with poly1305_blocks alone.
25 # Apple A7 1.86/+5% 0.72
26 # Cortex-A53 2.69/+58% 1.47
27 # Cortex-A57 2.70/+7% 1.14
28 # Denver 1.64/+50% 1.18(*)
29 # X-Gene 2.13/+68% 2.27
30 # Mongoose 1.77/+75% 1.12
32 # (*) estimate based on resources availability is less than 1.0,
33 # i.e. measured result is worse than expected, presumably binary
34 # translator is not almighty;
39 $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
40 ( $xlate="${dir}arm-xlate.pl" and -f $xlate ) or
41 ( $xlate="${dir}../../perlasm/arm-xlate.pl" and -f $xlate) or
42 die "can't locate arm-xlate.pl";
44 open OUT,"| \"$^X\" $xlate $flavour $output";
47 my ($ctx,$inp,$len,$padbit) = map("x$_",(0..3));
48 my ($mac,$nonce)=($inp,$len);
50 my ($h0,$h1,$h2,$r0,$r1,$s1,$t0,$t1,$d0,$d1,$d2) = map("x$_",(4..14));
57 // forward "declarations" are required for Apple
58 .extern OPENSSL_armcap_P
59 .globl poly1305_blocks
63 .type poly1305_init,%function
67 stp xzr,xzr,[$ctx] // zero hash value
68 stp xzr,xzr,[$ctx,#16] // [along with is_base2_26]
74 ldrsw $t1,.LOPENSSL_armcap_P
76 ldr $t1,.LOPENSSL_armcap_P
78 adr $t0,.LOPENSSL_armcap_P
80 ldp $r0,$r1,[$inp] // load key
81 mov $s1,#0xfffffffc0fffffff
82 movk $s1,#0x0fff,lsl#48
85 rev $r0,$r0 // flip bytes
88 and $r0,$r0,$s1 // &=0ffffffc0fffffff
90 and $r1,$r1,$s1 // &=0ffffffc0ffffffc
91 stp $r0,$r1,[$ctx,#32] // save key value
95 adr $d0,poly1305_blocks
96 adr $r0,poly1305_blocks_neon
98 adr $r1,poly1305_emit_neon
112 .size poly1305_init,.-poly1305_init
114 .type poly1305_blocks,%function
120 ldp $h0,$h1,[$ctx] // load hash value
121 ldp $r0,$r1,[$ctx,#32] // load key value
123 add $s1,$r1,$r1,lsr#2 // s1 = r1 + (r1 >> 2)
128 ldp $t0,$t1,[$inp],#16 // load input
134 adds $h0,$h0,$t0 // accumulate input
137 mul $d0,$h0,$r0 // h0*r0
141 mul $t0,$h1,$s1 // h1*5*r1
145 mul $t0,$h0,$r1 // h0*r1
150 mul $t0,$h1,$r0 // h1*r0
155 mul $t0,$h2,$s1 // h2*5*r1
157 mul $t1,$h2,$r0 // h2*r0
162 and $t0,$d2,#-4 // final reduction
164 add $t0,$t0,$d2,lsr#2
171 stp $h0,$h1,[$ctx] // store hash value
176 .size poly1305_blocks,.-poly1305_blocks
178 .type poly1305_emit,%function
181 ldp $h0,$h1,[$ctx] // load hash base 2^64
183 ldp $t0,$t1,[$nonce] // load nonce
185 adds $d0,$h0,#5 // compare to modulus
189 tst $d2,#-4 // see if it's carried/borrowed
195 ror $t0,$t0,#32 // flip nonce words
198 adds $h0,$h0,$t0 // accumulate nonce
201 rev $h0,$h0 // flip output bytes
204 stp $h0,$h1,[$mac] // write result
207 .size poly1305_emit,.-poly1305_emit
209 my ($R0,$R1,$S1,$R2,$S2,$R3,$S3,$R4,$S4) = map("v$_.4s",(0..8));
210 my ($IN01_0,$IN01_1,$IN01_2,$IN01_3,$IN01_4) = map("v$_.2s",(9..13));
211 my ($IN23_0,$IN23_1,$IN23_2,$IN23_3,$IN23_4) = map("v$_.2s",(14..18));
212 my ($ACC0,$ACC1,$ACC2,$ACC3,$ACC4) = map("v$_.2d",(19..23));
213 my ($H0,$H1,$H2,$H3,$H4) = map("v$_.2s",(24..28));
214 my ($T0,$T1,$MASK) = map("v$_",(29..31));
216 my ($in2,$zeros)=("x16","x17");
217 my $is_base2_26 = $zeros; # borrow
220 .type poly1305_mult,%function
223 mul $d0,$h0,$r0 // h0*r0
226 mul $t0,$h1,$s1 // h1*5*r1
230 mul $t0,$h0,$r1 // h0*r1
235 mul $t0,$h1,$r0 // h1*r0
240 mul $t0,$h2,$s1 // h2*5*r1
242 mul $t1,$h2,$r0 // h2*r0
247 and $t0,$d2,#-4 // final reduction
249 add $t0,$t0,$d2,lsr#2
255 .size poly1305_mult,.-poly1305_mult
257 .type poly1305_splat,%function
260 and x12,$h0,#0x03ffffff // base 2^64 -> base 2^26
263 and x14,x14,#0x03ffffff
267 str w12,[$ctx,#16*0] // r0
268 add w12,w13,w13,lsl#2 // r1*5
269 str w13,[$ctx,#16*1] // r1
270 add w13,w14,w14,lsl#2 // r2*5
271 str w12,[$ctx,#16*2] // s1
272 str w14,[$ctx,#16*3] // r2
273 add w14,w15,w15,lsl#2 // r3*5
274 str w13,[$ctx,#16*4] // s2
275 str w15,[$ctx,#16*5] // r3
276 add w15,w16,w16,lsl#2 // r4*5
277 str w14,[$ctx,#16*6] // s3
278 str w16,[$ctx,#16*7] // r4
279 str w15,[$ctx,#16*8] // s4
282 .size poly1305_splat,.-poly1305_splat
284 .type poly1305_blocks_neon,%function
286 poly1305_blocks_neon:
287 ldr $is_base2_26,[$ctx,#24]
290 cbz $is_base2_26,poly1305_blocks
293 stp x29,x30,[sp,#-80]!
299 cbz $is_base2_26,.Lbase2_64_neon
301 ldp w10,w11,[$ctx] // load hash value base 2^26
302 ldp w12,w13,[$ctx,#8]
308 ldp $r0,$r1,[$ctx,#32] // load key value
310 add $h0,x10,x11,lsl#26 // base 2^26 -> base 2^64
312 adds $h0,$h0,x12,lsl#52
313 add $h1,$h1,x13,lsl#14
316 adds $h1,$h1,x14,lsl#40
317 adc $d2,$h2,xzr // can be partially reduced...
319 ldp $d0,$d1,[$inp],#16 // load input
321 add $s1,$r1,$r1,lsr#2 // s1 = r1 + (r1 >> 2)
323 and $t0,$d2,#-4 // ... so reduce
325 add $t0,$t0,$d2,lsr#2
334 adds $h0,$h0,$d0 // accumulate input
341 cbz $padbit,.Lstore_base2_64_neon
343 and x10,$h0,#0x03ffffff // base 2^64 -> base 2^26
346 and x12,x12,#0x03ffffff
350 cbnz $len,.Leven_neon
352 stp w10,w11,[$ctx] // store hash value base 2^26
353 stp w12,w13,[$ctx,#8]
358 .Lstore_base2_64_neon:
359 stp $h0,$h1,[$ctx] // store hash value base 2^64
360 stp $h2,xzr,[$ctx,#16] // note that is_base2_26 is zeroed
365 ldp $r0,$r1,[$ctx,#32] // load key value
367 ldp $h0,$h1,[$ctx] // load hash value base 2^64
373 ldp $d0,$d1,[$inp],#16 // load input
375 add $s1,$r1,$r1,lsr#2 // s1 = r1 + (r1 >> 2)
380 adds $h0,$h0,$d0 // accumulate input
387 and x10,$h0,#0x03ffffff // base 2^64 -> base 2^26
390 and x12,x12,#0x03ffffff
394 stp d8,d9,[sp,#16] // meet ABI requirements
405 ////////////////////////////////// initialize r^n table
407 add $s1,$r1,$r1,lsr#2 // s1 = r1 + (r1 >> 2)
413 bl poly1305_mult // r^2
417 bl poly1305_mult // r^3
421 bl poly1305_mult // r^4
429 csel $in2,$zeros,$in2,lo
432 str x4,[$ctx,#-24] // set is_base2_26
433 sub $ctx,$ctx,#48 // restore original $ctx
441 csel $in2,$zeros,$in2,lo
443 stp d8,d9,[sp,#16] // meet ABI requirements
455 ldp x8,x12,[$in2],#16 // inp[2:3] (or zero)
456 ldp x9,x13,[$in2],#48
458 lsl $padbit,$padbit,#24
467 and x4,x8,#0x03ffffff // base 2^64 -> base 2^26
468 and x5,x9,#0x03ffffff
471 add x4,x4,x5,lsl#32 // bfi x4,x5,#32,#32
474 add x6,x6,x7,lsl#32 // bfi x6,x7,#32,#32
476 and x8,x8,#0x03ffffff
477 and x9,x9,#0x03ffffff
480 add x12,$padbit,x12,lsr#40
481 add x13,$padbit,x13,lsr#40
482 add x8,x8,x9,lsl#32 // bfi x8,x9,#32,#32
484 add x10,x10,x11,lsl#32 // bfi x10,x11,#32,#32
485 add x12,x12,x13,lsl#32 // bfi x12,x13,#32,#32
490 ldp x8,x12,[$inp],#16 // inp[0:1]
491 ldp x9,x13,[$inp],#48
493 ld1 {$R0,$R1,$S1,$R2},[x15],#64
494 ld1 {$S2,$R3,$S3,$R4},[x15],#64
503 and x4,x8,#0x03ffffff // base 2^64 -> base 2^26
504 and x5,x9,#0x03ffffff
507 add x4,x4,x5,lsl#32 // bfi x4,x5,#32,#32
510 add x6,x6,x7,lsl#32 // bfi x6,x7,#32,#32
512 and x8,x8,#0x03ffffff
513 and x9,x9,#0x03ffffff
516 add x12,$padbit,x12,lsr#40
517 add x13,$padbit,x13,lsr#40
518 add x8,x8,x9,lsl#32 // bfi x8,x9,#32,#32
520 add x10,x10,x11,lsl#32 // bfi x10,x11,#32,#32
521 add x12,x12,x13,lsl#32 // bfi x12,x13,#32,#32
526 ushr $MASK.2d,$MASK.2d,#38
532 ////////////////////////////////////////////////////////////////
533 // ((inp[0]*r^4+inp[2]*r^2+inp[4])*r^4+inp[6]*r^2
534 // ((inp[1]*r^4+inp[3]*r^2+inp[5])*r^3+inp[7]*r
535 // \___________________/
536 // ((inp[0]*r^4+inp[2]*r^2+inp[4])*r^4+inp[6]*r^2+inp[8])*r^2
537 // ((inp[1]*r^4+inp[3]*r^2+inp[5])*r^4+inp[7]*r^2+inp[9])*r
538 // \___________________/ \____________________/
540 // Note that we start with inp[2:3]*r^2. This is because it
541 // doesn't depend on reduction in previous iteration.
542 ////////////////////////////////////////////////////////////////
543 // d4 = h0*r4 + h1*r3 + h2*r2 + h3*r1 + h4*r0
544 // d3 = h0*r3 + h1*r2 + h2*r1 + h3*r0 + h4*5*r4
545 // d2 = h0*r2 + h1*r1 + h2*r0 + h3*5*r4 + h4*5*r3
546 // d1 = h0*r1 + h1*r0 + h2*5*r4 + h3*5*r3 + h4*5*r2
547 // d0 = h0*r0 + h1*5*r4 + h2*5*r3 + h3*5*r2 + h4*5*r1
550 umull $ACC4,$IN23_0,${R4}[2]
551 csel $in2,$zeros,$in2,lo
552 umull $ACC3,$IN23_0,${R3}[2]
553 umull $ACC2,$IN23_0,${R2}[2]
554 ldp x8,x12,[$in2],#16 // inp[2:3] (or zero)
555 umull $ACC1,$IN23_0,${R1}[2]
556 ldp x9,x13,[$in2],#48
557 umull $ACC0,$IN23_0,${R0}[2]
565 umlal $ACC4,$IN23_1,${R3}[2]
566 and x4,x8,#0x03ffffff // base 2^64 -> base 2^26
567 umlal $ACC3,$IN23_1,${R2}[2]
568 and x5,x9,#0x03ffffff
569 umlal $ACC2,$IN23_1,${R1}[2]
571 umlal $ACC1,$IN23_1,${R0}[2]
573 umlal $ACC0,$IN23_1,${S4}[2]
574 add x4,x4,x5,lsl#32 // bfi x4,x5,#32,#32
576 umlal $ACC4,$IN23_2,${R2}[2]
578 umlal $ACC3,$IN23_2,${R1}[2]
580 umlal $ACC2,$IN23_2,${R0}[2]
581 add x6,x6,x7,lsl#32 // bfi x6,x7,#32,#32
582 umlal $ACC1,$IN23_2,${S4}[2]
584 umlal $ACC0,$IN23_2,${S3}[2]
585 and x8,x8,#0x03ffffff
587 umlal $ACC4,$IN23_3,${R1}[2]
588 and x9,x9,#0x03ffffff
589 umlal $ACC3,$IN23_3,${R0}[2]
591 umlal $ACC2,$IN23_3,${S4}[2]
593 umlal $ACC1,$IN23_3,${S3}[2]
594 add x8,x8,x9,lsl#32 // bfi x8,x9,#32,#32
595 umlal $ACC0,$IN23_3,${S2}[2]
598 add $IN01_2,$IN01_2,$H2
599 add x12,$padbit,x12,lsr#40
600 umlal $ACC4,$IN23_4,${R0}[2]
601 add x13,$padbit,x13,lsr#40
602 umlal $ACC3,$IN23_4,${S4}[2]
603 add x10,x10,x11,lsl#32 // bfi x10,x11,#32,#32
604 umlal $ACC2,$IN23_4,${S3}[2]
605 add x12,x12,x13,lsl#32 // bfi x12,x13,#32,#32
606 umlal $ACC1,$IN23_4,${S2}[2]
608 umlal $ACC0,$IN23_4,${S1}[2]
611 ////////////////////////////////////////////////////////////////
612 // (hash+inp[0:1])*r^4 and accumulate
614 add $IN01_0,$IN01_0,$H0
616 umlal $ACC3,$IN01_2,${R1}[0]
617 ldp x8,x12,[$inp],#16 // inp[0:1]
618 umlal $ACC0,$IN01_2,${S3}[0]
619 ldp x9,x13,[$inp],#48
620 umlal $ACC4,$IN01_2,${R2}[0]
621 umlal $ACC1,$IN01_2,${S4}[0]
622 umlal $ACC2,$IN01_2,${R0}[0]
630 add $IN01_1,$IN01_1,$H1
631 umlal $ACC3,$IN01_0,${R3}[0]
632 umlal $ACC4,$IN01_0,${R4}[0]
633 and x4,x8,#0x03ffffff // base 2^64 -> base 2^26
634 umlal $ACC2,$IN01_0,${R2}[0]
635 and x5,x9,#0x03ffffff
636 umlal $ACC0,$IN01_0,${R0}[0]
638 umlal $ACC1,$IN01_0,${R1}[0]
641 add $IN01_3,$IN01_3,$H3
642 add x4,x4,x5,lsl#32 // bfi x4,x5,#32,#32
643 umlal $ACC3,$IN01_1,${R2}[0]
645 umlal $ACC4,$IN01_1,${R3}[0]
647 umlal $ACC0,$IN01_1,${S4}[0]
648 add x6,x6,x7,lsl#32 // bfi x6,x7,#32,#32
649 umlal $ACC2,$IN01_1,${R1}[0]
651 umlal $ACC1,$IN01_1,${R0}[0]
652 and x8,x8,#0x03ffffff
654 add $IN01_4,$IN01_4,$H4
655 and x9,x9,#0x03ffffff
656 umlal $ACC3,$IN01_3,${R0}[0]
658 umlal $ACC0,$IN01_3,${S2}[0]
660 umlal $ACC4,$IN01_3,${R1}[0]
661 add x8,x8,x9,lsl#32 // bfi x8,x9,#32,#32
662 umlal $ACC1,$IN01_3,${S3}[0]
664 umlal $ACC2,$IN01_3,${S4}[0]
665 add x12,$padbit,x12,lsr#40
667 umlal $ACC3,$IN01_4,${S4}[0]
668 add x13,$padbit,x13,lsr#40
669 umlal $ACC0,$IN01_4,${S1}[0]
670 add x10,x10,x11,lsl#32 // bfi x10,x11,#32,#32
671 umlal $ACC4,$IN01_4,${R0}[0]
672 add x12,x12,x13,lsl#32 // bfi x12,x13,#32,#32
673 umlal $ACC1,$IN01_4,${S2}[0]
675 umlal $ACC2,$IN01_4,${S3}[0]
679 /////////////////////////////////////////////////////////////////
680 // lazy reduction as discussed in "NEON crypto" by D.J. Bernstein
683 // [see discussion in poly1305-armv4 module]
685 ushr $T0.2d,$ACC3,#26
687 ushr $T1.2d,$ACC0,#26
688 and $ACC0,$ACC0,$MASK.2d
689 add $ACC4,$ACC4,$T0.2d // h3 -> h4
690 bic $H3,#0xfc,lsl#24 // &=0x03ffffff
691 add $ACC1,$ACC1,$T1.2d // h0 -> h1
693 ushr $T0.2d,$ACC4,#26
695 ushr $T1.2d,$ACC1,#26
698 add $ACC2,$ACC2,$T1.2d // h1 -> h2
700 add $ACC0,$ACC0,$T0.2d
702 shrn $T1.2s,$ACC2,#26
704 add $ACC0,$ACC0,$T0.2d // h4 -> h0
706 add $H3,$H3,$T1.2s // h2 -> h3
709 shrn $T0.2s,$ACC0,#26
714 add $H1,$H1,$T0.2s // h0 -> h1
715 add $H4,$H4,$T1.2s // h3 -> h4
720 dup $IN23_2,${IN23_2}[0]
721 add $IN01_2,$IN01_2,$H2
723 ////////////////////////////////////////////////////////////////
724 // multiply (inp[0:1]+hash) or inp[2:3] by r^2:r^1
729 dup $IN23_2,${IN01_2}[0]
730 add $IN23_0,$IN01_0,$H0
731 add $IN23_3,$IN01_3,$H3
732 add $IN23_1,$IN01_1,$H1
733 add $IN23_4,$IN01_4,$H4
736 dup $IN23_0,${IN23_0}[0]
737 umull2 $ACC0,$IN23_2,${S3}
738 umull2 $ACC3,$IN23_2,${R1}
739 umull2 $ACC4,$IN23_2,${R2}
740 umull2 $ACC2,$IN23_2,${R0}
741 umull2 $ACC1,$IN23_2,${S4}
743 dup $IN23_1,${IN23_1}[0]
744 umlal2 $ACC0,$IN23_0,${R0}
745 umlal2 $ACC2,$IN23_0,${R2}
746 umlal2 $ACC3,$IN23_0,${R3}
747 umlal2 $ACC4,$IN23_0,${R4}
748 umlal2 $ACC1,$IN23_0,${R1}
750 dup $IN23_3,${IN23_3}[0]
751 umlal2 $ACC0,$IN23_1,${S4}
752 umlal2 $ACC3,$IN23_1,${R2}
753 umlal2 $ACC2,$IN23_1,${R1}
754 umlal2 $ACC4,$IN23_1,${R3}
755 umlal2 $ACC1,$IN23_1,${R0}
757 dup $IN23_4,${IN23_4}[0]
758 umlal2 $ACC3,$IN23_3,${R0}
759 umlal2 $ACC4,$IN23_3,${R1}
760 umlal2 $ACC0,$IN23_3,${S2}
761 umlal2 $ACC1,$IN23_3,${S3}
762 umlal2 $ACC2,$IN23_3,${S4}
764 umlal2 $ACC3,$IN23_4,${S4}
765 umlal2 $ACC0,$IN23_4,${S1}
766 umlal2 $ACC4,$IN23_4,${R0}
767 umlal2 $ACC1,$IN23_4,${S2}
768 umlal2 $ACC2,$IN23_4,${S3}
772 ////////////////////////////////////////////////////////////////
773 // (hash+inp[0:1])*r^4:r^3 and accumulate
775 add $IN01_0,$IN01_0,$H0
776 umlal $ACC3,$IN01_2,${R1}
777 umlal $ACC0,$IN01_2,${S3}
778 umlal $ACC4,$IN01_2,${R2}
779 umlal $ACC1,$IN01_2,${S4}
780 umlal $ACC2,$IN01_2,${R0}
782 add $IN01_1,$IN01_1,$H1
783 umlal $ACC3,$IN01_0,${R3}
784 umlal $ACC0,$IN01_0,${R0}
785 umlal $ACC4,$IN01_0,${R4}
786 umlal $ACC1,$IN01_0,${R1}
787 umlal $ACC2,$IN01_0,${R2}
789 add $IN01_3,$IN01_3,$H3
790 umlal $ACC3,$IN01_1,${R2}
791 umlal $ACC0,$IN01_1,${S4}
792 umlal $ACC4,$IN01_1,${R3}
793 umlal $ACC1,$IN01_1,${R0}
794 umlal $ACC2,$IN01_1,${R1}
796 add $IN01_4,$IN01_4,$H4
797 umlal $ACC3,$IN01_3,${R0}
798 umlal $ACC0,$IN01_3,${S2}
799 umlal $ACC4,$IN01_3,${R1}
800 umlal $ACC1,$IN01_3,${S3}
801 umlal $ACC2,$IN01_3,${S4}
803 umlal $ACC3,$IN01_4,${S4}
804 umlal $ACC0,$IN01_4,${S1}
805 umlal $ACC4,$IN01_4,${R0}
806 umlal $ACC1,$IN01_4,${S2}
807 umlal $ACC2,$IN01_4,${S3}
810 ////////////////////////////////////////////////////////////////
813 addp $ACC3,$ACC3,$ACC3
814 ldp d8,d9,[sp,#16] // meet ABI requirements
815 addp $ACC0,$ACC0,$ACC0
817 addp $ACC4,$ACC4,$ACC4
819 addp $ACC1,$ACC1,$ACC1
821 addp $ACC2,$ACC2,$ACC2
823 ////////////////////////////////////////////////////////////////
824 // lazy reduction, but without narrowing
826 ushr $T0.2d,$ACC3,#26
827 and $ACC3,$ACC3,$MASK.2d
828 ushr $T1.2d,$ACC0,#26
829 and $ACC0,$ACC0,$MASK.2d
831 add $ACC4,$ACC4,$T0.2d // h3 -> h4
832 add $ACC1,$ACC1,$T1.2d // h0 -> h1
834 ushr $T0.2d,$ACC4,#26
835 and $ACC4,$ACC4,$MASK.2d
836 ushr $T1.2d,$ACC1,#26
837 and $ACC1,$ACC1,$MASK.2d
838 add $ACC2,$ACC2,$T1.2d // h1 -> h2
840 add $ACC0,$ACC0,$T0.2d
842 ushr $T1.2d,$ACC2,#26
843 and $ACC2,$ACC2,$MASK.2d
844 add $ACC0,$ACC0,$T0.2d // h4 -> h0
845 add $ACC3,$ACC3,$T1.2d // h2 -> h3
847 ushr $T0.2d,$ACC0,#26
848 and $ACC0,$ACC0,$MASK.2d
849 ushr $T1.2d,$ACC3,#26
850 and $ACC3,$ACC3,$MASK.2d
851 add $ACC1,$ACC1,$T0.2d // h0 -> h1
852 add $ACC4,$ACC4,$T1.2d // h3 -> h4
854 ////////////////////////////////////////////////////////////////
855 // write the result, can be partially reduced
857 st4 {$ACC0,$ACC1,$ACC2,$ACC3}[0],[$ctx],#16
858 st1 {$ACC4}[0],[$ctx]
863 .size poly1305_blocks_neon,.-poly1305_blocks_neon
865 .type poly1305_emit_neon,%function
868 ldr $is_base2_26,[$ctx,#24]
869 cbz $is_base2_26,poly1305_emit
871 ldp w10,w11,[$ctx] // load hash value base 2^26
872 ldp w12,w13,[$ctx,#8]
875 add $h0,x10,x11,lsl#26 // base 2^26 -> base 2^64
877 adds $h0,$h0,x12,lsl#52
878 add $h1,$h1,x13,lsl#14
881 adds $h1,$h1,x14,lsl#40
882 adc $h2,$h2,xzr // can be partially reduced...
884 ldp $t0,$t1,[$nonce] // load nonce
886 and $d0,$h2,#-4 // ... so reduce
887 add $d0,$d0,$h2,lsr#2
893 adds $d0,$h0,#5 // compare to modulus
897 tst $d2,#-4 // see if it's carried/borrowed
903 ror $t0,$t0,#32 // flip nonce words
906 adds $h0,$h0,$t0 // accumulate nonce
909 rev $h0,$h0 // flip output bytes
912 stp $h0,$h1,[$mac] // write result
915 .size poly1305_emit_neon,.-poly1305_emit_neon
919 .long 0,0,0,0,0,0,0,0
922 .long OPENSSL_armcap_P-.
924 .quad OPENSSL_armcap_P-.
926 .asciz "Poly1305 for ARMv8, CRYPTOGAMS by <appro\@openssl.org>"
930 foreach (split("\n",$code)) {
931 s/\b(shrn\s+v[0-9]+)\.[24]d/$1.2s/ or
932 s/\b(fmov\s+)v([0-9]+)[^,]*,\s*x([0-9]+)/$1d$2,x$3/ or
933 (m/\bdup\b/ and (s/\.[24]s/.2d/g or 1)) or
934 (m/\b(eor|and)/ and (s/\.[248][sdh]/.16b/g or 1)) or
935 (m/\bum(ul|la)l\b/ and (s/\.4s/.2s/g or 1)) or
936 (m/\bum(ul|la)l2\b/ and (s/\.2s/.4s/g or 1)) or
937 (m/\bst[1-4]\s+{[^}]+}\[/ and (s/\.[24]d/.s/g or 1));
939 s/\.[124]([sd])\[/.$1\[/;