2 # Copyright 2016-2020 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 .hidden OPENSSL_armcap_P
63 .globl poly1305_blocks
64 .hidden poly1305_blocks
68 .type poly1305_init,%function
72 stp xzr,xzr,[$ctx] // zero hash value
73 stp xzr,xzr,[$ctx,#16] // [along with is_base2_26]
79 ldrsw $t1,.LOPENSSL_armcap_P
81 ldr $t1,.LOPENSSL_armcap_P
83 adr $t0,.LOPENSSL_armcap_P
85 ldp $r0,$r1,[$inp] // load key
86 mov $s1,#0xfffffffc0fffffff
87 movk $s1,#0x0fff,lsl#48
90 rev $r0,$r0 // flip bytes
93 and $r0,$r0,$s1 // &=0ffffffc0fffffff
95 and $r1,$r1,$s1 // &=0ffffffc0ffffffc
96 stp $r0,$r1,[$ctx,#32] // save key value
100 adr $d0,poly1305_blocks
101 adr $r0,poly1305_blocks_neon
102 adr $d1,poly1305_emit
103 adr $r1,poly1305_emit_neon
117 .size poly1305_init,.-poly1305_init
119 .type poly1305_blocks,%function
125 ldp $h0,$h1,[$ctx] // load hash value
126 ldp $r0,$r1,[$ctx,#32] // load key value
128 add $s1,$r1,$r1,lsr#2 // s1 = r1 + (r1 >> 2)
133 ldp $t0,$t1,[$inp],#16 // load input
139 adds $h0,$h0,$t0 // accumulate input
142 mul $d0,$h0,$r0 // h0*r0
146 mul $t0,$h1,$s1 // h1*5*r1
150 mul $t0,$h0,$r1 // h0*r1
155 mul $t0,$h1,$r0 // h1*r0
160 mul $t0,$h2,$s1 // h2*5*r1
162 mul $t1,$h2,$r0 // h2*r0
167 and $t0,$d2,#-4 // final reduction
169 add $t0,$t0,$d2,lsr#2
176 stp $h0,$h1,[$ctx] // store hash value
181 .size poly1305_blocks,.-poly1305_blocks
183 .type poly1305_emit,%function
186 ldp $h0,$h1,[$ctx] // load hash base 2^64
188 ldp $t0,$t1,[$nonce] // load nonce
190 adds $d0,$h0,#5 // compare to modulus
194 tst $d2,#-4 // see if it's carried/borrowed
200 ror $t0,$t0,#32 // flip nonce words
203 adds $h0,$h0,$t0 // accumulate nonce
206 rev $h0,$h0 // flip output bytes
209 stp $h0,$h1,[$mac] // write result
212 .size poly1305_emit,.-poly1305_emit
214 my ($R0,$R1,$S1,$R2,$S2,$R3,$S3,$R4,$S4) = map("v$_.4s",(0..8));
215 my ($IN01_0,$IN01_1,$IN01_2,$IN01_3,$IN01_4) = map("v$_.2s",(9..13));
216 my ($IN23_0,$IN23_1,$IN23_2,$IN23_3,$IN23_4) = map("v$_.2s",(14..18));
217 my ($ACC0,$ACC1,$ACC2,$ACC3,$ACC4) = map("v$_.2d",(19..23));
218 my ($H0,$H1,$H2,$H3,$H4) = map("v$_.2s",(24..28));
219 my ($T0,$T1,$MASK) = map("v$_",(29..31));
221 my ($in2,$zeros)=("x16","x17");
222 my $is_base2_26 = $zeros; # borrow
225 .type poly1305_mult,%function
228 mul $d0,$h0,$r0 // h0*r0
231 mul $t0,$h1,$s1 // h1*5*r1
235 mul $t0,$h0,$r1 // h0*r1
240 mul $t0,$h1,$r0 // h1*r0
245 mul $t0,$h2,$s1 // h2*5*r1
247 mul $t1,$h2,$r0 // h2*r0
252 and $t0,$d2,#-4 // final reduction
254 add $t0,$t0,$d2,lsr#2
260 .size poly1305_mult,.-poly1305_mult
262 .type poly1305_splat,%function
265 and x12,$h0,#0x03ffffff // base 2^64 -> base 2^26
268 and x14,x14,#0x03ffffff
272 str w12,[$ctx,#16*0] // r0
273 add w12,w13,w13,lsl#2 // r1*5
274 str w13,[$ctx,#16*1] // r1
275 add w13,w14,w14,lsl#2 // r2*5
276 str w12,[$ctx,#16*2] // s1
277 str w14,[$ctx,#16*3] // r2
278 add w14,w15,w15,lsl#2 // r3*5
279 str w13,[$ctx,#16*4] // s2
280 str w15,[$ctx,#16*5] // r3
281 add w15,w16,w16,lsl#2 // r4*5
282 str w14,[$ctx,#16*6] // s3
283 str w16,[$ctx,#16*7] // r4
284 str w15,[$ctx,#16*8] // s4
287 .size poly1305_splat,.-poly1305_splat
289 .type poly1305_blocks_neon,%function
291 poly1305_blocks_neon:
292 ldr $is_base2_26,[$ctx,#24]
295 cbz $is_base2_26,poly1305_blocks
298 .inst 0xd503233f // paciasp
299 stp x29,x30,[sp,#-80]!
305 cbz $is_base2_26,.Lbase2_64_neon
307 ldp w10,w11,[$ctx] // load hash value base 2^26
308 ldp w12,w13,[$ctx,#8]
314 ldp $r0,$r1,[$ctx,#32] // load key value
316 add $h0,x10,x11,lsl#26 // base 2^26 -> base 2^64
318 adds $h0,$h0,x12,lsl#52
319 add $h1,$h1,x13,lsl#14
322 adds $h1,$h1,x14,lsl#40
323 adc $d2,$h2,xzr // can be partially reduced...
325 ldp $d0,$d1,[$inp],#16 // load input
327 add $s1,$r1,$r1,lsr#2 // s1 = r1 + (r1 >> 2)
329 and $t0,$d2,#-4 // ... so reduce
331 add $t0,$t0,$d2,lsr#2
340 adds $h0,$h0,$d0 // accumulate input
347 cbz $padbit,.Lstore_base2_64_neon
349 and x10,$h0,#0x03ffffff // base 2^64 -> base 2^26
352 and x12,x12,#0x03ffffff
356 cbnz $len,.Leven_neon
358 stp w10,w11,[$ctx] // store hash value base 2^26
359 stp w12,w13,[$ctx,#8]
364 .Lstore_base2_64_neon:
365 stp $h0,$h1,[$ctx] // store hash value base 2^64
366 stp $h2,xzr,[$ctx,#16] // note that is_base2_26 is zeroed
371 ldp $r0,$r1,[$ctx,#32] // load key value
373 ldp $h0,$h1,[$ctx] // load hash value base 2^64
379 ldp $d0,$d1,[$inp],#16 // load input
381 add $s1,$r1,$r1,lsr#2 // s1 = r1 + (r1 >> 2)
386 adds $h0,$h0,$d0 // accumulate input
393 and x10,$h0,#0x03ffffff // base 2^64 -> base 2^26
396 and x12,x12,#0x03ffffff
400 stp d8,d9,[sp,#16] // meet ABI requirements
411 ////////////////////////////////// initialize r^n table
413 add $s1,$r1,$r1,lsr#2 // s1 = r1 + (r1 >> 2)
419 bl poly1305_mult // r^2
423 bl poly1305_mult // r^3
427 bl poly1305_mult // r^4
435 csel $in2,$zeros,$in2,lo
438 str x4,[$ctx,#-24] // set is_base2_26
439 sub $ctx,$ctx,#48 // restore original $ctx
447 csel $in2,$zeros,$in2,lo
449 stp d8,d9,[sp,#16] // meet ABI requirements
461 ldp x8,x12,[$in2],#16 // inp[2:3] (or zero)
462 ldp x9,x13,[$in2],#48
464 lsl $padbit,$padbit,#24
473 and x4,x8,#0x03ffffff // base 2^64 -> base 2^26
474 and x5,x9,#0x03ffffff
477 add x4,x4,x5,lsl#32 // bfi x4,x5,#32,#32
480 add x6,x6,x7,lsl#32 // bfi x6,x7,#32,#32
482 and x8,x8,#0x03ffffff
483 and x9,x9,#0x03ffffff
486 add x12,$padbit,x12,lsr#40
487 add x13,$padbit,x13,lsr#40
488 add x8,x8,x9,lsl#32 // bfi x8,x9,#32,#32
490 add x10,x10,x11,lsl#32 // bfi x10,x11,#32,#32
491 add x12,x12,x13,lsl#32 // bfi x12,x13,#32,#32
496 ldp x8,x12,[$inp],#16 // inp[0:1]
497 ldp x9,x13,[$inp],#48
499 ld1 {$R0,$R1,$S1,$R2},[x15],#64
500 ld1 {$S2,$R3,$S3,$R4},[x15],#64
509 and x4,x8,#0x03ffffff // base 2^64 -> base 2^26
510 and x5,x9,#0x03ffffff
513 add x4,x4,x5,lsl#32 // bfi x4,x5,#32,#32
516 add x6,x6,x7,lsl#32 // bfi x6,x7,#32,#32
518 and x8,x8,#0x03ffffff
519 and x9,x9,#0x03ffffff
522 add x12,$padbit,x12,lsr#40
523 add x13,$padbit,x13,lsr#40
524 add x8,x8,x9,lsl#32 // bfi x8,x9,#32,#32
526 add x10,x10,x11,lsl#32 // bfi x10,x11,#32,#32
527 add x12,x12,x13,lsl#32 // bfi x12,x13,#32,#32
532 ushr $MASK.2d,$MASK.2d,#38
538 ////////////////////////////////////////////////////////////////
539 // ((inp[0]*r^4+inp[2]*r^2+inp[4])*r^4+inp[6]*r^2
540 // ((inp[1]*r^4+inp[3]*r^2+inp[5])*r^3+inp[7]*r
541 // \___________________/
542 // ((inp[0]*r^4+inp[2]*r^2+inp[4])*r^4+inp[6]*r^2+inp[8])*r^2
543 // ((inp[1]*r^4+inp[3]*r^2+inp[5])*r^4+inp[7]*r^2+inp[9])*r
544 // \___________________/ \____________________/
546 // Note that we start with inp[2:3]*r^2. This is because it
547 // doesn't depend on reduction in previous iteration.
548 ////////////////////////////////////////////////////////////////
549 // d4 = h0*r4 + h1*r3 + h2*r2 + h3*r1 + h4*r0
550 // d3 = h0*r3 + h1*r2 + h2*r1 + h3*r0 + h4*5*r4
551 // d2 = h0*r2 + h1*r1 + h2*r0 + h3*5*r4 + h4*5*r3
552 // d1 = h0*r1 + h1*r0 + h2*5*r4 + h3*5*r3 + h4*5*r2
553 // d0 = h0*r0 + h1*5*r4 + h2*5*r3 + h3*5*r2 + h4*5*r1
556 umull $ACC4,$IN23_0,${R4}[2]
557 csel $in2,$zeros,$in2,lo
558 umull $ACC3,$IN23_0,${R3}[2]
559 umull $ACC2,$IN23_0,${R2}[2]
560 ldp x8,x12,[$in2],#16 // inp[2:3] (or zero)
561 umull $ACC1,$IN23_0,${R1}[2]
562 ldp x9,x13,[$in2],#48
563 umull $ACC0,$IN23_0,${R0}[2]
571 umlal $ACC4,$IN23_1,${R3}[2]
572 and x4,x8,#0x03ffffff // base 2^64 -> base 2^26
573 umlal $ACC3,$IN23_1,${R2}[2]
574 and x5,x9,#0x03ffffff
575 umlal $ACC2,$IN23_1,${R1}[2]
577 umlal $ACC1,$IN23_1,${R0}[2]
579 umlal $ACC0,$IN23_1,${S4}[2]
580 add x4,x4,x5,lsl#32 // bfi x4,x5,#32,#32
582 umlal $ACC4,$IN23_2,${R2}[2]
584 umlal $ACC3,$IN23_2,${R1}[2]
586 umlal $ACC2,$IN23_2,${R0}[2]
587 add x6,x6,x7,lsl#32 // bfi x6,x7,#32,#32
588 umlal $ACC1,$IN23_2,${S4}[2]
590 umlal $ACC0,$IN23_2,${S3}[2]
591 and x8,x8,#0x03ffffff
593 umlal $ACC4,$IN23_3,${R1}[2]
594 and x9,x9,#0x03ffffff
595 umlal $ACC3,$IN23_3,${R0}[2]
597 umlal $ACC2,$IN23_3,${S4}[2]
599 umlal $ACC1,$IN23_3,${S3}[2]
600 add x8,x8,x9,lsl#32 // bfi x8,x9,#32,#32
601 umlal $ACC0,$IN23_3,${S2}[2]
604 add $IN01_2,$IN01_2,$H2
605 add x12,$padbit,x12,lsr#40
606 umlal $ACC4,$IN23_4,${R0}[2]
607 add x13,$padbit,x13,lsr#40
608 umlal $ACC3,$IN23_4,${S4}[2]
609 add x10,x10,x11,lsl#32 // bfi x10,x11,#32,#32
610 umlal $ACC2,$IN23_4,${S3}[2]
611 add x12,x12,x13,lsl#32 // bfi x12,x13,#32,#32
612 umlal $ACC1,$IN23_4,${S2}[2]
614 umlal $ACC0,$IN23_4,${S1}[2]
617 ////////////////////////////////////////////////////////////////
618 // (hash+inp[0:1])*r^4 and accumulate
620 add $IN01_0,$IN01_0,$H0
622 umlal $ACC3,$IN01_2,${R1}[0]
623 ldp x8,x12,[$inp],#16 // inp[0:1]
624 umlal $ACC0,$IN01_2,${S3}[0]
625 ldp x9,x13,[$inp],#48
626 umlal $ACC4,$IN01_2,${R2}[0]
627 umlal $ACC1,$IN01_2,${S4}[0]
628 umlal $ACC2,$IN01_2,${R0}[0]
636 add $IN01_1,$IN01_1,$H1
637 umlal $ACC3,$IN01_0,${R3}[0]
638 umlal $ACC4,$IN01_0,${R4}[0]
639 and x4,x8,#0x03ffffff // base 2^64 -> base 2^26
640 umlal $ACC2,$IN01_0,${R2}[0]
641 and x5,x9,#0x03ffffff
642 umlal $ACC0,$IN01_0,${R0}[0]
644 umlal $ACC1,$IN01_0,${R1}[0]
647 add $IN01_3,$IN01_3,$H3
648 add x4,x4,x5,lsl#32 // bfi x4,x5,#32,#32
649 umlal $ACC3,$IN01_1,${R2}[0]
651 umlal $ACC4,$IN01_1,${R3}[0]
653 umlal $ACC0,$IN01_1,${S4}[0]
654 add x6,x6,x7,lsl#32 // bfi x6,x7,#32,#32
655 umlal $ACC2,$IN01_1,${R1}[0]
657 umlal $ACC1,$IN01_1,${R0}[0]
658 and x8,x8,#0x03ffffff
660 add $IN01_4,$IN01_4,$H4
661 and x9,x9,#0x03ffffff
662 umlal $ACC3,$IN01_3,${R0}[0]
664 umlal $ACC0,$IN01_3,${S2}[0]
666 umlal $ACC4,$IN01_3,${R1}[0]
667 add x8,x8,x9,lsl#32 // bfi x8,x9,#32,#32
668 umlal $ACC1,$IN01_3,${S3}[0]
670 umlal $ACC2,$IN01_3,${S4}[0]
671 add x12,$padbit,x12,lsr#40
673 umlal $ACC3,$IN01_4,${S4}[0]
674 add x13,$padbit,x13,lsr#40
675 umlal $ACC0,$IN01_4,${S1}[0]
676 add x10,x10,x11,lsl#32 // bfi x10,x11,#32,#32
677 umlal $ACC4,$IN01_4,${R0}[0]
678 add x12,x12,x13,lsl#32 // bfi x12,x13,#32,#32
679 umlal $ACC1,$IN01_4,${S2}[0]
681 umlal $ACC2,$IN01_4,${S3}[0]
685 /////////////////////////////////////////////////////////////////
686 // lazy reduction as discussed in "NEON crypto" by D.J. Bernstein
689 // [see discussion in poly1305-armv4 module]
691 ushr $T0.2d,$ACC3,#26
693 ushr $T1.2d,$ACC0,#26
694 and $ACC0,$ACC0,$MASK.2d
695 add $ACC4,$ACC4,$T0.2d // h3 -> h4
696 bic $H3,#0xfc,lsl#24 // &=0x03ffffff
697 add $ACC1,$ACC1,$T1.2d // h0 -> h1
699 ushr $T0.2d,$ACC4,#26
701 ushr $T1.2d,$ACC1,#26
704 add $ACC2,$ACC2,$T1.2d // h1 -> h2
706 add $ACC0,$ACC0,$T0.2d
708 shrn $T1.2s,$ACC2,#26
710 add $ACC0,$ACC0,$T0.2d // h4 -> h0
712 add $H3,$H3,$T1.2s // h2 -> h3
715 shrn $T0.2s,$ACC0,#26
720 add $H1,$H1,$T0.2s // h0 -> h1
721 add $H4,$H4,$T1.2s // h3 -> h4
726 dup $IN23_2,${IN23_2}[0]
727 add $IN01_2,$IN01_2,$H2
729 ////////////////////////////////////////////////////////////////
730 // multiply (inp[0:1]+hash) or inp[2:3] by r^2:r^1
735 dup $IN23_2,${IN01_2}[0]
736 add $IN23_0,$IN01_0,$H0
737 add $IN23_3,$IN01_3,$H3
738 add $IN23_1,$IN01_1,$H1
739 add $IN23_4,$IN01_4,$H4
742 dup $IN23_0,${IN23_0}[0]
743 umull2 $ACC0,$IN23_2,${S3}
744 umull2 $ACC3,$IN23_2,${R1}
745 umull2 $ACC4,$IN23_2,${R2}
746 umull2 $ACC2,$IN23_2,${R0}
747 umull2 $ACC1,$IN23_2,${S4}
749 dup $IN23_1,${IN23_1}[0]
750 umlal2 $ACC0,$IN23_0,${R0}
751 umlal2 $ACC2,$IN23_0,${R2}
752 umlal2 $ACC3,$IN23_0,${R3}
753 umlal2 $ACC4,$IN23_0,${R4}
754 umlal2 $ACC1,$IN23_0,${R1}
756 dup $IN23_3,${IN23_3}[0]
757 umlal2 $ACC0,$IN23_1,${S4}
758 umlal2 $ACC3,$IN23_1,${R2}
759 umlal2 $ACC2,$IN23_1,${R1}
760 umlal2 $ACC4,$IN23_1,${R3}
761 umlal2 $ACC1,$IN23_1,${R0}
763 dup $IN23_4,${IN23_4}[0]
764 umlal2 $ACC3,$IN23_3,${R0}
765 umlal2 $ACC4,$IN23_3,${R1}
766 umlal2 $ACC0,$IN23_3,${S2}
767 umlal2 $ACC1,$IN23_3,${S3}
768 umlal2 $ACC2,$IN23_3,${S4}
770 umlal2 $ACC3,$IN23_4,${S4}
771 umlal2 $ACC0,$IN23_4,${S1}
772 umlal2 $ACC4,$IN23_4,${R0}
773 umlal2 $ACC1,$IN23_4,${S2}
774 umlal2 $ACC2,$IN23_4,${S3}
778 ////////////////////////////////////////////////////////////////
779 // (hash+inp[0:1])*r^4:r^3 and accumulate
781 add $IN01_0,$IN01_0,$H0
782 umlal $ACC3,$IN01_2,${R1}
783 umlal $ACC0,$IN01_2,${S3}
784 umlal $ACC4,$IN01_2,${R2}
785 umlal $ACC1,$IN01_2,${S4}
786 umlal $ACC2,$IN01_2,${R0}
788 add $IN01_1,$IN01_1,$H1
789 umlal $ACC3,$IN01_0,${R3}
790 umlal $ACC0,$IN01_0,${R0}
791 umlal $ACC4,$IN01_0,${R4}
792 umlal $ACC1,$IN01_0,${R1}
793 umlal $ACC2,$IN01_0,${R2}
795 add $IN01_3,$IN01_3,$H3
796 umlal $ACC3,$IN01_1,${R2}
797 umlal $ACC0,$IN01_1,${S4}
798 umlal $ACC4,$IN01_1,${R3}
799 umlal $ACC1,$IN01_1,${R0}
800 umlal $ACC2,$IN01_1,${R1}
802 add $IN01_4,$IN01_4,$H4
803 umlal $ACC3,$IN01_3,${R0}
804 umlal $ACC0,$IN01_3,${S2}
805 umlal $ACC4,$IN01_3,${R1}
806 umlal $ACC1,$IN01_3,${S3}
807 umlal $ACC2,$IN01_3,${S4}
809 umlal $ACC3,$IN01_4,${S4}
810 umlal $ACC0,$IN01_4,${S1}
811 umlal $ACC4,$IN01_4,${R0}
812 umlal $ACC1,$IN01_4,${S2}
813 umlal $ACC2,$IN01_4,${S3}
816 ////////////////////////////////////////////////////////////////
819 addp $ACC3,$ACC3,$ACC3
820 ldp d8,d9,[sp,#16] // meet ABI requirements
821 addp $ACC0,$ACC0,$ACC0
823 addp $ACC4,$ACC4,$ACC4
825 addp $ACC1,$ACC1,$ACC1
827 addp $ACC2,$ACC2,$ACC2
829 ////////////////////////////////////////////////////////////////
830 // lazy reduction, but without narrowing
832 ushr $T0.2d,$ACC3,#26
833 and $ACC3,$ACC3,$MASK.2d
834 ushr $T1.2d,$ACC0,#26
835 and $ACC0,$ACC0,$MASK.2d
837 add $ACC4,$ACC4,$T0.2d // h3 -> h4
838 add $ACC1,$ACC1,$T1.2d // h0 -> h1
840 ushr $T0.2d,$ACC4,#26
841 and $ACC4,$ACC4,$MASK.2d
842 ushr $T1.2d,$ACC1,#26
843 and $ACC1,$ACC1,$MASK.2d
844 add $ACC2,$ACC2,$T1.2d // h1 -> h2
846 add $ACC0,$ACC0,$T0.2d
848 ushr $T1.2d,$ACC2,#26
849 and $ACC2,$ACC2,$MASK.2d
850 add $ACC0,$ACC0,$T0.2d // h4 -> h0
851 add $ACC3,$ACC3,$T1.2d // h2 -> h3
853 ushr $T0.2d,$ACC0,#26
854 and $ACC0,$ACC0,$MASK.2d
855 ushr $T1.2d,$ACC3,#26
856 and $ACC3,$ACC3,$MASK.2d
857 add $ACC1,$ACC1,$T0.2d // h0 -> h1
858 add $ACC4,$ACC4,$T1.2d // h3 -> h4
860 ////////////////////////////////////////////////////////////////
861 // write the result, can be partially reduced
863 st4 {$ACC0,$ACC1,$ACC2,$ACC3}[0],[$ctx],#16
864 st1 {$ACC4}[0],[$ctx]
868 .inst 0xd50323bf // autiasp
870 .size poly1305_blocks_neon,.-poly1305_blocks_neon
872 .type poly1305_emit_neon,%function
875 ldr $is_base2_26,[$ctx,#24]
876 cbz $is_base2_26,poly1305_emit
878 ldp w10,w11,[$ctx] // load hash value base 2^26
879 ldp w12,w13,[$ctx,#8]
882 add $h0,x10,x11,lsl#26 // base 2^26 -> base 2^64
884 adds $h0,$h0,x12,lsl#52
885 add $h1,$h1,x13,lsl#14
888 adds $h1,$h1,x14,lsl#40
889 adc $h2,$h2,xzr // can be partially reduced...
891 ldp $t0,$t1,[$nonce] // load nonce
893 and $d0,$h2,#-4 // ... so reduce
894 add $d0,$d0,$h2,lsr#2
900 adds $d0,$h0,#5 // compare to modulus
904 tst $d2,#-4 // see if it's carried/borrowed
910 ror $t0,$t0,#32 // flip nonce words
913 adds $h0,$h0,$t0 // accumulate nonce
916 rev $h0,$h0 // flip output bytes
919 stp $h0,$h1,[$mac] // write result
922 .size poly1305_emit_neon,.-poly1305_emit_neon
926 .long 0,0,0,0,0,0,0,0
929 .long OPENSSL_armcap_P-.
931 .quad OPENSSL_armcap_P-.
933 .asciz "Poly1305 for ARMv8, CRYPTOGAMS by <appro\@openssl.org>"
937 foreach (split("\n",$code)) {
938 s/\b(shrn\s+v[0-9]+)\.[24]d/$1.2s/ or
939 s/\b(fmov\s+)v([0-9]+)[^,]*,\s*x([0-9]+)/$1d$2,x$3/ or
940 (m/\bdup\b/ and (s/\.[24]s/.2d/g or 1)) or
941 (m/\b(eor|and)/ and (s/\.[248][sdh]/.16b/g or 1)) or
942 (m/\bum(ul|la)l\b/ and (s/\.4s/.2s/g or 1)) or
943 (m/\bum(ul|la)l2\b/ and (s/\.2s/.4s/g or 1)) or
944 (m/\bst[1-4]\s+{[^}]+}\[/ and (s/\.[24]d/.s/g or 1));
946 s/\.[124]([sd])\[/.$1\[/;
950 close STDOUT or die "error closing STDOUT: $!";