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
33 # (*) estimate based on resources availability is less than 1.0,
34 # i.e. measured result is worse than expected, presumably binary
35 # translator is not almighty;
40 $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
41 ( $xlate="${dir}arm-xlate.pl" and -f $xlate ) or
42 ( $xlate="${dir}../../perlasm/arm-xlate.pl" and -f $xlate) or
43 die "can't locate arm-xlate.pl";
45 open OUT,"| \"$^X\" $xlate $flavour $output";
48 my ($ctx,$inp,$len,$padbit) = map("x$_",(0..3));
49 my ($mac,$nonce)=($inp,$len);
51 my ($h0,$h1,$h2,$r0,$r1,$s1,$t0,$t1,$d0,$d1,$d2) = map("x$_",(4..14));
58 // forward "declarations" are required for Apple
59 .extern OPENSSL_armcap_P
60 .globl poly1305_blocks
64 .type poly1305_init,%function
68 stp xzr,xzr,[$ctx] // zero hash value
69 stp xzr,xzr,[$ctx,#16] // [along with is_base2_26]
75 ldrsw $t1,.LOPENSSL_armcap_P
77 ldr $t1,.LOPENSSL_armcap_P
79 adr $t0,.LOPENSSL_armcap_P
81 ldp $r0,$r1,[$inp] // load key
82 mov $s1,#0xfffffffc0fffffff
83 movk $s1,#0x0fff,lsl#48
86 rev $r0,$r0 // flip bytes
89 and $r0,$r0,$s1 // &=0ffffffc0fffffff
91 and $r1,$r1,$s1 // &=0ffffffc0ffffffc
92 stp $r0,$r1,[$ctx,#32] // save key value
96 adr $d0,poly1305_blocks
97 adr $r0,poly1305_blocks_neon
99 adr $r1,poly1305_emit_neon
113 .size poly1305_init,.-poly1305_init
115 .type poly1305_blocks,%function
121 ldp $h0,$h1,[$ctx] // load hash value
122 ldp $r0,$r1,[$ctx,#32] // load key value
124 add $s1,$r1,$r1,lsr#2 // s1 = r1 + (r1 >> 2)
129 ldp $t0,$t1,[$inp],#16 // load input
135 adds $h0,$h0,$t0 // accumulate input
138 mul $d0,$h0,$r0 // h0*r0
142 mul $t0,$h1,$s1 // h1*5*r1
146 mul $t0,$h0,$r1 // h0*r1
151 mul $t0,$h1,$r0 // h1*r0
156 mul $t0,$h2,$s1 // h2*5*r1
158 mul $t1,$h2,$r0 // h2*r0
163 and $t0,$d2,#-4 // final reduction
165 add $t0,$t0,$d2,lsr#2
172 stp $h0,$h1,[$ctx] // store hash value
177 .size poly1305_blocks,.-poly1305_blocks
179 .type poly1305_emit,%function
182 ldp $h0,$h1,[$ctx] // load hash base 2^64
184 ldp $t0,$t1,[$nonce] // load nonce
186 adds $d0,$h0,#5 // compare to modulus
190 tst $d2,#-4 // see if it's carried/borrowed
196 ror $t0,$t0,#32 // flip nonce words
199 adds $h0,$h0,$t0 // accumulate nonce
202 rev $h0,$h0 // flip output bytes
205 stp $h0,$h1,[$mac] // write result
208 .size poly1305_emit,.-poly1305_emit
210 my ($R0,$R1,$S1,$R2,$S2,$R3,$S3,$R4,$S4) = map("v$_.4s",(0..8));
211 my ($IN01_0,$IN01_1,$IN01_2,$IN01_3,$IN01_4) = map("v$_.2s",(9..13));
212 my ($IN23_0,$IN23_1,$IN23_2,$IN23_3,$IN23_4) = map("v$_.2s",(14..18));
213 my ($ACC0,$ACC1,$ACC2,$ACC3,$ACC4) = map("v$_.2d",(19..23));
214 my ($H0,$H1,$H2,$H3,$H4) = map("v$_.2s",(24..28));
215 my ($T0,$T1,$MASK) = map("v$_",(29..31));
217 my ($in2,$zeros)=("x16","x17");
218 my $is_base2_26 = $zeros; # borrow
221 .type poly1305_mult,%function
224 mul $d0,$h0,$r0 // h0*r0
227 mul $t0,$h1,$s1 // h1*5*r1
231 mul $t0,$h0,$r1 // h0*r1
236 mul $t0,$h1,$r0 // h1*r0
241 mul $t0,$h2,$s1 // h2*5*r1
243 mul $t1,$h2,$r0 // h2*r0
248 and $t0,$d2,#-4 // final reduction
250 add $t0,$t0,$d2,lsr#2
256 .size poly1305_mult,.-poly1305_mult
258 .type poly1305_splat,%function
261 and x12,$h0,#0x03ffffff // base 2^64 -> base 2^26
264 and x14,x14,#0x03ffffff
268 str w12,[$ctx,#16*0] // r0
269 add w12,w13,w13,lsl#2 // r1*5
270 str w13,[$ctx,#16*1] // r1
271 add w13,w14,w14,lsl#2 // r2*5
272 str w12,[$ctx,#16*2] // s1
273 str w14,[$ctx,#16*3] // r2
274 add w14,w15,w15,lsl#2 // r3*5
275 str w13,[$ctx,#16*4] // s2
276 str w15,[$ctx,#16*5] // r3
277 add w15,w16,w16,lsl#2 // r4*5
278 str w14,[$ctx,#16*6] // s3
279 str w16,[$ctx,#16*7] // r4
280 str w15,[$ctx,#16*8] // s4
283 .size poly1305_splat,.-poly1305_splat
285 .type poly1305_blocks_neon,%function
287 poly1305_blocks_neon:
288 ldr $is_base2_26,[$ctx,#24]
291 cbz $is_base2_26,poly1305_blocks
294 .inst 0xd503233f // paciasp
295 stp x29,x30,[sp,#-80]!
301 cbz $is_base2_26,.Lbase2_64_neon
303 ldp w10,w11,[$ctx] // load hash value base 2^26
304 ldp w12,w13,[$ctx,#8]
310 ldp $r0,$r1,[$ctx,#32] // load key value
312 add $h0,x10,x11,lsl#26 // base 2^26 -> base 2^64
314 adds $h0,$h0,x12,lsl#52
315 add $h1,$h1,x13,lsl#14
318 adds $h1,$h1,x14,lsl#40
319 adc $d2,$h2,xzr // can be partially reduced...
321 ldp $d0,$d1,[$inp],#16 // load input
323 add $s1,$r1,$r1,lsr#2 // s1 = r1 + (r1 >> 2)
325 and $t0,$d2,#-4 // ... so reduce
327 add $t0,$t0,$d2,lsr#2
336 adds $h0,$h0,$d0 // accumulate input
343 cbz $padbit,.Lstore_base2_64_neon
345 and x10,$h0,#0x03ffffff // base 2^64 -> base 2^26
348 and x12,x12,#0x03ffffff
352 cbnz $len,.Leven_neon
354 stp w10,w11,[$ctx] // store hash value base 2^26
355 stp w12,w13,[$ctx,#8]
360 .Lstore_base2_64_neon:
361 stp $h0,$h1,[$ctx] // store hash value base 2^64
362 stp $h2,xzr,[$ctx,#16] // note that is_base2_26 is zeroed
367 ldp $r0,$r1,[$ctx,#32] // load key value
369 ldp $h0,$h1,[$ctx] // load hash value base 2^64
375 ldp $d0,$d1,[$inp],#16 // load input
377 add $s1,$r1,$r1,lsr#2 // s1 = r1 + (r1 >> 2)
382 adds $h0,$h0,$d0 // accumulate input
389 and x10,$h0,#0x03ffffff // base 2^64 -> base 2^26
392 and x12,x12,#0x03ffffff
396 stp d8,d9,[sp,#16] // meet ABI requirements
407 ////////////////////////////////// initialize r^n table
409 add $s1,$r1,$r1,lsr#2 // s1 = r1 + (r1 >> 2)
415 bl poly1305_mult // r^2
419 bl poly1305_mult // r^3
423 bl poly1305_mult // r^4
431 csel $in2,$zeros,$in2,lo
434 str x4,[$ctx,#-24] // set is_base2_26
435 sub $ctx,$ctx,#48 // restore original $ctx
443 csel $in2,$zeros,$in2,lo
445 stp d8,d9,[sp,#16] // meet ABI requirements
457 ldp x8,x12,[$in2],#16 // inp[2:3] (or zero)
458 ldp x9,x13,[$in2],#48
460 lsl $padbit,$padbit,#24
469 and x4,x8,#0x03ffffff // base 2^64 -> base 2^26
470 and x5,x9,#0x03ffffff
473 add x4,x4,x5,lsl#32 // bfi x4,x5,#32,#32
476 add x6,x6,x7,lsl#32 // bfi x6,x7,#32,#32
478 and x8,x8,#0x03ffffff
479 and x9,x9,#0x03ffffff
482 add x12,$padbit,x12,lsr#40
483 add x13,$padbit,x13,lsr#40
484 add x8,x8,x9,lsl#32 // bfi x8,x9,#32,#32
486 add x10,x10,x11,lsl#32 // bfi x10,x11,#32,#32
487 add x12,x12,x13,lsl#32 // bfi x12,x13,#32,#32
492 ldp x8,x12,[$inp],#16 // inp[0:1]
493 ldp x9,x13,[$inp],#48
495 ld1 {$R0,$R1,$S1,$R2},[x15],#64
496 ld1 {$S2,$R3,$S3,$R4},[x15],#64
505 and x4,x8,#0x03ffffff // base 2^64 -> base 2^26
506 and x5,x9,#0x03ffffff
509 add x4,x4,x5,lsl#32 // bfi x4,x5,#32,#32
512 add x6,x6,x7,lsl#32 // bfi x6,x7,#32,#32
514 and x8,x8,#0x03ffffff
515 and x9,x9,#0x03ffffff
518 add x12,$padbit,x12,lsr#40
519 add x13,$padbit,x13,lsr#40
520 add x8,x8,x9,lsl#32 // bfi x8,x9,#32,#32
522 add x10,x10,x11,lsl#32 // bfi x10,x11,#32,#32
523 add x12,x12,x13,lsl#32 // bfi x12,x13,#32,#32
528 ushr $MASK.2d,$MASK.2d,#38
534 ////////////////////////////////////////////////////////////////
535 // ((inp[0]*r^4+inp[2]*r^2+inp[4])*r^4+inp[6]*r^2
536 // ((inp[1]*r^4+inp[3]*r^2+inp[5])*r^3+inp[7]*r
537 // \___________________/
538 // ((inp[0]*r^4+inp[2]*r^2+inp[4])*r^4+inp[6]*r^2+inp[8])*r^2
539 // ((inp[1]*r^4+inp[3]*r^2+inp[5])*r^4+inp[7]*r^2+inp[9])*r
540 // \___________________/ \____________________/
542 // Note that we start with inp[2:3]*r^2. This is because it
543 // doesn't depend on reduction in previous iteration.
544 ////////////////////////////////////////////////////////////////
545 // d4 = h0*r4 + h1*r3 + h2*r2 + h3*r1 + h4*r0
546 // d3 = h0*r3 + h1*r2 + h2*r1 + h3*r0 + h4*5*r4
547 // d2 = h0*r2 + h1*r1 + h2*r0 + h3*5*r4 + h4*5*r3
548 // d1 = h0*r1 + h1*r0 + h2*5*r4 + h3*5*r3 + h4*5*r2
549 // d0 = h0*r0 + h1*5*r4 + h2*5*r3 + h3*5*r2 + h4*5*r1
552 umull $ACC4,$IN23_0,${R4}[2]
553 csel $in2,$zeros,$in2,lo
554 umull $ACC3,$IN23_0,${R3}[2]
555 umull $ACC2,$IN23_0,${R2}[2]
556 ldp x8,x12,[$in2],#16 // inp[2:3] (or zero)
557 umull $ACC1,$IN23_0,${R1}[2]
558 ldp x9,x13,[$in2],#48
559 umull $ACC0,$IN23_0,${R0}[2]
567 umlal $ACC4,$IN23_1,${R3}[2]
568 and x4,x8,#0x03ffffff // base 2^64 -> base 2^26
569 umlal $ACC3,$IN23_1,${R2}[2]
570 and x5,x9,#0x03ffffff
571 umlal $ACC2,$IN23_1,${R1}[2]
573 umlal $ACC1,$IN23_1,${R0}[2]
575 umlal $ACC0,$IN23_1,${S4}[2]
576 add x4,x4,x5,lsl#32 // bfi x4,x5,#32,#32
578 umlal $ACC4,$IN23_2,${R2}[2]
580 umlal $ACC3,$IN23_2,${R1}[2]
582 umlal $ACC2,$IN23_2,${R0}[2]
583 add x6,x6,x7,lsl#32 // bfi x6,x7,#32,#32
584 umlal $ACC1,$IN23_2,${S4}[2]
586 umlal $ACC0,$IN23_2,${S3}[2]
587 and x8,x8,#0x03ffffff
589 umlal $ACC4,$IN23_3,${R1}[2]
590 and x9,x9,#0x03ffffff
591 umlal $ACC3,$IN23_3,${R0}[2]
593 umlal $ACC2,$IN23_3,${S4}[2]
595 umlal $ACC1,$IN23_3,${S3}[2]
596 add x8,x8,x9,lsl#32 // bfi x8,x9,#32,#32
597 umlal $ACC0,$IN23_3,${S2}[2]
600 add $IN01_2,$IN01_2,$H2
601 add x12,$padbit,x12,lsr#40
602 umlal $ACC4,$IN23_4,${R0}[2]
603 add x13,$padbit,x13,lsr#40
604 umlal $ACC3,$IN23_4,${S4}[2]
605 add x10,x10,x11,lsl#32 // bfi x10,x11,#32,#32
606 umlal $ACC2,$IN23_4,${S3}[2]
607 add x12,x12,x13,lsl#32 // bfi x12,x13,#32,#32
608 umlal $ACC1,$IN23_4,${S2}[2]
610 umlal $ACC0,$IN23_4,${S1}[2]
613 ////////////////////////////////////////////////////////////////
614 // (hash+inp[0:1])*r^4 and accumulate
616 add $IN01_0,$IN01_0,$H0
618 umlal $ACC3,$IN01_2,${R1}[0]
619 ldp x8,x12,[$inp],#16 // inp[0:1]
620 umlal $ACC0,$IN01_2,${S3}[0]
621 ldp x9,x13,[$inp],#48
622 umlal $ACC4,$IN01_2,${R2}[0]
623 umlal $ACC1,$IN01_2,${S4}[0]
624 umlal $ACC2,$IN01_2,${R0}[0]
632 add $IN01_1,$IN01_1,$H1
633 umlal $ACC3,$IN01_0,${R3}[0]
634 umlal $ACC4,$IN01_0,${R4}[0]
635 and x4,x8,#0x03ffffff // base 2^64 -> base 2^26
636 umlal $ACC2,$IN01_0,${R2}[0]
637 and x5,x9,#0x03ffffff
638 umlal $ACC0,$IN01_0,${R0}[0]
640 umlal $ACC1,$IN01_0,${R1}[0]
643 add $IN01_3,$IN01_3,$H3
644 add x4,x4,x5,lsl#32 // bfi x4,x5,#32,#32
645 umlal $ACC3,$IN01_1,${R2}[0]
647 umlal $ACC4,$IN01_1,${R3}[0]
649 umlal $ACC0,$IN01_1,${S4}[0]
650 add x6,x6,x7,lsl#32 // bfi x6,x7,#32,#32
651 umlal $ACC2,$IN01_1,${R1}[0]
653 umlal $ACC1,$IN01_1,${R0}[0]
654 and x8,x8,#0x03ffffff
656 add $IN01_4,$IN01_4,$H4
657 and x9,x9,#0x03ffffff
658 umlal $ACC3,$IN01_3,${R0}[0]
660 umlal $ACC0,$IN01_3,${S2}[0]
662 umlal $ACC4,$IN01_3,${R1}[0]
663 add x8,x8,x9,lsl#32 // bfi x8,x9,#32,#32
664 umlal $ACC1,$IN01_3,${S3}[0]
666 umlal $ACC2,$IN01_3,${S4}[0]
667 add x12,$padbit,x12,lsr#40
669 umlal $ACC3,$IN01_4,${S4}[0]
670 add x13,$padbit,x13,lsr#40
671 umlal $ACC0,$IN01_4,${S1}[0]
672 add x10,x10,x11,lsl#32 // bfi x10,x11,#32,#32
673 umlal $ACC4,$IN01_4,${R0}[0]
674 add x12,x12,x13,lsl#32 // bfi x12,x13,#32,#32
675 umlal $ACC1,$IN01_4,${S2}[0]
677 umlal $ACC2,$IN01_4,${S3}[0]
681 /////////////////////////////////////////////////////////////////
682 // lazy reduction as discussed in "NEON crypto" by D.J. Bernstein
685 // [see discussion in poly1305-armv4 module]
687 ushr $T0.2d,$ACC3,#26
689 ushr $T1.2d,$ACC0,#26
690 and $ACC0,$ACC0,$MASK.2d
691 add $ACC4,$ACC4,$T0.2d // h3 -> h4
692 bic $H3,#0xfc,lsl#24 // &=0x03ffffff
693 add $ACC1,$ACC1,$T1.2d // h0 -> h1
695 ushr $T0.2d,$ACC4,#26
697 ushr $T1.2d,$ACC1,#26
700 add $ACC2,$ACC2,$T1.2d // h1 -> h2
702 add $ACC0,$ACC0,$T0.2d
704 shrn $T1.2s,$ACC2,#26
706 add $ACC0,$ACC0,$T0.2d // h4 -> h0
708 add $H3,$H3,$T1.2s // h2 -> h3
711 shrn $T0.2s,$ACC0,#26
716 add $H1,$H1,$T0.2s // h0 -> h1
717 add $H4,$H4,$T1.2s // h3 -> h4
722 dup $IN23_2,${IN23_2}[0]
723 add $IN01_2,$IN01_2,$H2
725 ////////////////////////////////////////////////////////////////
726 // multiply (inp[0:1]+hash) or inp[2:3] by r^2:r^1
731 dup $IN23_2,${IN01_2}[0]
732 add $IN23_0,$IN01_0,$H0
733 add $IN23_3,$IN01_3,$H3
734 add $IN23_1,$IN01_1,$H1
735 add $IN23_4,$IN01_4,$H4
738 dup $IN23_0,${IN23_0}[0]
739 umull2 $ACC0,$IN23_2,${S3}
740 umull2 $ACC3,$IN23_2,${R1}
741 umull2 $ACC4,$IN23_2,${R2}
742 umull2 $ACC2,$IN23_2,${R0}
743 umull2 $ACC1,$IN23_2,${S4}
745 dup $IN23_1,${IN23_1}[0]
746 umlal2 $ACC0,$IN23_0,${R0}
747 umlal2 $ACC2,$IN23_0,${R2}
748 umlal2 $ACC3,$IN23_0,${R3}
749 umlal2 $ACC4,$IN23_0,${R4}
750 umlal2 $ACC1,$IN23_0,${R1}
752 dup $IN23_3,${IN23_3}[0]
753 umlal2 $ACC0,$IN23_1,${S4}
754 umlal2 $ACC3,$IN23_1,${R2}
755 umlal2 $ACC2,$IN23_1,${R1}
756 umlal2 $ACC4,$IN23_1,${R3}
757 umlal2 $ACC1,$IN23_1,${R0}
759 dup $IN23_4,${IN23_4}[0]
760 umlal2 $ACC3,$IN23_3,${R0}
761 umlal2 $ACC4,$IN23_3,${R1}
762 umlal2 $ACC0,$IN23_3,${S2}
763 umlal2 $ACC1,$IN23_3,${S3}
764 umlal2 $ACC2,$IN23_3,${S4}
766 umlal2 $ACC3,$IN23_4,${S4}
767 umlal2 $ACC0,$IN23_4,${S1}
768 umlal2 $ACC4,$IN23_4,${R0}
769 umlal2 $ACC1,$IN23_4,${S2}
770 umlal2 $ACC2,$IN23_4,${S3}
774 ////////////////////////////////////////////////////////////////
775 // (hash+inp[0:1])*r^4:r^3 and accumulate
777 add $IN01_0,$IN01_0,$H0
778 umlal $ACC3,$IN01_2,${R1}
779 umlal $ACC0,$IN01_2,${S3}
780 umlal $ACC4,$IN01_2,${R2}
781 umlal $ACC1,$IN01_2,${S4}
782 umlal $ACC2,$IN01_2,${R0}
784 add $IN01_1,$IN01_1,$H1
785 umlal $ACC3,$IN01_0,${R3}
786 umlal $ACC0,$IN01_0,${R0}
787 umlal $ACC4,$IN01_0,${R4}
788 umlal $ACC1,$IN01_0,${R1}
789 umlal $ACC2,$IN01_0,${R2}
791 add $IN01_3,$IN01_3,$H3
792 umlal $ACC3,$IN01_1,${R2}
793 umlal $ACC0,$IN01_1,${S4}
794 umlal $ACC4,$IN01_1,${R3}
795 umlal $ACC1,$IN01_1,${R0}
796 umlal $ACC2,$IN01_1,${R1}
798 add $IN01_4,$IN01_4,$H4
799 umlal $ACC3,$IN01_3,${R0}
800 umlal $ACC0,$IN01_3,${S2}
801 umlal $ACC4,$IN01_3,${R1}
802 umlal $ACC1,$IN01_3,${S3}
803 umlal $ACC2,$IN01_3,${S4}
805 umlal $ACC3,$IN01_4,${S4}
806 umlal $ACC0,$IN01_4,${S1}
807 umlal $ACC4,$IN01_4,${R0}
808 umlal $ACC1,$IN01_4,${S2}
809 umlal $ACC2,$IN01_4,${S3}
812 ////////////////////////////////////////////////////////////////
815 addp $ACC3,$ACC3,$ACC3
816 ldp d8,d9,[sp,#16] // meet ABI requirements
817 addp $ACC0,$ACC0,$ACC0
819 addp $ACC4,$ACC4,$ACC4
821 addp $ACC1,$ACC1,$ACC1
823 addp $ACC2,$ACC2,$ACC2
825 ////////////////////////////////////////////////////////////////
826 // lazy reduction, but without narrowing
828 ushr $T0.2d,$ACC3,#26
829 and $ACC3,$ACC3,$MASK.2d
830 ushr $T1.2d,$ACC0,#26
831 and $ACC0,$ACC0,$MASK.2d
833 add $ACC4,$ACC4,$T0.2d // h3 -> h4
834 add $ACC1,$ACC1,$T1.2d // h0 -> h1
836 ushr $T0.2d,$ACC4,#26
837 and $ACC4,$ACC4,$MASK.2d
838 ushr $T1.2d,$ACC1,#26
839 and $ACC1,$ACC1,$MASK.2d
840 add $ACC2,$ACC2,$T1.2d // h1 -> h2
842 add $ACC0,$ACC0,$T0.2d
844 ushr $T1.2d,$ACC2,#26
845 and $ACC2,$ACC2,$MASK.2d
846 add $ACC0,$ACC0,$T0.2d // h4 -> h0
847 add $ACC3,$ACC3,$T1.2d // h2 -> h3
849 ushr $T0.2d,$ACC0,#26
850 and $ACC0,$ACC0,$MASK.2d
851 ushr $T1.2d,$ACC3,#26
852 and $ACC3,$ACC3,$MASK.2d
853 add $ACC1,$ACC1,$T0.2d // h0 -> h1
854 add $ACC4,$ACC4,$T1.2d // h3 -> h4
856 ////////////////////////////////////////////////////////////////
857 // write the result, can be partially reduced
859 st4 {$ACC0,$ACC1,$ACC2,$ACC3}[0],[$ctx],#16
860 st1 {$ACC4}[0],[$ctx]
863 .inst 0xd50323bf // autiasp
866 .size poly1305_blocks_neon,.-poly1305_blocks_neon
868 .type poly1305_emit_neon,%function
871 ldr $is_base2_26,[$ctx,#24]
872 cbz $is_base2_26,poly1305_emit
874 ldp w10,w11,[$ctx] // load hash value base 2^26
875 ldp w12,w13,[$ctx,#8]
878 add $h0,x10,x11,lsl#26 // base 2^26 -> base 2^64
880 adds $h0,$h0,x12,lsl#52
881 add $h1,$h1,x13,lsl#14
884 adds $h1,$h1,x14,lsl#40
885 adc $h2,$h2,xzr // can be partially reduced...
887 ldp $t0,$t1,[$nonce] // load nonce
889 and $d0,$h2,#-4 // ... so reduce
890 add $d0,$d0,$h2,lsr#2
896 adds $d0,$h0,#5 // compare to modulus
900 tst $d2,#-4 // see if it's carried/borrowed
906 ror $t0,$t0,#32 // flip nonce words
909 adds $h0,$h0,$t0 // accumulate nonce
912 rev $h0,$h0 // flip output bytes
915 stp $h0,$h1,[$mac] // write result
918 .size poly1305_emit_neon,.-poly1305_emit_neon
922 .long 0,0,0,0,0,0,0,0
925 .long OPENSSL_armcap_P-.
927 .quad OPENSSL_armcap_P-.
929 .asciz "Poly1305 for ARMv8, CRYPTOGAMS by <appro\@openssl.org>"
933 foreach (split("\n",$code)) {
934 s/\b(shrn\s+v[0-9]+)\.[24]d/$1.2s/ or
935 s/\b(fmov\s+)v([0-9]+)[^,]*,\s*x([0-9]+)/$1d$2,x$3/ or
936 (m/\bdup\b/ and (s/\.[24]s/.2d/g or 1)) or
937 (m/\b(eor|and)/ and (s/\.[248][sdh]/.16b/g or 1)) or
938 (m/\bum(ul|la)l\b/ and (s/\.4s/.2s/g or 1)) or
939 (m/\bum(ul|la)l2\b/ and (s/\.2s/.4s/g or 1)) or
940 (m/\bst[1-4]\s+{[^}]+}\[/ and (s/\.[24]d/.s/g or 1));
942 s/\.[124]([sd])\[/.$1\[/;