3 # ====================================================================
4 # Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
5 # project. The module is, however, dual licensed under OpenSSL and
6 # CRYPTOGAMS licenses depending on where you obtain it. For further
7 # details see http://www.openssl.org/~appro/cryptogams/.
8 # ====================================================================
10 # This module implements Poly1305 hash for ARMv8.
14 # Numbers are cycles per processed byte with poly1305_blocks alone.
18 # Apple A7 1.86/+5% 0.72
19 # Cortex-A53 2.69/+58% 1.47
20 # Cortex-A57 2.70/+7% 1.14
21 # Denver 1.64/+50% 1.18(*)
22 # X-Gene 2.13/+68% 2.19
24 # (*) estimate based on resources availability is less than 1.0,
25 # i.e. measured result is worse than expected, presumably binary
26 # translator is not almighty;
31 $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
32 ( $xlate="${dir}arm-xlate.pl" and -f $xlate ) or
33 ( $xlate="${dir}../../perlasm/arm-xlate.pl" and -f $xlate) or
34 die "can't locate arm-xlate.pl";
36 open OUT,"| \"$^X\" $xlate $flavour $output";
39 my ($ctx,$inp,$len,$padbit) = map("x$_",(0..3));
40 my ($mac,$nonce)=($inp,$len);
42 my ($h0,$h1,$h2,$r0,$r1,$s1,$t0,$t1,$d0,$d1,$d2) = map("x$_",(4..14));
49 // forward "declarations" are required for Apple
50 .extern OPENSSL_armcap_P
51 .globl poly1305_blocks
55 .type poly1305_init,%function
59 stp xzr,xzr,[$ctx] // zero hash value
60 stp xzr,xzr,[$ctx,#16] // [along with is_base2_26]
66 ldrsw $t1,.LOPENSSL_armcap_P
68 ldr $t1,.LOPENSSL_armcap_P
70 adr $t0,.LOPENSSL_armcap_P
72 ldp $r0,$r1,[$inp] // load key
73 mov $s1,#0xfffffffc0fffffff
74 movk $s1,#0x0fff,lsl#48
77 rev $r0,$r0 // flip bytes
80 and $r0,$r0,$s1 // &=0ffffffc0fffffff
82 and $r1,$r1,$s1 // &=0ffffffc0ffffffc
83 stp $r0,$r1,[$ctx,#32] // save key value
87 adr $d0,poly1305_blocks
88 adr $r0,poly1305_blocks_neon
90 adr $r1,poly1305_emit_neon
100 .size poly1305_init,.-poly1305_init
102 .type poly1305_blocks,%function
108 ldp $h0,$h1,[$ctx] // load hash value
109 ldp $r0,$r1,[$ctx,#32] // load key value
111 add $s1,$r1,$r1,lsr#2 // s1 = r1 + (r1 >> 2)
116 ldp $t0,$t1,[$inp],#16 // load input
122 adds $h0,$h0,$t0 // accumulate input
125 mul $d0,$h0,$r0 // h0*r0
129 mul $t0,$h1,$s1 // h1*5*r1
133 mul $t0,$h0,$r1 // h0*r1
138 mul $t0,$h1,$r0 // h1*r0
143 mul $t0,$h2,$s1 // h2*5*r1
145 mul $t1,$h2,$r0 // h2*r0
150 and $t0,$d2,#-4 // final reduction
152 add $t0,$t0,$d2,lsr#2
159 stp $h0,$h1,[$ctx] // store hash value
164 .size poly1305_blocks,.-poly1305_blocks
166 .type poly1305_emit,%function
169 ldp $h0,$h1,[$ctx] // load hash base 2^64
171 ldp $t0,$t1,[$nonce] // load nonce
173 adds $d0,$h0,#5 // compare to modulus
177 tst $d2,#-4 // see if it's carried/borrowed
183 ror $t0,$t0,#32 // flip nonce words
186 adds $h0,$h0,$t0 // accumulate nonce
189 rev $h0,$h0 // flip output bytes
192 stp $h0,$h1,[$mac] // write result
195 .size poly1305_emit,.-poly1305_emit
197 my ($R0,$R1,$S1,$R2,$S2,$R3,$S3,$R4,$S4) = map("v$_.4s",(0..8));
198 my ($IN01_0,$IN01_1,$IN01_2,$IN01_3,$IN01_4) = map("v$_.2s",(9..13));
199 my ($IN23_0,$IN23_1,$IN23_2,$IN23_3,$IN23_4) = map("v$_.2s",(14..18));
200 my ($ACC0,$ACC1,$ACC2,$ACC3,$ACC4) = map("v$_.2d",(19..23));
201 my ($H0,$H1,$H2,$H3,$H4) = map("v$_.2s",(24..28));
202 my ($T0,$T1,$MASK) = map("v$_",(29..31));
204 my ($in2,$zeros)=("x16","x17");
205 my $is_base2_26 = $zeros; # borrow
208 .type poly1305_mult,%function
211 mul $d0,$h0,$r0 // h0*r0
214 mul $t0,$h1,$s1 // h1*5*r1
218 mul $t0,$h0,$r1 // h0*r1
223 mul $t0,$h1,$r0 // h1*r0
228 mul $t0,$h2,$s1 // h2*5*r1
230 mul $t1,$h2,$r0 // h2*r0
235 and $t0,$d2,#-4 // final reduction
237 add $t0,$t0,$d2,lsr#2
243 .size poly1305_mult,.-poly1305_mult
245 .type poly1305_splat,%function
248 and x12,$h0,#0x03ffffff // base 2^64 -> base 2^26
251 and x14,x14,#0x03ffffff
255 str w12,[$ctx,#16*0] // r0
256 add w12,w13,w13,lsl#2 // r1*5
257 str w13,[$ctx,#16*1] // r1
258 add w13,w14,w14,lsl#2 // r2*5
259 str w12,[$ctx,#16*2] // s1
260 str w14,[$ctx,#16*3] // r2
261 add w14,w15,w15,lsl#2 // r3*5
262 str w13,[$ctx,#16*4] // s2
263 str w15,[$ctx,#16*5] // r3
264 add w15,w16,w16,lsl#2 // r4*5
265 str w14,[$ctx,#16*6] // s3
266 str w16,[$ctx,#16*7] // r4
267 str w15,[$ctx,#16*8] // s4
270 .size poly1305_splat,.-poly1305_splat
272 .type poly1305_blocks_neon,%function
274 poly1305_blocks_neon:
275 ldr $is_base2_26,[$ctx,#24]
278 cbz $is_base2_26,poly1305_blocks
281 stp x29,x30,[sp,#-80]!
287 cbz $is_base2_26,.Lbase2_64_neon
289 ldp w10,w11,[$ctx] // load hash value base 2^26
290 ldp w12,w13,[$ctx,#8]
296 ldp $r0,$r1,[$ctx,#32] // load key value
298 add $h0,x10,x11,lsl#26 // base 2^26 -> base 2^64
300 adds $h0,$h0,x12,lsl#52
301 add $h1,$h1,x13,lsl#14
304 adds $h1,$h1,x14,lsl#40
305 adc $d2,$h2,xzr // can be partially reduced...
307 ldp $d0,$d1,[$inp],#16 // load input
309 add $s1,$r1,$r1,lsr#2 // s1 = r1 + (r1 >> 2)
311 and $t0,$d2,#-4 // ... so reduce
313 add $t0,$t0,$d2,lsr#2
322 adds $h0,$h0,$d0 // accumulate input
329 cbz $padbit,.Lstore_base2_64_neon
331 and x10,$h0,#0x03ffffff // base 2^64 -> base 2^26
334 and x12,x12,#0x03ffffff
338 cbnz $len,.Leven_neon
340 stp w10,w11,[$ctx] // store hash value base 2^26
341 stp w12,w13,[$ctx,#8]
346 .Lstore_base2_64_neon:
347 stp $h0,$h1,[$ctx] // store hash value base 2^64
348 stp $h2,xzr,[$ctx,#16] // note that is_base2_26 is zeroed
353 ldp $r0,$r1,[$ctx,#32] // load key value
355 ldp $h0,$h1,[$ctx] // load hash value base 2^64
361 ldp $d0,$d1,[$inp],#16 // load input
363 add $s1,$r1,$r1,lsr#2 // s1 = r1 + (r1 >> 2)
368 adds $h0,$h0,$d0 // accumulate input
375 and x10,$h0,#0x03ffffff // base 2^64 -> base 2^26
378 and x12,x12,#0x03ffffff
382 stp d8,d9,[sp,#16] // meet ABI requirements
393 ////////////////////////////////// initialize r^n table
395 add $s1,$r1,$r1,lsr#2 // s1 = r1 + (r1 >> 2)
401 bl poly1305_mult // r^2
405 bl poly1305_mult // r^3
409 bl poly1305_mult // r^4
417 csel $in2,$zeros,$in2,lo
420 str x4,[$ctx,#-24] // set is_base2_26
421 sub $ctx,$ctx,#48 // restore original $ctx
429 csel $in2,$zeros,$in2,lo
431 stp d8,d9,[sp,#16] // meet ABI requirements
443 ldp x8,x12,[$in2],#16 // inp[2:3] (or zero)
444 ldp x9,x13,[$in2],#48
446 lsl $padbit,$padbit,#24
455 and x4,x8,#0x03ffffff // base 2^64 -> base 2^26
456 and x5,x9,#0x03ffffff
459 add x4,x4,x5,lsl#32 // bfi x4,x5,#32,#32
462 add x6,x6,x7,lsl#32 // bfi x6,x7,#32,#32
464 and x8,x8,#0x03ffffff
465 and x9,x9,#0x03ffffff
468 add x12,$padbit,x12,lsr#40
469 add x13,$padbit,x13,lsr#40
470 add x8,x8,x9,lsl#32 // bfi x8,x9,#32,#32
472 add x10,x10,x11,lsl#32 // bfi x10,x11,#32,#32
473 add x12,x12,x13,lsl#32 // bfi x12,x13,#32,#32
478 ldp x8,x12,[$inp],#16 // inp[0:1]
479 ldp x9,x13,[$inp],#48
481 ld1 {$R0,$R1,$S1,$R2},[x15],#64
482 ld1 {$S2,$R3,$S3,$R4},[x15],#64
491 and x4,x8,#0x03ffffff // base 2^64 -> base 2^26
492 and x5,x9,#0x03ffffff
495 add x4,x4,x5,lsl#32 // bfi x4,x5,#32,#32
498 add x6,x6,x7,lsl#32 // bfi x6,x7,#32,#32
500 and x8,x8,#0x03ffffff
501 and x9,x9,#0x03ffffff
504 add x12,$padbit,x12,lsr#40
505 add x13,$padbit,x13,lsr#40
506 add x8,x8,x9,lsl#32 // bfi x8,x9,#32,#32
508 add x10,x10,x11,lsl#32 // bfi x10,x11,#32,#32
509 add x12,x12,x13,lsl#32 // bfi x12,x13,#32,#32
518 ////////////////////////////////////////////////////////////////
519 // ((inp[0]*r^4+inp[2]*r^2+inp[4])*r^4+inp[6]*r^2
520 // ((inp[1]*r^4+inp[3]*r^2+inp[5])*r^3+inp[7]*r
521 // \___________________/
522 // ((inp[0]*r^4+inp[2]*r^2+inp[4])*r^4+inp[6]*r^2+inp[8])*r^2
523 // ((inp[1]*r^4+inp[3]*r^2+inp[5])*r^4+inp[7]*r^2+inp[9])*r
524 // \___________________/ \____________________/
526 // Note that we start with inp[2:3]*r^2. This is because it
527 // doesn't depend on reduction in previous iteration.
528 ////////////////////////////////////////////////////////////////
529 // d4 = h0*r4 + h1*r3 + h2*r2 + h3*r1 + h4*r0
530 // d3 = h0*r3 + h1*r2 + h2*r1 + h3*r0 + h4*5*r4
531 // d2 = h0*r2 + h1*r1 + h2*r0 + h3*5*r4 + h4*5*r3
532 // d1 = h0*r1 + h1*r0 + h2*5*r4 + h3*5*r3 + h4*5*r2
533 // d0 = h0*r0 + h1*5*r4 + h2*5*r3 + h3*5*r2 + h4*5*r1
536 umull $ACC4,$IN23_0,${R4}[2]
537 csel $in2,$zeros,$in2,lo
538 umull $ACC3,$IN23_0,${R3}[2]
539 umull $ACC2,$IN23_0,${R2}[2]
540 ldp x8,x12,[$in2],#16 // inp[2:3] (or zero)
541 umull $ACC1,$IN23_0,${R1}[2]
542 ldp x9,x13,[$in2],#48
543 umull $ACC0,$IN23_0,${R0}[2]
551 umlal $ACC4,$IN23_1,${R3}[2]
552 and x4,x8,#0x03ffffff // base 2^64 -> base 2^26
553 umlal $ACC3,$IN23_1,${R2}[2]
554 and x5,x9,#0x03ffffff
555 umlal $ACC2,$IN23_1,${R1}[2]
557 umlal $ACC1,$IN23_1,${R0}[2]
559 umlal $ACC0,$IN23_1,${S4}[2]
560 add x4,x4,x5,lsl#32 // bfi x4,x5,#32,#32
562 umlal $ACC4,$IN23_2,${R2}[2]
564 umlal $ACC3,$IN23_2,${R1}[2]
566 umlal $ACC2,$IN23_2,${R0}[2]
567 add x6,x6,x7,lsl#32 // bfi x6,x7,#32,#32
568 umlal $ACC1,$IN23_2,${S4}[2]
570 umlal $ACC0,$IN23_2,${S3}[2]
571 and x8,x8,#0x03ffffff
573 umlal $ACC4,$IN23_3,${R1}[2]
574 and x9,x9,#0x03ffffff
575 umlal $ACC3,$IN23_3,${R0}[2]
577 umlal $ACC2,$IN23_3,${S4}[2]
579 umlal $ACC1,$IN23_3,${S3}[2]
580 add x8,x8,x9,lsl#32 // bfi x8,x9,#32,#32
581 umlal $ACC0,$IN23_3,${S2}[2]
584 add $IN01_2,$IN01_2,$H2
585 add x12,$padbit,x12,lsr#40
586 umlal $ACC4,$IN23_4,${R0}[2]
587 add x13,$padbit,x13,lsr#40
588 umlal $ACC3,$IN23_4,${S4}[2]
589 add x10,x10,x11,lsl#32 // bfi x10,x11,#32,#32
590 umlal $ACC2,$IN23_4,${S3}[2]
591 add x12,x12,x13,lsl#32 // bfi x12,x13,#32,#32
592 umlal $ACC1,$IN23_4,${S2}[2]
594 umlal $ACC0,$IN23_4,${S1}[2]
597 ////////////////////////////////////////////////////////////////
598 // (hash+inp[0:1])*r^4 and accumulate
600 add $IN01_0,$IN01_0,$H0
602 umlal $ACC3,$IN01_2,${R1}[0]
603 ldp x8,x12,[$inp],#16 // inp[0:1]
604 umlal $ACC0,$IN01_2,${S3}[0]
605 ldp x9,x13,[$inp],#48
606 umlal $ACC4,$IN01_2,${R2}[0]
607 umlal $ACC1,$IN01_2,${S4}[0]
608 umlal $ACC2,$IN01_2,${R0}[0]
616 add $IN01_1,$IN01_1,$H1
617 umlal $ACC3,$IN01_0,${R3}[0]
618 umlal $ACC4,$IN01_0,${R4}[0]
619 and x4,x8,#0x03ffffff // base 2^64 -> base 2^26
620 umlal $ACC2,$IN01_0,${R2}[0]
621 and x5,x9,#0x03ffffff
622 umlal $ACC0,$IN01_0,${R0}[0]
624 umlal $ACC1,$IN01_0,${R1}[0]
627 add $IN01_3,$IN01_3,$H3
628 add x4,x4,x5,lsl#32 // bfi x4,x5,#32,#32
629 umlal $ACC3,$IN01_1,${R2}[0]
631 umlal $ACC4,$IN01_1,${R3}[0]
633 umlal $ACC0,$IN01_1,${S4}[0]
634 add x6,x6,x7,lsl#32 // bfi x6,x7,#32,#32
635 umlal $ACC2,$IN01_1,${R1}[0]
637 umlal $ACC1,$IN01_1,${R0}[0]
638 and x8,x8,#0x03ffffff
640 add $IN01_4,$IN01_4,$H4
641 and x9,x9,#0x03ffffff
642 umlal $ACC3,$IN01_3,${R0}[0]
644 umlal $ACC0,$IN01_3,${S2}[0]
646 umlal $ACC4,$IN01_3,${R1}[0]
647 add x8,x8,x9,lsl#32 // bfi x8,x9,#32,#32
648 umlal $ACC1,$IN01_3,${S3}[0]
650 umlal $ACC2,$IN01_3,${S4}[0]
651 add x12,$padbit,x12,lsr#40
653 umlal $ACC3,$IN01_4,${S4}[0]
654 add x13,$padbit,x13,lsr#40
655 umlal $ACC0,$IN01_4,${S1}[0]
656 add x10,x10,x11,lsl#32 // bfi x10,x11,#32,#32
657 umlal $ACC4,$IN01_4,${R0}[0]
658 add x12,x12,x13,lsl#32 // bfi x12,x13,#32,#32
659 umlal $ACC1,$IN01_4,${S2}[0]
661 umlal $ACC2,$IN01_4,${S3}[0]
664 /////////////////////////////////////////////////////////////////
665 // lazy reduction as discussed in "NEON crypto" by D.J. Bernstein
668 ushr $T0.2d,$ACC3,#26
671 ushr $T1.2d,$ACC0,#26
673 add $ACC4,$ACC4,$T0.2d // h3 -> h4
674 bic $H3,#0xfc,lsl#24 // &=0x03ffffff
675 add $ACC1,$ACC1,$T1.2d // h0 -> h1
678 shrn $T0.2s,$ACC4,#26
680 ushr $T1.2d,$ACC1,#26
682 add $ACC2,$ACC2,$T1.2d // h1 -> h2
688 shrn $T1.2s,$ACC2,#26
690 add $H0,$H0,$T0.2s // h4 -> h0
691 add $H3,$H3,$T1.2s // h2 -> h3
698 add $H1,$H1,$T0.2s // h0 -> h1
699 add $H4,$H4,$T1.2s // h3 -> h4
704 dup $IN23_2,${IN23_2}[0]
706 add $IN01_2,$IN01_2,$H2
707 ushr $MASK.2d,$MASK.2d,#38
709 ////////////////////////////////////////////////////////////////
710 // multiply (inp[0:1]+hash) or inp[2:3] by r^2:r^1
715 dup $IN23_2,${IN01_2}[0]
716 add $IN23_0,$IN01_0,$H0
717 add $IN23_3,$IN01_3,$H3
718 add $IN23_1,$IN01_1,$H1
719 add $IN23_4,$IN01_4,$H4
722 dup $IN23_0,${IN23_0}[0]
723 umull2 $ACC0,$IN23_2,${S3}
724 umull2 $ACC3,$IN23_2,${R1}
725 umull2 $ACC4,$IN23_2,${R2}
726 umull2 $ACC2,$IN23_2,${R0}
727 umull2 $ACC1,$IN23_2,${S4}
729 dup $IN23_1,${IN23_1}[0]
730 umlal2 $ACC0,$IN23_0,${R0}
731 umlal2 $ACC2,$IN23_0,${R2}
732 umlal2 $ACC3,$IN23_0,${R3}
733 umlal2 $ACC4,$IN23_0,${R4}
734 umlal2 $ACC1,$IN23_0,${R1}
736 dup $IN23_3,${IN23_3}[0]
737 umlal2 $ACC0,$IN23_1,${S4}
738 umlal2 $ACC3,$IN23_1,${R2}
739 umlal2 $ACC2,$IN23_1,${R1}
740 umlal2 $ACC4,$IN23_1,${R3}
741 umlal2 $ACC1,$IN23_1,${R0}
743 dup $IN23_4,${IN23_4}[0]
744 umlal2 $ACC3,$IN23_3,${R0}
745 umlal2 $ACC4,$IN23_3,${R1}
746 umlal2 $ACC0,$IN23_3,${S2}
747 umlal2 $ACC1,$IN23_3,${S3}
748 umlal2 $ACC2,$IN23_3,${S4}
750 umlal2 $ACC3,$IN23_4,${S4}
751 umlal2 $ACC0,$IN23_4,${S1}
752 umlal2 $ACC4,$IN23_4,${R0}
753 umlal2 $ACC1,$IN23_4,${S2}
754 umlal2 $ACC2,$IN23_4,${S3}
758 ////////////////////////////////////////////////////////////////
759 // (hash+inp[0:1])*r^4:r^3 and accumulate
761 add $IN01_0,$IN01_0,$H0
762 umlal $ACC3,$IN01_2,${R1}
763 umlal $ACC0,$IN01_2,${S3}
764 umlal $ACC4,$IN01_2,${R2}
765 umlal $ACC1,$IN01_2,${S4}
766 umlal $ACC2,$IN01_2,${R0}
768 add $IN01_1,$IN01_1,$H1
769 umlal $ACC3,$IN01_0,${R3}
770 umlal $ACC0,$IN01_0,${R0}
771 umlal $ACC4,$IN01_0,${R4}
772 umlal $ACC1,$IN01_0,${R1}
773 umlal $ACC2,$IN01_0,${R2}
775 add $IN01_3,$IN01_3,$H3
776 umlal $ACC3,$IN01_1,${R2}
777 umlal $ACC0,$IN01_1,${S4}
778 umlal $ACC4,$IN01_1,${R3}
779 umlal $ACC1,$IN01_1,${R0}
780 umlal $ACC2,$IN01_1,${R1}
782 add $IN01_4,$IN01_4,$H4
783 umlal $ACC3,$IN01_3,${R0}
784 umlal $ACC0,$IN01_3,${S2}
785 umlal $ACC4,$IN01_3,${R1}
786 umlal $ACC1,$IN01_3,${S3}
787 umlal $ACC2,$IN01_3,${S4}
789 umlal $ACC3,$IN01_4,${S4}
790 umlal $ACC0,$IN01_4,${S1}
791 umlal $ACC4,$IN01_4,${R0}
792 umlal $ACC1,$IN01_4,${S2}
793 umlal $ACC2,$IN01_4,${S3}
796 ////////////////////////////////////////////////////////////////
799 addp $ACC3,$ACC3,$ACC3
800 ldp d8,d9,[sp,#16] // meet ABI requirements
801 addp $ACC0,$ACC0,$ACC0
803 addp $ACC4,$ACC4,$ACC4
805 addp $ACC1,$ACC1,$ACC1
807 addp $ACC2,$ACC2,$ACC2
809 ////////////////////////////////////////////////////////////////
810 // lazy reduction, but without narrowing
812 ushr $T0.2d,$ACC3,#26
813 and $ACC3,$ACC3,$MASK.2d
814 ushr $T1.2d,$ACC0,#26
815 and $ACC0,$ACC0,$MASK.2d
817 add $ACC4,$ACC4,$T0.2d // h3 -> h4
818 add $ACC1,$ACC1,$T1.2d // h0 -> h1
820 ushr $T0.2d,$ACC4,#26
821 and $ACC4,$ACC4,$MASK.2d
822 ushr $T1.2d,$ACC1,#26
823 and $ACC1,$ACC1,$MASK.2d
824 add $ACC2,$ACC2,$T1.2d // h1 -> h2
826 add $ACC0,$ACC0,$T0.2d
828 ushr $T1.2d,$ACC2,#26
829 and $ACC2,$ACC2,$MASK.2d
830 add $ACC0,$ACC0,$T0.2d // h4 -> h0
831 add $ACC3,$ACC3,$T1.2d // h2 -> h3
833 ushr $T0.2d,$ACC0,#26
834 and $ACC0,$ACC0,$MASK.2d
835 ushr $T1.2d,$ACC3,#26
836 and $ACC3,$ACC3,$MASK.2d
837 add $ACC1,$ACC1,$T0.2d // h0 -> h1
838 add $ACC4,$ACC4,$T1.2d // h3 -> h4
840 ////////////////////////////////////////////////////////////////
841 // write the result, can be partially reduced
843 st4 {$ACC0,$ACC1,$ACC2,$ACC3}[0],[$ctx],#16
844 st1 {$ACC4}[0],[$ctx]
849 .size poly1305_blocks_neon,.-poly1305_blocks_neon
851 .type poly1305_emit_neon,%function
854 ldr $is_base2_26,[$ctx,#24]
855 cbz $is_base2_26,poly1305_emit
857 ldp w10,w11,[$ctx] // load hash value base 2^26
858 ldp w12,w13,[$ctx,#8]
861 add $h0,x10,x11,lsl#26 // base 2^26 -> base 2^64
863 adds $h0,$h0,x12,lsl#52
864 add $h1,$h1,x13,lsl#14
867 adds $h1,$h1,x14,lsl#40
868 adc $h2,$h2,xzr // can be partially reduced...
870 ldp $t0,$t1,[$nonce] // load nonce
872 and $d0,$h2,#-4 // ... so reduce
873 add $d0,$d0,$h2,lsr#2
879 adds $d0,$h0,#5 // compare to modulus
883 tst $d2,#-4 // see if it's carried/borrowed
889 ror $t0,$t0,#32 // flip nonce words
892 adds $h0,$h0,$t0 // accumulate nonce
895 rev $h0,$h0 // flip output bytes
898 stp $h0,$h1,[$mac] // write result
901 .size poly1305_emit_neon,.-poly1305_emit_neon
905 .long 0,0,0,0,0,0,0,0
908 .long OPENSSL_armcap_P-.
910 .quad OPENSSL_armcap_P-.
912 .asciz "Poly1305 for ARMv8, CRYPTOGAMS by <appro\@openssl.org>"
916 foreach (split("\n",$code)) {
917 s/\b(shrn\s+v[0-9]+)\.[24]d/$1.2s/ or
918 s/\b(fmov\s+)v([0-9]+)[^,]*,\s*x([0-9]+)/$1d$2,x$3/ or
919 (m/\bdup\b/ and (s/\.[24]s/.2d/g or 1)) or
920 (m/\b(eor|and)/ and (s/\.[248][sdh]/.16b/g or 1)) or
921 (m/\bum(ul|la)l\b/ and (s/\.4s/.2s/g or 1)) or
922 (m/\bum(ul|la)l2\b/ and (s/\.2s/.4s/g or 1)) or
923 (m/\bst[1-4]\s+{[^}]+}\[/ and (s/\.[24]d/.s/g or 1));
925 s/\.[124]([sd])\[/.$1\[/;