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
31 # (*) estimate based on resources availability is less than 1.0,
32 # i.e. measured result is worse than expected, presumably binary
33 # translator is not almighty;
38 $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
39 ( $xlate="${dir}arm-xlate.pl" and -f $xlate ) or
40 ( $xlate="${dir}../../perlasm/arm-xlate.pl" and -f $xlate) or
41 die "can't locate arm-xlate.pl";
43 open OUT,"| \"$^X\" $xlate $flavour $output";
46 my ($ctx,$inp,$len,$padbit) = map("x$_",(0..3));
47 my ($mac,$nonce)=($inp,$len);
49 my ($h0,$h1,$h2,$r0,$r1,$s1,$t0,$t1,$d0,$d1,$d2) = map("x$_",(4..14));
56 // forward "declarations" are required for Apple
57 .extern OPENSSL_armcap_P
58 .globl poly1305_blocks
62 .type poly1305_init,%function
66 stp xzr,xzr,[$ctx] // zero hash value
67 stp xzr,xzr,[$ctx,#16] // [along with is_base2_26]
73 ldrsw $t1,.LOPENSSL_armcap_P
75 ldr $t1,.LOPENSSL_armcap_P
77 adr $t0,.LOPENSSL_armcap_P
79 ldp $r0,$r1,[$inp] // load key
80 mov $s1,#0xfffffffc0fffffff
81 movk $s1,#0x0fff,lsl#48
84 rev $r0,$r0 // flip bytes
87 and $r0,$r0,$s1 // &=0ffffffc0fffffff
89 and $r1,$r1,$s1 // &=0ffffffc0ffffffc
90 stp $r0,$r1,[$ctx,#32] // save key value
94 adr $d0,poly1305_blocks
95 adr $r0,poly1305_blocks_neon
97 adr $r1,poly1305_emit_neon
107 .size poly1305_init,.-poly1305_init
109 .type poly1305_blocks,%function
115 ldp $h0,$h1,[$ctx] // load hash value
116 ldp $r0,$r1,[$ctx,#32] // load key value
118 add $s1,$r1,$r1,lsr#2 // s1 = r1 + (r1 >> 2)
123 ldp $t0,$t1,[$inp],#16 // load input
129 adds $h0,$h0,$t0 // accumulate input
132 mul $d0,$h0,$r0 // h0*r0
136 mul $t0,$h1,$s1 // h1*5*r1
140 mul $t0,$h0,$r1 // h0*r1
145 mul $t0,$h1,$r0 // h1*r0
150 mul $t0,$h2,$s1 // h2*5*r1
152 mul $t1,$h2,$r0 // h2*r0
157 and $t0,$d2,#-4 // final reduction
159 add $t0,$t0,$d2,lsr#2
166 stp $h0,$h1,[$ctx] // store hash value
171 .size poly1305_blocks,.-poly1305_blocks
173 .type poly1305_emit,%function
176 ldp $h0,$h1,[$ctx] // load hash base 2^64
178 ldp $t0,$t1,[$nonce] // load nonce
180 adds $d0,$h0,#5 // compare to modulus
184 tst $d2,#-4 // see if it's carried/borrowed
190 ror $t0,$t0,#32 // flip nonce words
193 adds $h0,$h0,$t0 // accumulate nonce
196 rev $h0,$h0 // flip output bytes
199 stp $h0,$h1,[$mac] // write result
202 .size poly1305_emit,.-poly1305_emit
204 my ($R0,$R1,$S1,$R2,$S2,$R3,$S3,$R4,$S4) = map("v$_.4s",(0..8));
205 my ($IN01_0,$IN01_1,$IN01_2,$IN01_3,$IN01_4) = map("v$_.2s",(9..13));
206 my ($IN23_0,$IN23_1,$IN23_2,$IN23_3,$IN23_4) = map("v$_.2s",(14..18));
207 my ($ACC0,$ACC1,$ACC2,$ACC3,$ACC4) = map("v$_.2d",(19..23));
208 my ($H0,$H1,$H2,$H3,$H4) = map("v$_.2s",(24..28));
209 my ($T0,$T1,$MASK) = map("v$_",(29..31));
211 my ($in2,$zeros)=("x16","x17");
212 my $is_base2_26 = $zeros; # borrow
215 .type poly1305_mult,%function
218 mul $d0,$h0,$r0 // h0*r0
221 mul $t0,$h1,$s1 // h1*5*r1
225 mul $t0,$h0,$r1 // h0*r1
230 mul $t0,$h1,$r0 // h1*r0
235 mul $t0,$h2,$s1 // h2*5*r1
237 mul $t1,$h2,$r0 // h2*r0
242 and $t0,$d2,#-4 // final reduction
244 add $t0,$t0,$d2,lsr#2
250 .size poly1305_mult,.-poly1305_mult
252 .type poly1305_splat,%function
255 and x12,$h0,#0x03ffffff // base 2^64 -> base 2^26
258 and x14,x14,#0x03ffffff
262 str w12,[$ctx,#16*0] // r0
263 add w12,w13,w13,lsl#2 // r1*5
264 str w13,[$ctx,#16*1] // r1
265 add w13,w14,w14,lsl#2 // r2*5
266 str w12,[$ctx,#16*2] // s1
267 str w14,[$ctx,#16*3] // r2
268 add w14,w15,w15,lsl#2 // r3*5
269 str w13,[$ctx,#16*4] // s2
270 str w15,[$ctx,#16*5] // r3
271 add w15,w16,w16,lsl#2 // r4*5
272 str w14,[$ctx,#16*6] // s3
273 str w16,[$ctx,#16*7] // r4
274 str w15,[$ctx,#16*8] // s4
277 .size poly1305_splat,.-poly1305_splat
279 .type poly1305_blocks_neon,%function
281 poly1305_blocks_neon:
282 ldr $is_base2_26,[$ctx,#24]
285 cbz $is_base2_26,poly1305_blocks
288 stp x29,x30,[sp,#-80]!
294 cbz $is_base2_26,.Lbase2_64_neon
296 ldp w10,w11,[$ctx] // load hash value base 2^26
297 ldp w12,w13,[$ctx,#8]
303 ldp $r0,$r1,[$ctx,#32] // load key value
305 add $h0,x10,x11,lsl#26 // base 2^26 -> base 2^64
307 adds $h0,$h0,x12,lsl#52
308 add $h1,$h1,x13,lsl#14
311 adds $h1,$h1,x14,lsl#40
312 adc $d2,$h2,xzr // can be partially reduced...
314 ldp $d0,$d1,[$inp],#16 // load input
316 add $s1,$r1,$r1,lsr#2 // s1 = r1 + (r1 >> 2)
318 and $t0,$d2,#-4 // ... so reduce
320 add $t0,$t0,$d2,lsr#2
329 adds $h0,$h0,$d0 // accumulate input
336 cbz $padbit,.Lstore_base2_64_neon
338 and x10,$h0,#0x03ffffff // base 2^64 -> base 2^26
341 and x12,x12,#0x03ffffff
345 cbnz $len,.Leven_neon
347 stp w10,w11,[$ctx] // store hash value base 2^26
348 stp w12,w13,[$ctx,#8]
353 .Lstore_base2_64_neon:
354 stp $h0,$h1,[$ctx] // store hash value base 2^64
355 stp $h2,xzr,[$ctx,#16] // note that is_base2_26 is zeroed
360 ldp $r0,$r1,[$ctx,#32] // load key value
362 ldp $h0,$h1,[$ctx] // load hash value base 2^64
368 ldp $d0,$d1,[$inp],#16 // load input
370 add $s1,$r1,$r1,lsr#2 // s1 = r1 + (r1 >> 2)
375 adds $h0,$h0,$d0 // accumulate input
382 and x10,$h0,#0x03ffffff // base 2^64 -> base 2^26
385 and x12,x12,#0x03ffffff
389 stp d8,d9,[sp,#16] // meet ABI requirements
400 ////////////////////////////////// initialize r^n table
402 add $s1,$r1,$r1,lsr#2 // s1 = r1 + (r1 >> 2)
408 bl poly1305_mult // r^2
412 bl poly1305_mult // r^3
416 bl poly1305_mult // r^4
424 csel $in2,$zeros,$in2,lo
427 str x4,[$ctx,#-24] // set is_base2_26
428 sub $ctx,$ctx,#48 // restore original $ctx
436 csel $in2,$zeros,$in2,lo
438 stp d8,d9,[sp,#16] // meet ABI requirements
450 ldp x8,x12,[$in2],#16 // inp[2:3] (or zero)
451 ldp x9,x13,[$in2],#48
453 lsl $padbit,$padbit,#24
462 and x4,x8,#0x03ffffff // base 2^64 -> base 2^26
463 and x5,x9,#0x03ffffff
466 add x4,x4,x5,lsl#32 // bfi x4,x5,#32,#32
469 add x6,x6,x7,lsl#32 // bfi x6,x7,#32,#32
471 and x8,x8,#0x03ffffff
472 and x9,x9,#0x03ffffff
475 add x12,$padbit,x12,lsr#40
476 add x13,$padbit,x13,lsr#40
477 add x8,x8,x9,lsl#32 // bfi x8,x9,#32,#32
479 add x10,x10,x11,lsl#32 // bfi x10,x11,#32,#32
480 add x12,x12,x13,lsl#32 // bfi x12,x13,#32,#32
485 ldp x8,x12,[$inp],#16 // inp[0:1]
486 ldp x9,x13,[$inp],#48
488 ld1 {$R0,$R1,$S1,$R2},[x15],#64
489 ld1 {$S2,$R3,$S3,$R4},[x15],#64
498 and x4,x8,#0x03ffffff // base 2^64 -> base 2^26
499 and x5,x9,#0x03ffffff
502 add x4,x4,x5,lsl#32 // bfi x4,x5,#32,#32
505 add x6,x6,x7,lsl#32 // bfi x6,x7,#32,#32
507 and x8,x8,#0x03ffffff
508 and x9,x9,#0x03ffffff
511 add x12,$padbit,x12,lsr#40
512 add x13,$padbit,x13,lsr#40
513 add x8,x8,x9,lsl#32 // bfi x8,x9,#32,#32
515 add x10,x10,x11,lsl#32 // bfi x10,x11,#32,#32
516 add x12,x12,x13,lsl#32 // bfi x12,x13,#32,#32
521 ushr $MASK.2d,$MASK.2d,#38
527 ////////////////////////////////////////////////////////////////
528 // ((inp[0]*r^4+inp[2]*r^2+inp[4])*r^4+inp[6]*r^2
529 // ((inp[1]*r^4+inp[3]*r^2+inp[5])*r^3+inp[7]*r
530 // \___________________/
531 // ((inp[0]*r^4+inp[2]*r^2+inp[4])*r^4+inp[6]*r^2+inp[8])*r^2
532 // ((inp[1]*r^4+inp[3]*r^2+inp[5])*r^4+inp[7]*r^2+inp[9])*r
533 // \___________________/ \____________________/
535 // Note that we start with inp[2:3]*r^2. This is because it
536 // doesn't depend on reduction in previous iteration.
537 ////////////////////////////////////////////////////////////////
538 // d4 = h0*r4 + h1*r3 + h2*r2 + h3*r1 + h4*r0
539 // d3 = h0*r3 + h1*r2 + h2*r1 + h3*r0 + h4*5*r4
540 // d2 = h0*r2 + h1*r1 + h2*r0 + h3*5*r4 + h4*5*r3
541 // d1 = h0*r1 + h1*r0 + h2*5*r4 + h3*5*r3 + h4*5*r2
542 // d0 = h0*r0 + h1*5*r4 + h2*5*r3 + h3*5*r2 + h4*5*r1
545 umull $ACC4,$IN23_0,${R4}[2]
546 csel $in2,$zeros,$in2,lo
547 umull $ACC3,$IN23_0,${R3}[2]
548 umull $ACC2,$IN23_0,${R2}[2]
549 ldp x8,x12,[$in2],#16 // inp[2:3] (or zero)
550 umull $ACC1,$IN23_0,${R1}[2]
551 ldp x9,x13,[$in2],#48
552 umull $ACC0,$IN23_0,${R0}[2]
560 umlal $ACC4,$IN23_1,${R3}[2]
561 and x4,x8,#0x03ffffff // base 2^64 -> base 2^26
562 umlal $ACC3,$IN23_1,${R2}[2]
563 and x5,x9,#0x03ffffff
564 umlal $ACC2,$IN23_1,${R1}[2]
566 umlal $ACC1,$IN23_1,${R0}[2]
568 umlal $ACC0,$IN23_1,${S4}[2]
569 add x4,x4,x5,lsl#32 // bfi x4,x5,#32,#32
571 umlal $ACC4,$IN23_2,${R2}[2]
573 umlal $ACC3,$IN23_2,${R1}[2]
575 umlal $ACC2,$IN23_2,${R0}[2]
576 add x6,x6,x7,lsl#32 // bfi x6,x7,#32,#32
577 umlal $ACC1,$IN23_2,${S4}[2]
579 umlal $ACC0,$IN23_2,${S3}[2]
580 and x8,x8,#0x03ffffff
582 umlal $ACC4,$IN23_3,${R1}[2]
583 and x9,x9,#0x03ffffff
584 umlal $ACC3,$IN23_3,${R0}[2]
586 umlal $ACC2,$IN23_3,${S4}[2]
588 umlal $ACC1,$IN23_3,${S3}[2]
589 add x8,x8,x9,lsl#32 // bfi x8,x9,#32,#32
590 umlal $ACC0,$IN23_3,${S2}[2]
593 add $IN01_2,$IN01_2,$H2
594 add x12,$padbit,x12,lsr#40
595 umlal $ACC4,$IN23_4,${R0}[2]
596 add x13,$padbit,x13,lsr#40
597 umlal $ACC3,$IN23_4,${S4}[2]
598 add x10,x10,x11,lsl#32 // bfi x10,x11,#32,#32
599 umlal $ACC2,$IN23_4,${S3}[2]
600 add x12,x12,x13,lsl#32 // bfi x12,x13,#32,#32
601 umlal $ACC1,$IN23_4,${S2}[2]
603 umlal $ACC0,$IN23_4,${S1}[2]
606 ////////////////////////////////////////////////////////////////
607 // (hash+inp[0:1])*r^4 and accumulate
609 add $IN01_0,$IN01_0,$H0
611 umlal $ACC3,$IN01_2,${R1}[0]
612 ldp x8,x12,[$inp],#16 // inp[0:1]
613 umlal $ACC0,$IN01_2,${S3}[0]
614 ldp x9,x13,[$inp],#48
615 umlal $ACC4,$IN01_2,${R2}[0]
616 umlal $ACC1,$IN01_2,${S4}[0]
617 umlal $ACC2,$IN01_2,${R0}[0]
625 add $IN01_1,$IN01_1,$H1
626 umlal $ACC3,$IN01_0,${R3}[0]
627 umlal $ACC4,$IN01_0,${R4}[0]
628 and x4,x8,#0x03ffffff // base 2^64 -> base 2^26
629 umlal $ACC2,$IN01_0,${R2}[0]
630 and x5,x9,#0x03ffffff
631 umlal $ACC0,$IN01_0,${R0}[0]
633 umlal $ACC1,$IN01_0,${R1}[0]
636 add $IN01_3,$IN01_3,$H3
637 add x4,x4,x5,lsl#32 // bfi x4,x5,#32,#32
638 umlal $ACC3,$IN01_1,${R2}[0]
640 umlal $ACC4,$IN01_1,${R3}[0]
642 umlal $ACC0,$IN01_1,${S4}[0]
643 add x6,x6,x7,lsl#32 // bfi x6,x7,#32,#32
644 umlal $ACC2,$IN01_1,${R1}[0]
646 umlal $ACC1,$IN01_1,${R0}[0]
647 and x8,x8,#0x03ffffff
649 add $IN01_4,$IN01_4,$H4
650 and x9,x9,#0x03ffffff
651 umlal $ACC3,$IN01_3,${R0}[0]
653 umlal $ACC0,$IN01_3,${S2}[0]
655 umlal $ACC4,$IN01_3,${R1}[0]
656 add x8,x8,x9,lsl#32 // bfi x8,x9,#32,#32
657 umlal $ACC1,$IN01_3,${S3}[0]
659 umlal $ACC2,$IN01_3,${S4}[0]
660 add x12,$padbit,x12,lsr#40
662 umlal $ACC3,$IN01_4,${S4}[0]
663 add x13,$padbit,x13,lsr#40
664 umlal $ACC0,$IN01_4,${S1}[0]
665 add x10,x10,x11,lsl#32 // bfi x10,x11,#32,#32
666 umlal $ACC4,$IN01_4,${R0}[0]
667 add x12,x12,x13,lsl#32 // bfi x12,x13,#32,#32
668 umlal $ACC1,$IN01_4,${S2}[0]
670 umlal $ACC2,$IN01_4,${S3}[0]
674 /////////////////////////////////////////////////////////////////
675 // lazy reduction as discussed in "NEON crypto" by D.J. Bernstein
678 // [see discussion in poly1305-armv4 module]
680 ushr $T0.2d,$ACC3,#26
682 ushr $T1.2d,$ACC0,#26
683 and $ACC0,$ACC0,$MASK.2d
684 add $ACC4,$ACC4,$T0.2d // h3 -> h4
685 bic $H3,#0xfc,lsl#24 // &=0x03ffffff
686 add $ACC1,$ACC1,$T1.2d // h0 -> h1
688 ushr $T0.2d,$ACC4,#26
690 ushr $T1.2d,$ACC1,#26
693 add $ACC2,$ACC2,$T1.2d // h1 -> h2
695 add $ACC0,$ACC0,$T0.2d
697 shrn $T1.2s,$ACC2,#26
699 add $ACC0,$ACC0,$T0.2d // h4 -> h0
701 add $H3,$H3,$T1.2s // h2 -> h3
704 shrn $T0.2s,$ACC0,#26
709 add $H1,$H1,$T0.2s // h0 -> h1
710 add $H4,$H4,$T1.2s // h3 -> h4
715 dup $IN23_2,${IN23_2}[0]
716 add $IN01_2,$IN01_2,$H2
718 ////////////////////////////////////////////////////////////////
719 // multiply (inp[0:1]+hash) or inp[2:3] by r^2:r^1
724 dup $IN23_2,${IN01_2}[0]
725 add $IN23_0,$IN01_0,$H0
726 add $IN23_3,$IN01_3,$H3
727 add $IN23_1,$IN01_1,$H1
728 add $IN23_4,$IN01_4,$H4
731 dup $IN23_0,${IN23_0}[0]
732 umull2 $ACC0,$IN23_2,${S3}
733 umull2 $ACC3,$IN23_2,${R1}
734 umull2 $ACC4,$IN23_2,${R2}
735 umull2 $ACC2,$IN23_2,${R0}
736 umull2 $ACC1,$IN23_2,${S4}
738 dup $IN23_1,${IN23_1}[0]
739 umlal2 $ACC0,$IN23_0,${R0}
740 umlal2 $ACC2,$IN23_0,${R2}
741 umlal2 $ACC3,$IN23_0,${R3}
742 umlal2 $ACC4,$IN23_0,${R4}
743 umlal2 $ACC1,$IN23_0,${R1}
745 dup $IN23_3,${IN23_3}[0]
746 umlal2 $ACC0,$IN23_1,${S4}
747 umlal2 $ACC3,$IN23_1,${R2}
748 umlal2 $ACC2,$IN23_1,${R1}
749 umlal2 $ACC4,$IN23_1,${R3}
750 umlal2 $ACC1,$IN23_1,${R0}
752 dup $IN23_4,${IN23_4}[0]
753 umlal2 $ACC3,$IN23_3,${R0}
754 umlal2 $ACC4,$IN23_3,${R1}
755 umlal2 $ACC0,$IN23_3,${S2}
756 umlal2 $ACC1,$IN23_3,${S3}
757 umlal2 $ACC2,$IN23_3,${S4}
759 umlal2 $ACC3,$IN23_4,${S4}
760 umlal2 $ACC0,$IN23_4,${S1}
761 umlal2 $ACC4,$IN23_4,${R0}
762 umlal2 $ACC1,$IN23_4,${S2}
763 umlal2 $ACC2,$IN23_4,${S3}
767 ////////////////////////////////////////////////////////////////
768 // (hash+inp[0:1])*r^4:r^3 and accumulate
770 add $IN01_0,$IN01_0,$H0
771 umlal $ACC3,$IN01_2,${R1}
772 umlal $ACC0,$IN01_2,${S3}
773 umlal $ACC4,$IN01_2,${R2}
774 umlal $ACC1,$IN01_2,${S4}
775 umlal $ACC2,$IN01_2,${R0}
777 add $IN01_1,$IN01_1,$H1
778 umlal $ACC3,$IN01_0,${R3}
779 umlal $ACC0,$IN01_0,${R0}
780 umlal $ACC4,$IN01_0,${R4}
781 umlal $ACC1,$IN01_0,${R1}
782 umlal $ACC2,$IN01_0,${R2}
784 add $IN01_3,$IN01_3,$H3
785 umlal $ACC3,$IN01_1,${R2}
786 umlal $ACC0,$IN01_1,${S4}
787 umlal $ACC4,$IN01_1,${R3}
788 umlal $ACC1,$IN01_1,${R0}
789 umlal $ACC2,$IN01_1,${R1}
791 add $IN01_4,$IN01_4,$H4
792 umlal $ACC3,$IN01_3,${R0}
793 umlal $ACC0,$IN01_3,${S2}
794 umlal $ACC4,$IN01_3,${R1}
795 umlal $ACC1,$IN01_3,${S3}
796 umlal $ACC2,$IN01_3,${S4}
798 umlal $ACC3,$IN01_4,${S4}
799 umlal $ACC0,$IN01_4,${S1}
800 umlal $ACC4,$IN01_4,${R0}
801 umlal $ACC1,$IN01_4,${S2}
802 umlal $ACC2,$IN01_4,${S3}
805 ////////////////////////////////////////////////////////////////
808 addp $ACC3,$ACC3,$ACC3
809 ldp d8,d9,[sp,#16] // meet ABI requirements
810 addp $ACC0,$ACC0,$ACC0
812 addp $ACC4,$ACC4,$ACC4
814 addp $ACC1,$ACC1,$ACC1
816 addp $ACC2,$ACC2,$ACC2
818 ////////////////////////////////////////////////////////////////
819 // lazy reduction, but without narrowing
821 ushr $T0.2d,$ACC3,#26
822 and $ACC3,$ACC3,$MASK.2d
823 ushr $T1.2d,$ACC0,#26
824 and $ACC0,$ACC0,$MASK.2d
826 add $ACC4,$ACC4,$T0.2d // h3 -> h4
827 add $ACC1,$ACC1,$T1.2d // h0 -> h1
829 ushr $T0.2d,$ACC4,#26
830 and $ACC4,$ACC4,$MASK.2d
831 ushr $T1.2d,$ACC1,#26
832 and $ACC1,$ACC1,$MASK.2d
833 add $ACC2,$ACC2,$T1.2d // h1 -> h2
835 add $ACC0,$ACC0,$T0.2d
837 ushr $T1.2d,$ACC2,#26
838 and $ACC2,$ACC2,$MASK.2d
839 add $ACC0,$ACC0,$T0.2d // h4 -> h0
840 add $ACC3,$ACC3,$T1.2d // h2 -> h3
842 ushr $T0.2d,$ACC0,#26
843 and $ACC0,$ACC0,$MASK.2d
844 ushr $T1.2d,$ACC3,#26
845 and $ACC3,$ACC3,$MASK.2d
846 add $ACC1,$ACC1,$T0.2d // h0 -> h1
847 add $ACC4,$ACC4,$T1.2d // h3 -> h4
849 ////////////////////////////////////////////////////////////////
850 // write the result, can be partially reduced
852 st4 {$ACC0,$ACC1,$ACC2,$ACC3}[0],[$ctx],#16
853 st1 {$ACC4}[0],[$ctx]
858 .size poly1305_blocks_neon,.-poly1305_blocks_neon
860 .type poly1305_emit_neon,%function
863 ldr $is_base2_26,[$ctx,#24]
864 cbz $is_base2_26,poly1305_emit
866 ldp w10,w11,[$ctx] // load hash value base 2^26
867 ldp w12,w13,[$ctx,#8]
870 add $h0,x10,x11,lsl#26 // base 2^26 -> base 2^64
872 adds $h0,$h0,x12,lsl#52
873 add $h1,$h1,x13,lsl#14
876 adds $h1,$h1,x14,lsl#40
877 adc $h2,$h2,xzr // can be partially reduced...
879 ldp $t0,$t1,[$nonce] // load nonce
881 and $d0,$h2,#-4 // ... so reduce
882 add $d0,$d0,$h2,lsr#2
888 adds $d0,$h0,#5 // compare to modulus
892 tst $d2,#-4 // see if it's carried/borrowed
898 ror $t0,$t0,#32 // flip nonce words
901 adds $h0,$h0,$t0 // accumulate nonce
904 rev $h0,$h0 // flip output bytes
907 stp $h0,$h1,[$mac] // write result
910 .size poly1305_emit_neon,.-poly1305_emit_neon
914 .long 0,0,0,0,0,0,0,0
917 .long OPENSSL_armcap_P-.
919 .quad OPENSSL_armcap_P-.
921 .asciz "Poly1305 for ARMv8, CRYPTOGAMS by <appro\@openssl.org>"
925 foreach (split("\n",$code)) {
926 s/\b(shrn\s+v[0-9]+)\.[24]d/$1.2s/ or
927 s/\b(fmov\s+)v([0-9]+)[^,]*,\s*x([0-9]+)/$1d$2,x$3/ or
928 (m/\bdup\b/ and (s/\.[24]s/.2d/g or 1)) or
929 (m/\b(eor|and)/ and (s/\.[248][sdh]/.16b/g or 1)) or
930 (m/\bum(ul|la)l\b/ and (s/\.4s/.2s/g or 1)) or
931 (m/\bum(ul|la)l2\b/ and (s/\.2s/.4s/g or 1)) or
932 (m/\bst[1-4]\s+{[^}]+}\[/ and (s/\.[24]d/.s/g or 1));
934 s/\.[124]([sd])\[/.$1\[/;