zte's code,first commit
Change-Id: I9a04da59e459a9bc0d67f101f700d9d7dc8d681b
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/README.pod b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/README.pod
new file mode 100644
index 0000000..5d5c4fa
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/README.pod
@@ -0,0 +1,241 @@
+=pod
+
+=head1 NAME
+
+bn_mul_words, bn_mul_add_words, bn_sqr_words, bn_div_words,
+bn_add_words, bn_sub_words, bn_mul_comba4, bn_mul_comba8,
+bn_sqr_comba4, bn_sqr_comba8, bn_cmp_words, bn_mul_normal,
+bn_mul_low_normal, bn_mul_recursive, bn_mul_part_recursive,
+bn_mul_low_recursive, bn_sqr_normal, bn_sqr_recursive,
+bn_expand, bn_wexpand, bn_expand2, bn_fix_top, bn_check_top,
+bn_print, bn_dump, bn_set_max, bn_set_high, bn_set_low - BIGNUM
+library internal functions
+
+=head1 SYNOPSIS
+
+ #include <openssl/bn.h>
+
+ BN_ULONG bn_mul_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w);
+ BN_ULONG bn_mul_add_words(BN_ULONG *rp, BN_ULONG *ap, int num,
+ BN_ULONG w);
+ void bn_sqr_words(BN_ULONG *rp, BN_ULONG *ap, int num);
+ BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d);
+ BN_ULONG bn_add_words(BN_ULONG *rp, BN_ULONG *ap, BN_ULONG *bp,
+ int num);
+ BN_ULONG bn_sub_words(BN_ULONG *rp, BN_ULONG *ap, BN_ULONG *bp,
+ int num);
+
+ void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b);
+ void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b);
+ void bn_sqr_comba4(BN_ULONG *r, BN_ULONG *a);
+ void bn_sqr_comba8(BN_ULONG *r, BN_ULONG *a);
+
+ int bn_cmp_words(BN_ULONG *a, BN_ULONG *b, int n);
+
+ void bn_mul_normal(BN_ULONG *r, BN_ULONG *a, int na, BN_ULONG *b,
+ int nb);
+ void bn_mul_low_normal(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n);
+ void bn_mul_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2,
+ int dna, int dnb, BN_ULONG *tmp);
+ void bn_mul_part_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b,
+ int n, int tna, int tnb, BN_ULONG *tmp);
+ void bn_mul_low_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b,
+ int n2, BN_ULONG *tmp);
+
+ void bn_sqr_normal(BN_ULONG *r, BN_ULONG *a, int n, BN_ULONG *tmp);
+ void bn_sqr_recursive(BN_ULONG *r, BN_ULONG *a, int n2, BN_ULONG *tmp);
+
+ void mul(BN_ULONG r, BN_ULONG a, BN_ULONG w, BN_ULONG c);
+ void mul_add(BN_ULONG r, BN_ULONG a, BN_ULONG w, BN_ULONG c);
+ void sqr(BN_ULONG r0, BN_ULONG r1, BN_ULONG a);
+
+ BIGNUM *bn_expand(BIGNUM *a, int bits);
+ BIGNUM *bn_wexpand(BIGNUM *a, int n);
+ BIGNUM *bn_expand2(BIGNUM *a, int n);
+ void bn_fix_top(BIGNUM *a);
+
+ void bn_check_top(BIGNUM *a);
+ void bn_print(BIGNUM *a);
+ void bn_dump(BN_ULONG *d, int n);
+ void bn_set_max(BIGNUM *a);
+ void bn_set_high(BIGNUM *r, BIGNUM *a, int n);
+ void bn_set_low(BIGNUM *r, BIGNUM *a, int n);
+
+=head1 DESCRIPTION
+
+This page documents the internal functions used by the OpenSSL
+B<BIGNUM> implementation. They are described here to facilitate
+debugging and extending the library. They are I<not> to be used by
+applications.
+
+=head2 The BIGNUM structure
+
+ typedef struct bignum_st BIGNUM;
+
+ struct bignum_st
+ {
+ BN_ULONG *d; /* Pointer to an array of 'BN_BITS2' bit chunks. */
+ int top; /* Index of last used d +1. */
+ /* The next are internal book keeping for bn_expand. */
+ int dmax; /* Size of the d array. */
+ int neg; /* one if the number is negative */
+ int flags;
+ };
+
+
+The integer value is stored in B<d>, a malloc()ed array of words (B<BN_ULONG>),
+least significant word first. A B<BN_ULONG> can be either 16, 32 or 64 bits
+in size, depending on the 'number of bits' (B<BITS2>) specified in
+C<openssl/bn.h>.
+
+B<dmax> is the size of the B<d> array that has been allocated. B<top>
+is the number of words being used, so for a value of 4, bn.d[0]=4 and
+bn.top=1. B<neg> is 1 if the number is negative. When a B<BIGNUM> is
+B<0>, the B<d> field can be B<NULL> and B<top> == B<0>.
+
+B<flags> is a bit field of flags which are defined in C<openssl/bn.h>. The
+flags begin with B<BN_FLG_>. The macros BN_set_flags(b, n) and
+BN_get_flags(b, n) exist to enable or fetch flag(s) B<n> from B<BIGNUM>
+structure B<b>.
+
+Various routines in this library require the use of temporary
+B<BIGNUM> variables during their execution. Since dynamic memory
+allocation to create B<BIGNUM>s is rather expensive when used in
+conjunction with repeated subroutine calls, the B<BN_CTX> structure is
+used. This structure contains B<BN_CTX_NUM> B<BIGNUM>s, see
+L<BN_CTX_start(3)>.
+
+=head2 Low-level arithmetic operations
+
+These functions are implemented in C and for several platforms in
+assembly language:
+
+bn_mul_words(B<rp>, B<ap>, B<num>, B<w>) operates on the B<num> word
+arrays B<rp> and B<ap>. It computes B<ap> * B<w>, places the result
+in B<rp>, and returns the high word (carry).
+
+bn_mul_add_words(B<rp>, B<ap>, B<num>, B<w>) operates on the B<num>
+word arrays B<rp> and B<ap>. It computes B<ap> * B<w> + B<rp>, places
+the result in B<rp>, and returns the high word (carry).
+
+bn_sqr_words(B<rp>, B<ap>, B<n>) operates on the B<num> word array
+B<ap> and the 2*B<num> word array B<ap>. It computes B<ap> * B<ap>
+word-wise, and places the low and high bytes of the result in B<rp>.
+
+bn_div_words(B<h>, B<l>, B<d>) divides the two word number (B<h>, B<l>)
+by B<d> and returns the result.
+
+bn_add_words(B<rp>, B<ap>, B<bp>, B<num>) operates on the B<num> word
+arrays B<ap>, B<bp> and B<rp>. It computes B<ap> + B<bp>, places the
+result in B<rp>, and returns the high word (carry).
+
+bn_sub_words(B<rp>, B<ap>, B<bp>, B<num>) operates on the B<num> word
+arrays B<ap>, B<bp> and B<rp>. It computes B<ap> - B<bp>, places the
+result in B<rp>, and returns the carry (1 if B<bp> E<gt> B<ap>, 0
+otherwise).
+
+bn_mul_comba4(B<r>, B<a>, B<b>) operates on the 4 word arrays B<a> and
+B<b> and the 8 word array B<r>. It computes B<a>*B<b> and places the
+result in B<r>.
+
+bn_mul_comba8(B<r>, B<a>, B<b>) operates on the 8 word arrays B<a> and
+B<b> and the 16 word array B<r>. It computes B<a>*B<b> and places the
+result in B<r>.
+
+bn_sqr_comba4(B<r>, B<a>, B<b>) operates on the 4 word arrays B<a> and
+B<b> and the 8 word array B<r>.
+
+bn_sqr_comba8(B<r>, B<a>, B<b>) operates on the 8 word arrays B<a> and
+B<b> and the 16 word array B<r>.
+
+The following functions are implemented in C:
+
+bn_cmp_words(B<a>, B<b>, B<n>) operates on the B<n> word arrays B<a>
+and B<b>. It returns 1, 0 and -1 if B<a> is greater than, equal and
+less than B<b>.
+
+bn_mul_normal(B<r>, B<a>, B<na>, B<b>, B<nb>) operates on the B<na>
+word array B<a>, the B<nb> word array B<b> and the B<na>+B<nb> word
+array B<r>. It computes B<a>*B<b> and places the result in B<r>.
+
+bn_mul_low_normal(B<r>, B<a>, B<b>, B<n>) operates on the B<n> word
+arrays B<r>, B<a> and B<b>. It computes the B<n> low words of
+B<a>*B<b> and places the result in B<r>.
+
+bn_mul_recursive(B<r>, B<a>, B<b>, B<n2>, B<dna>, B<dnb>, B<t>) operates
+on the word arrays B<a> and B<b> of length B<n2>+B<dna> and B<n2>+B<dnb>
+(B<dna> and B<dnb> are currently allowed to be 0 or negative) and the 2*B<n2>
+word arrays B<r> and B<t>. B<n2> must be a power of 2. It computes
+B<a>*B<b> and places the result in B<r>.
+
+bn_mul_part_recursive(B<r>, B<a>, B<b>, B<n>, B<tna>, B<tnb>, B<tmp>)
+operates on the word arrays B<a> and B<b> of length B<n>+B<tna> and
+B<n>+B<tnb> and the 4*B<n> word arrays B<r> and B<tmp>.
+
+bn_mul_low_recursive(B<r>, B<a>, B<b>, B<n2>, B<tmp>) operates on the
+B<n2> word arrays B<r> and B<tmp> and the B<n2>/2 word arrays B<a>
+and B<b>.
+
+BN_mul() calls bn_mul_normal(), or an optimized implementation if the
+factors have the same size: bn_mul_comba8() is used if they are 8
+words long, bn_mul_recursive() if they are larger than
+B<BN_MULL_SIZE_NORMAL> and the size is an exact multiple of the word
+size, and bn_mul_part_recursive() for others that are larger than
+B<BN_MULL_SIZE_NORMAL>.
+
+bn_sqr_normal(B<r>, B<a>, B<n>, B<tmp>) operates on the B<n> word array
+B<a> and the 2*B<n> word arrays B<tmp> and B<r>.
+
+The implementations use the following macros which, depending on the
+architecture, may use "long long" C operations or inline assembler.
+They are defined in C<bn_local.h>.
+
+mul(B<r>, B<a>, B<w>, B<c>) computes B<w>*B<a>+B<c> and places the
+low word of the result in B<r> and the high word in B<c>.
+
+mul_add(B<r>, B<a>, B<w>, B<c>) computes B<w>*B<a>+B<r>+B<c> and
+places the low word of the result in B<r> and the high word in B<c>.
+
+sqr(B<r0>, B<r1>, B<a>) computes B<a>*B<a> and places the low word
+of the result in B<r0> and the high word in B<r1>.
+
+=head2 Size changes
+
+bn_expand() ensures that B<b> has enough space for a B<bits> bit
+number. bn_wexpand() ensures that B<b> has enough space for an
+B<n> word number. If the number has to be expanded, both macros
+call bn_expand2(), which allocates a new B<d> array and copies the
+data. They return B<NULL> on error, B<b> otherwise.
+
+The bn_fix_top() macro reduces B<a-E<gt>top> to point to the most
+significant non-zero word plus one when B<a> has shrunk.
+
+=head2 Debugging
+
+bn_check_top() verifies that C<((a)-E<gt>top E<gt>= 0 && (a)-E<gt>top
+E<lt>= (a)-E<gt>dmax)>. A violation will cause the program to abort.
+
+bn_print() prints B<a> to stderr. bn_dump() prints B<n> words at B<d>
+(in reverse order, i.e. most significant word first) to stderr.
+
+bn_set_max() makes B<a> a static number with a B<dmax> of its current size.
+This is used by bn_set_low() and bn_set_high() to make B<r> a read-only
+B<BIGNUM> that contains the B<n> low or high words of B<a>.
+
+If B<BN_DEBUG> is not defined, bn_check_top(), bn_print(), bn_dump()
+and bn_set_max() are defined as empty macros.
+
+=head1 SEE ALSO
+
+L<bn(3)>
+
+=head1 COPYRIGHT
+
+Copyright 2000-2016 The OpenSSL Project Authors. All Rights Reserved.
+
+Licensed under the OpenSSL license (the "License"). You may not use
+this file except in compliance with the License. You can obtain a copy
+in the file LICENSE in the source distribution or at
+L<https://www.openssl.org/source/license.html>.
+
+=cut
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/alpha-mont.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/alpha-mont.pl
new file mode 100644
index 0000000..8371792
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/alpha-mont.pl
@@ -0,0 +1,328 @@
+#! /usr/bin/env perl
+# Copyright 2006-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+#
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+#
+# On 21264 RSA sign performance improves by 70/35/20/15 percent for
+# 512/1024/2048/4096 bit key lengths. This is against vendor compiler
+# instructed to '-tune host' code with in-line assembler. Other
+# benchmarks improve by 15-20%. To anchor it to something else, the
+# code provides approximately the same performance per GHz as AMD64.
+# I.e. if you compare 1GHz 21264 and 2GHz Opteron, you'll observe ~2x
+# difference.
+
+$output=pop;
+open STDOUT,">$output";
+
+# int bn_mul_mont(
+$rp="a0"; # BN_ULONG *rp,
+$ap="a1"; # const BN_ULONG *ap,
+$bp="a2"; # const BN_ULONG *bp,
+$np="a3"; # const BN_ULONG *np,
+$n0="a4"; # const BN_ULONG *n0,
+$num="a5"; # int num);
+
+$lo0="t0";
+$hi0="t1";
+$lo1="t2";
+$hi1="t3";
+$aj="t4";
+$bi="t5";
+$nj="t6";
+$tp="t7";
+$alo="t8";
+$ahi="t9";
+$nlo="t10";
+$nhi="t11";
+$tj="t12";
+$i="s3";
+$j="s4";
+$m1="s5";
+
+$code=<<___;
+#ifdef __linux__
+#include <asm/regdef.h>
+#else
+#include <asm.h>
+#include <regdef.h>
+#endif
+
+.text
+
+.set noat
+.set noreorder
+
+.globl bn_mul_mont
+.align 5
+.ent bn_mul_mont
+bn_mul_mont:
+ lda sp,-48(sp)
+ stq ra,0(sp)
+ stq s3,8(sp)
+ stq s4,16(sp)
+ stq s5,24(sp)
+ stq fp,32(sp)
+ mov sp,fp
+ .mask 0x0400f000,-48
+ .frame fp,48,ra
+ .prologue 0
+
+ .align 4
+ .set reorder
+ sextl $num,$num
+ mov 0,v0
+ cmplt $num,4,AT
+ bne AT,.Lexit
+
+ ldq $hi0,0($ap) # ap[0]
+ s8addq $num,16,AT
+ ldq $aj,8($ap)
+ subq sp,AT,sp
+ ldq $bi,0($bp) # bp[0]
+ lda AT,-4096(zero) # mov -4096,AT
+ ldq $n0,0($n0)
+ and sp,AT,sp
+
+ mulq $hi0,$bi,$lo0
+ ldq $hi1,0($np) # np[0]
+ umulh $hi0,$bi,$hi0
+ ldq $nj,8($np)
+
+ mulq $lo0,$n0,$m1
+
+ mulq $hi1,$m1,$lo1
+ umulh $hi1,$m1,$hi1
+
+ addq $lo1,$lo0,$lo1
+ cmpult $lo1,$lo0,AT
+ addq $hi1,AT,$hi1
+
+ mulq $aj,$bi,$alo
+ mov 2,$j
+ umulh $aj,$bi,$ahi
+ mov sp,$tp
+
+ mulq $nj,$m1,$nlo
+ s8addq $j,$ap,$aj
+ umulh $nj,$m1,$nhi
+ s8addq $j,$np,$nj
+.align 4
+.L1st:
+ .set noreorder
+ ldq $aj,0($aj)
+ addl $j,1,$j
+ ldq $nj,0($nj)
+ lda $tp,8($tp)
+
+ addq $alo,$hi0,$lo0
+ mulq $aj,$bi,$alo
+ cmpult $lo0,$hi0,AT
+ addq $nlo,$hi1,$lo1
+
+ mulq $nj,$m1,$nlo
+ addq $ahi,AT,$hi0
+ cmpult $lo1,$hi1,v0
+ cmplt $j,$num,$tj
+
+ umulh $aj,$bi,$ahi
+ addq $nhi,v0,$hi1
+ addq $lo1,$lo0,$lo1
+ s8addq $j,$ap,$aj
+
+ umulh $nj,$m1,$nhi
+ cmpult $lo1,$lo0,v0
+ addq $hi1,v0,$hi1
+ s8addq $j,$np,$nj
+
+ stq $lo1,-8($tp)
+ nop
+ unop
+ bne $tj,.L1st
+ .set reorder
+
+ addq $alo,$hi0,$lo0
+ addq $nlo,$hi1,$lo1
+ cmpult $lo0,$hi0,AT
+ cmpult $lo1,$hi1,v0
+ addq $ahi,AT,$hi0
+ addq $nhi,v0,$hi1
+
+ addq $lo1,$lo0,$lo1
+ cmpult $lo1,$lo0,v0
+ addq $hi1,v0,$hi1
+
+ stq $lo1,0($tp)
+
+ addq $hi1,$hi0,$hi1
+ cmpult $hi1,$hi0,AT
+ stq $hi1,8($tp)
+ stq AT,16($tp)
+
+ mov 1,$i
+.align 4
+.Louter:
+ s8addq $i,$bp,$bi
+ ldq $hi0,0($ap)
+ ldq $aj,8($ap)
+ ldq $bi,0($bi)
+ ldq $hi1,0($np)
+ ldq $nj,8($np)
+ ldq $tj,0(sp)
+
+ mulq $hi0,$bi,$lo0
+ umulh $hi0,$bi,$hi0
+
+ addq $lo0,$tj,$lo0
+ cmpult $lo0,$tj,AT
+ addq $hi0,AT,$hi0
+
+ mulq $lo0,$n0,$m1
+
+ mulq $hi1,$m1,$lo1
+ umulh $hi1,$m1,$hi1
+
+ addq $lo1,$lo0,$lo1
+ cmpult $lo1,$lo0,AT
+ mov 2,$j
+ addq $hi1,AT,$hi1
+
+ mulq $aj,$bi,$alo
+ mov sp,$tp
+ umulh $aj,$bi,$ahi
+
+ mulq $nj,$m1,$nlo
+ s8addq $j,$ap,$aj
+ umulh $nj,$m1,$nhi
+.align 4
+.Linner:
+ .set noreorder
+ ldq $tj,8($tp) #L0
+ nop #U1
+ ldq $aj,0($aj) #L1
+ s8addq $j,$np,$nj #U0
+
+ ldq $nj,0($nj) #L0
+ nop #U1
+ addq $alo,$hi0,$lo0 #L1
+ lda $tp,8($tp)
+
+ mulq $aj,$bi,$alo #U1
+ cmpult $lo0,$hi0,AT #L0
+ addq $nlo,$hi1,$lo1 #L1
+ addl $j,1,$j
+
+ mulq $nj,$m1,$nlo #U1
+ addq $ahi,AT,$hi0 #L0
+ addq $lo0,$tj,$lo0 #L1
+ cmpult $lo1,$hi1,v0 #U0
+
+ umulh $aj,$bi,$ahi #U1
+ cmpult $lo0,$tj,AT #L0
+ addq $lo1,$lo0,$lo1 #L1
+ addq $nhi,v0,$hi1 #U0
+
+ umulh $nj,$m1,$nhi #U1
+ s8addq $j,$ap,$aj #L0
+ cmpult $lo1,$lo0,v0 #L1
+ cmplt $j,$num,$tj #U0 # borrow $tj
+
+ addq $hi0,AT,$hi0 #L0
+ addq $hi1,v0,$hi1 #U1
+ stq $lo1,-8($tp) #L1
+ bne $tj,.Linner #U0
+ .set reorder
+
+ ldq $tj,8($tp)
+ addq $alo,$hi0,$lo0
+ addq $nlo,$hi1,$lo1
+ cmpult $lo0,$hi0,AT
+ cmpult $lo1,$hi1,v0
+ addq $ahi,AT,$hi0
+ addq $nhi,v0,$hi1
+
+ addq $lo0,$tj,$lo0
+ cmpult $lo0,$tj,AT
+ addq $hi0,AT,$hi0
+
+ ldq $tj,16($tp)
+ addq $lo1,$lo0,$j
+ cmpult $j,$lo0,v0
+ addq $hi1,v0,$hi1
+
+ addq $hi1,$hi0,$lo1
+ stq $j,0($tp)
+ cmpult $lo1,$hi0,$hi1
+ addq $lo1,$tj,$lo1
+ cmpult $lo1,$tj,AT
+ addl $i,1,$i
+ addq $hi1,AT,$hi1
+ stq $lo1,8($tp)
+ cmplt $i,$num,$tj # borrow $tj
+ stq $hi1,16($tp)
+ bne $tj,.Louter
+
+ s8addq $num,sp,$tj # &tp[num]
+ mov $rp,$bp # put rp aside
+ mov sp,$tp
+ mov sp,$ap
+ mov 0,$hi0 # clear borrow bit
+
+.align 4
+.Lsub: ldq $lo0,0($tp)
+ ldq $lo1,0($np)
+ lda $tp,8($tp)
+ lda $np,8($np)
+ subq $lo0,$lo1,$lo1 # tp[i]-np[i]
+ cmpult $lo0,$lo1,AT
+ subq $lo1,$hi0,$lo0
+ cmpult $lo1,$lo0,$hi0
+ or $hi0,AT,$hi0
+ stq $lo0,0($rp)
+ cmpult $tp,$tj,v0
+ lda $rp,8($rp)
+ bne v0,.Lsub
+
+ subq $hi1,$hi0,$hi0 # handle upmost overflow bit
+ mov sp,$tp
+ mov $bp,$rp # restore rp
+
+.align 4
+.Lcopy: ldq $aj,0($tp) # conditional copy
+ ldq $nj,0($rp)
+ lda $tp,8($tp)
+ lda $rp,8($rp)
+ cmoveq $hi0,$nj,$aj
+ stq zero,-8($tp) # zap tp
+ cmpult $tp,$tj,AT
+ stq $aj,-8($rp)
+ bne AT,.Lcopy
+ mov 1,v0
+
+.Lexit:
+ .set noreorder
+ mov fp,sp
+ /*ldq ra,0(sp)*/
+ ldq s3,8(sp)
+ ldq s4,16(sp)
+ ldq s5,24(sp)
+ ldq fp,32(sp)
+ lda sp,48(sp)
+ ret (ra)
+.end bn_mul_mont
+.ascii "Montgomery Multiplication for Alpha, CRYPTOGAMS by <appro\@openssl.org>"
+.align 2
+___
+
+print $code;
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/armv4-gf2m.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/armv4-gf2m.pl
new file mode 100644
index 0000000..3a83cb8
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/armv4-gf2m.pl
@@ -0,0 +1,332 @@
+#! /usr/bin/env perl
+# Copyright 2011-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+#
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+#
+# May 2011
+#
+# The module implements bn_GF2m_mul_2x2 polynomial multiplication
+# used in bn_gf2m.c. It's kind of low-hanging mechanical port from
+# C for the time being... Except that it has two code paths: pure
+# integer code suitable for any ARMv4 and later CPU and NEON code
+# suitable for ARMv7. Pure integer 1x1 multiplication subroutine runs
+# in ~45 cycles on dual-issue core such as Cortex A8, which is ~50%
+# faster than compiler-generated code. For ECDH and ECDSA verify (but
+# not for ECDSA sign) it means 25%-45% improvement depending on key
+# length, more for longer keys. Even though NEON 1x1 multiplication
+# runs in even less cycles, ~30, improvement is measurable only on
+# longer keys. One has to optimize code elsewhere to get NEON glow...
+#
+# April 2014
+#
+# Double bn_GF2m_mul_2x2 performance by using algorithm from paper
+# referred below, which improves ECDH and ECDSA verify benchmarks
+# by 18-40%.
+#
+# Câmara, D.; Gouvêa, C. P. L.; López, J. & Dahab, R.: Fast Software
+# Polynomial Multiplication on ARM Processors using the NEON Engine.
+#
+# http://conradoplg.cryptoland.net/files/2010/12/mocrysen13.pdf
+
+$flavour = shift;
+if ($flavour=~/\w[\w\-]*\.\w+$/) { $output=$flavour; undef $flavour; }
+else { while (($output=shift) && ($output!~/\w[\w\-]*\.\w+$/)) {} }
+
+if ($flavour && $flavour ne "void") {
+ $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+ ( $xlate="${dir}arm-xlate.pl" and -f $xlate ) or
+ ( $xlate="${dir}../../perlasm/arm-xlate.pl" and -f $xlate) or
+ die "can't locate arm-xlate.pl";
+
+ open STDOUT,"| \"$^X\" $xlate $flavour $output";
+} else {
+ open STDOUT,">$output";
+}
+
+$code=<<___;
+#include "arm_arch.h"
+
+.text
+#if defined(__thumb2__)
+.syntax unified
+.thumb
+#else
+.code 32
+#endif
+___
+################
+# private interface to mul_1x1_ialu
+#
+$a="r1";
+$b="r0";
+
+($a0,$a1,$a2,$a12,$a4,$a14)=
+($hi,$lo,$t0,$t1, $i0,$i1 )=map("r$_",(4..9),12);
+
+$mask="r12";
+
+$code.=<<___;
+.type mul_1x1_ialu,%function
+.align 5
+mul_1x1_ialu:
+ mov $a0,#0
+ bic $a1,$a,#3<<30 @ a1=a&0x3fffffff
+ str $a0,[sp,#0] @ tab[0]=0
+ add $a2,$a1,$a1 @ a2=a1<<1
+ str $a1,[sp,#4] @ tab[1]=a1
+ eor $a12,$a1,$a2 @ a1^a2
+ str $a2,[sp,#8] @ tab[2]=a2
+ mov $a4,$a1,lsl#2 @ a4=a1<<2
+ str $a12,[sp,#12] @ tab[3]=a1^a2
+ eor $a14,$a1,$a4 @ a1^a4
+ str $a4,[sp,#16] @ tab[4]=a4
+ eor $a0,$a2,$a4 @ a2^a4
+ str $a14,[sp,#20] @ tab[5]=a1^a4
+ eor $a12,$a12,$a4 @ a1^a2^a4
+ str $a0,[sp,#24] @ tab[6]=a2^a4
+ and $i0,$mask,$b,lsl#2
+ str $a12,[sp,#28] @ tab[7]=a1^a2^a4
+
+ and $i1,$mask,$b,lsr#1
+ ldr $lo,[sp,$i0] @ tab[b & 0x7]
+ and $i0,$mask,$b,lsr#4
+ ldr $t1,[sp,$i1] @ tab[b >> 3 & 0x7]
+ and $i1,$mask,$b,lsr#7
+ ldr $t0,[sp,$i0] @ tab[b >> 6 & 0x7]
+ eor $lo,$lo,$t1,lsl#3 @ stall
+ mov $hi,$t1,lsr#29
+ ldr $t1,[sp,$i1] @ tab[b >> 9 & 0x7]
+
+ and $i0,$mask,$b,lsr#10
+ eor $lo,$lo,$t0,lsl#6
+ eor $hi,$hi,$t0,lsr#26
+ ldr $t0,[sp,$i0] @ tab[b >> 12 & 0x7]
+
+ and $i1,$mask,$b,lsr#13
+ eor $lo,$lo,$t1,lsl#9
+ eor $hi,$hi,$t1,lsr#23
+ ldr $t1,[sp,$i1] @ tab[b >> 15 & 0x7]
+
+ and $i0,$mask,$b,lsr#16
+ eor $lo,$lo,$t0,lsl#12
+ eor $hi,$hi,$t0,lsr#20
+ ldr $t0,[sp,$i0] @ tab[b >> 18 & 0x7]
+
+ and $i1,$mask,$b,lsr#19
+ eor $lo,$lo,$t1,lsl#15
+ eor $hi,$hi,$t1,lsr#17
+ ldr $t1,[sp,$i1] @ tab[b >> 21 & 0x7]
+
+ and $i0,$mask,$b,lsr#22
+ eor $lo,$lo,$t0,lsl#18
+ eor $hi,$hi,$t0,lsr#14
+ ldr $t0,[sp,$i0] @ tab[b >> 24 & 0x7]
+
+ and $i1,$mask,$b,lsr#25
+ eor $lo,$lo,$t1,lsl#21
+ eor $hi,$hi,$t1,lsr#11
+ ldr $t1,[sp,$i1] @ tab[b >> 27 & 0x7]
+
+ tst $a,#1<<30
+ and $i0,$mask,$b,lsr#28
+ eor $lo,$lo,$t0,lsl#24
+ eor $hi,$hi,$t0,lsr#8
+ ldr $t0,[sp,$i0] @ tab[b >> 30 ]
+
+#ifdef __thumb2__
+ itt ne
+#endif
+ eorne $lo,$lo,$b,lsl#30
+ eorne $hi,$hi,$b,lsr#2
+ tst $a,#1<<31
+ eor $lo,$lo,$t1,lsl#27
+ eor $hi,$hi,$t1,lsr#5
+#ifdef __thumb2__
+ itt ne
+#endif
+ eorne $lo,$lo,$b,lsl#31
+ eorne $hi,$hi,$b,lsr#1
+ eor $lo,$lo,$t0,lsl#30
+ eor $hi,$hi,$t0,lsr#2
+
+ mov pc,lr
+.size mul_1x1_ialu,.-mul_1x1_ialu
+___
+################
+# void bn_GF2m_mul_2x2(BN_ULONG *r,
+# BN_ULONG a1,BN_ULONG a0,
+# BN_ULONG b1,BN_ULONG b0); # r[3..0]=a1a0·b1b0
+{
+$code.=<<___;
+.global bn_GF2m_mul_2x2
+.type bn_GF2m_mul_2x2,%function
+.align 5
+bn_GF2m_mul_2x2:
+#if __ARM_MAX_ARCH__>=7
+ stmdb sp!,{r10,lr}
+ ldr r12,.LOPENSSL_armcap
+ adr r10,.LOPENSSL_armcap
+ ldr r12,[r12,r10]
+#ifdef __APPLE__
+ ldr r12,[r12]
+#endif
+ tst r12,#ARMV7_NEON
+ itt ne
+ ldrne r10,[sp],#8
+ bne .LNEON
+ stmdb sp!,{r4-r9}
+#else
+ stmdb sp!,{r4-r10,lr}
+#endif
+___
+$ret="r10"; # reassigned 1st argument
+$code.=<<___;
+ mov $ret,r0 @ reassign 1st argument
+ mov $b,r3 @ $b=b1
+ sub r7,sp,#36
+ mov r8,sp
+ and r7,r7,#-32
+ ldr r3,[sp,#32] @ load b0
+ mov $mask,#7<<2
+ mov sp,r7 @ allocate tab[8]
+ str r8,[r7,#32]
+
+ bl mul_1x1_ialu @ a1·b1
+ str $lo,[$ret,#8]
+ str $hi,[$ret,#12]
+
+ eor $b,$b,r3 @ flip b0 and b1
+ eor $a,$a,r2 @ flip a0 and a1
+ eor r3,r3,$b
+ eor r2,r2,$a
+ eor $b,$b,r3
+ eor $a,$a,r2
+ bl mul_1x1_ialu @ a0·b0
+ str $lo,[$ret]
+ str $hi,[$ret,#4]
+
+ eor $a,$a,r2
+ eor $b,$b,r3
+ bl mul_1x1_ialu @ (a1+a0)·(b1+b0)
+___
+@r=map("r$_",(6..9));
+$code.=<<___;
+ ldmia $ret,{@r[0]-@r[3]}
+ eor $lo,$lo,$hi
+ ldr sp,[sp,#32] @ destroy tab[8]
+ eor $hi,$hi,@r[1]
+ eor $lo,$lo,@r[0]
+ eor $hi,$hi,@r[2]
+ eor $lo,$lo,@r[3]
+ eor $hi,$hi,@r[3]
+ str $hi,[$ret,#8]
+ eor $lo,$lo,$hi
+ str $lo,[$ret,#4]
+
+#if __ARM_ARCH__>=5
+ ldmia sp!,{r4-r10,pc}
+#else
+ ldmia sp!,{r4-r10,lr}
+ tst lr,#1
+ moveq pc,lr @ be binary compatible with V4, yet
+ bx lr @ interoperable with Thumb ISA:-)
+#endif
+___
+}
+{
+my ($r,$t0,$t1,$t2,$t3)=map("q$_",(0..3,8..12));
+my ($a,$b,$k48,$k32,$k16)=map("d$_",(26..31));
+
+$code.=<<___;
+#if __ARM_MAX_ARCH__>=7
+.arch armv7-a
+.fpu neon
+
+.align 5
+.LNEON:
+ ldr r12, [sp] @ 5th argument
+ vmov $a, r2, r1
+ vmov $b, r12, r3
+ vmov.i64 $k48, #0x0000ffffffffffff
+ vmov.i64 $k32, #0x00000000ffffffff
+ vmov.i64 $k16, #0x000000000000ffff
+
+ vext.8 $t0#lo, $a, $a, #1 @ A1
+ vmull.p8 $t0, $t0#lo, $b @ F = A1*B
+ vext.8 $r#lo, $b, $b, #1 @ B1
+ vmull.p8 $r, $a, $r#lo @ E = A*B1
+ vext.8 $t1#lo, $a, $a, #2 @ A2
+ vmull.p8 $t1, $t1#lo, $b @ H = A2*B
+ vext.8 $t3#lo, $b, $b, #2 @ B2
+ vmull.p8 $t3, $a, $t3#lo @ G = A*B2
+ vext.8 $t2#lo, $a, $a, #3 @ A3
+ veor $t0, $t0, $r @ L = E + F
+ vmull.p8 $t2, $t2#lo, $b @ J = A3*B
+ vext.8 $r#lo, $b, $b, #3 @ B3
+ veor $t1, $t1, $t3 @ M = G + H
+ vmull.p8 $r, $a, $r#lo @ I = A*B3
+ veor $t0#lo, $t0#lo, $t0#hi @ t0 = (L) (P0 + P1) << 8
+ vand $t0#hi, $t0#hi, $k48
+ vext.8 $t3#lo, $b, $b, #4 @ B4
+ veor $t1#lo, $t1#lo, $t1#hi @ t1 = (M) (P2 + P3) << 16
+ vand $t1#hi, $t1#hi, $k32
+ vmull.p8 $t3, $a, $t3#lo @ K = A*B4
+ veor $t2, $t2, $r @ N = I + J
+ veor $t0#lo, $t0#lo, $t0#hi
+ veor $t1#lo, $t1#lo, $t1#hi
+ veor $t2#lo, $t2#lo, $t2#hi @ t2 = (N) (P4 + P5) << 24
+ vand $t2#hi, $t2#hi, $k16
+ vext.8 $t0, $t0, $t0, #15
+ veor $t3#lo, $t3#lo, $t3#hi @ t3 = (K) (P6 + P7) << 32
+ vmov.i64 $t3#hi, #0
+ vext.8 $t1, $t1, $t1, #14
+ veor $t2#lo, $t2#lo, $t2#hi
+ vmull.p8 $r, $a, $b @ D = A*B
+ vext.8 $t3, $t3, $t3, #12
+ vext.8 $t2, $t2, $t2, #13
+ veor $t0, $t0, $t1
+ veor $t2, $t2, $t3
+ veor $r, $r, $t0
+ veor $r, $r, $t2
+
+ vst1.32 {$r}, [r0]
+ ret @ bx lr
+#endif
+___
+}
+$code.=<<___;
+.size bn_GF2m_mul_2x2,.-bn_GF2m_mul_2x2
+#if __ARM_MAX_ARCH__>=7
+.align 5
+.LOPENSSL_armcap:
+.word OPENSSL_armcap_P-.
+#endif
+.asciz "GF(2^m) Multiplication for ARMv4/NEON, CRYPTOGAMS by <appro\@openssl.org>"
+.align 5
+
+#if __ARM_MAX_ARCH__>=7
+.comm OPENSSL_armcap_P,4,4
+#endif
+___
+
+foreach (split("\n",$code)) {
+ s/\`([^\`]*)\`/eval $1/geo;
+
+ s/\bq([0-9]+)#(lo|hi)/sprintf "d%d",2*$1+($2 eq "hi")/geo or
+ s/\bret\b/bx lr/go or
+ s/\bbx\s+lr\b/.word\t0xe12fff1e/go; # make it possible to compile with -march=armv4
+
+ print $_,"\n";
+}
+close STDOUT or die "error closing STDOUT: $!"; # enforce flush
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/armv4-mont.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/armv4-mont.pl
new file mode 100644
index 0000000..eadc8bb
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/armv4-mont.pl
@@ -0,0 +1,757 @@
+#! /usr/bin/env perl
+# Copyright 2007-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+
+# January 2007.
+
+# Montgomery multiplication for ARMv4.
+#
+# Performance improvement naturally varies among CPU implementations
+# and compilers. The code was observed to provide +65-35% improvement
+# [depending on key length, less for longer keys] on ARM920T, and
+# +115-80% on Intel IXP425. This is compared to pre-bn_mul_mont code
+# base and compiler generated code with in-lined umull and even umlal
+# instructions. The latter means that this code didn't really have an
+# "advantage" of utilizing some "secret" instruction.
+#
+# The code is interoperable with Thumb ISA and is rather compact, less
+# than 1/2KB. Windows CE port would be trivial, as it's exclusively
+# about decorations, ABI and instruction syntax are identical.
+
+# November 2013
+#
+# Add NEON code path, which handles lengths divisible by 8. RSA/DSA
+# performance improvement on Cortex-A8 is ~45-100% depending on key
+# length, more for longer keys. On Cortex-A15 the span is ~10-105%.
+# On Snapdragon S4 improvement was measured to vary from ~70% to
+# incredible ~380%, yes, 4.8x faster, for RSA4096 sign. But this is
+# rather because original integer-only code seems to perform
+# suboptimally on S4. Situation on Cortex-A9 is unfortunately
+# different. It's being looked into, but the trouble is that
+# performance for vectors longer than 256 bits is actually couple
+# of percent worse than for integer-only code. The code is chosen
+# for execution on all NEON-capable processors, because gain on
+# others outweighs the marginal loss on Cortex-A9.
+
+# September 2015
+#
+# Align Cortex-A9 performance with November 2013 improvements, i.e.
+# NEON code is now ~20-105% faster than integer-only one on this
+# processor. But this optimization further improved performance even
+# on other processors: NEON code path is ~45-180% faster than original
+# integer-only on Cortex-A8, ~10-210% on Cortex-A15, ~70-450% on
+# Snapdragon S4.
+
+$flavour = shift;
+if ($flavour=~/\w[\w\-]*\.\w+$/) { $output=$flavour; undef $flavour; }
+else { while (($output=shift) && ($output!~/\w[\w\-]*\.\w+$/)) {} }
+
+if ($flavour && $flavour ne "void") {
+ $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+ ( $xlate="${dir}arm-xlate.pl" and -f $xlate ) or
+ ( $xlate="${dir}../../perlasm/arm-xlate.pl" and -f $xlate) or
+ die "can't locate arm-xlate.pl";
+
+ open STDOUT,"| \"$^X\" $xlate $flavour $output";
+} else {
+ open STDOUT,">$output";
+}
+
+$num="r0"; # starts as num argument, but holds &tp[num-1]
+$ap="r1";
+$bp="r2"; $bi="r2"; $rp="r2";
+$np="r3";
+$tp="r4";
+$aj="r5";
+$nj="r6";
+$tj="r7";
+$n0="r8";
+########### # r9 is reserved by ELF as platform specific, e.g. TLS pointer
+$alo="r10"; # sl, gcc uses it to keep @GOT
+$ahi="r11"; # fp
+$nlo="r12"; # ip
+########### # r13 is stack pointer
+$nhi="r14"; # lr
+########### # r15 is program counter
+
+#### argument block layout relative to &tp[num-1], a.k.a. $num
+$_rp="$num,#12*4";
+# ap permanently resides in r1
+$_bp="$num,#13*4";
+# np permanently resides in r3
+$_n0="$num,#14*4";
+$_num="$num,#15*4"; $_bpend=$_num;
+
+$code=<<___;
+#include "arm_arch.h"
+
+.text
+#if defined(__thumb2__)
+.syntax unified
+.thumb
+#else
+.code 32
+#endif
+
+#if __ARM_MAX_ARCH__>=7
+.align 5
+.LOPENSSL_armcap:
+.word OPENSSL_armcap_P-.Lbn_mul_mont
+#endif
+
+.global bn_mul_mont
+.type bn_mul_mont,%function
+
+.align 5
+bn_mul_mont:
+.Lbn_mul_mont:
+ ldr ip,[sp,#4] @ load num
+ stmdb sp!,{r0,r2} @ sp points at argument block
+#if __ARM_MAX_ARCH__>=7
+ tst ip,#7
+ bne .Lialu
+ adr r0,.Lbn_mul_mont
+ ldr r2,.LOPENSSL_armcap
+ ldr r0,[r0,r2]
+#ifdef __APPLE__
+ ldr r0,[r0]
+#endif
+ tst r0,#ARMV7_NEON @ NEON available?
+ ldmia sp, {r0,r2}
+ beq .Lialu
+ add sp,sp,#8
+ b bn_mul8x_mont_neon
+.align 4
+.Lialu:
+#endif
+ cmp ip,#2
+ mov $num,ip @ load num
+#ifdef __thumb2__
+ ittt lt
+#endif
+ movlt r0,#0
+ addlt sp,sp,#2*4
+ blt .Labrt
+
+ stmdb sp!,{r4-r12,lr} @ save 10 registers
+
+ mov $num,$num,lsl#2 @ rescale $num for byte count
+ sub sp,sp,$num @ alloca(4*num)
+ sub sp,sp,#4 @ +extra dword
+ sub $num,$num,#4 @ "num=num-1"
+ add $tp,$bp,$num @ &bp[num-1]
+
+ add $num,sp,$num @ $num to point at &tp[num-1]
+ ldr $n0,[$_n0] @ &n0
+ ldr $bi,[$bp] @ bp[0]
+ ldr $aj,[$ap],#4 @ ap[0],ap++
+ ldr $nj,[$np],#4 @ np[0],np++
+ ldr $n0,[$n0] @ *n0
+ str $tp,[$_bpend] @ save &bp[num]
+
+ umull $alo,$ahi,$aj,$bi @ ap[0]*bp[0]
+ str $n0,[$_n0] @ save n0 value
+ mul $n0,$alo,$n0 @ "tp[0]"*n0
+ mov $nlo,#0
+ umlal $alo,$nlo,$nj,$n0 @ np[0]*n0+"t[0]"
+ mov $tp,sp
+
+.L1st:
+ ldr $aj,[$ap],#4 @ ap[j],ap++
+ mov $alo,$ahi
+ ldr $nj,[$np],#4 @ np[j],np++
+ mov $ahi,#0
+ umlal $alo,$ahi,$aj,$bi @ ap[j]*bp[0]
+ mov $nhi,#0
+ umlal $nlo,$nhi,$nj,$n0 @ np[j]*n0
+ adds $nlo,$nlo,$alo
+ str $nlo,[$tp],#4 @ tp[j-1]=,tp++
+ adc $nlo,$nhi,#0
+ cmp $tp,$num
+ bne .L1st
+
+ adds $nlo,$nlo,$ahi
+ ldr $tp,[$_bp] @ restore bp
+ mov $nhi,#0
+ ldr $n0,[$_n0] @ restore n0
+ adc $nhi,$nhi,#0
+ str $nlo,[$num] @ tp[num-1]=
+ mov $tj,sp
+ str $nhi,[$num,#4] @ tp[num]=
+
+.Louter:
+ sub $tj,$num,$tj @ "original" $num-1 value
+ sub $ap,$ap,$tj @ "rewind" ap to &ap[1]
+ ldr $bi,[$tp,#4]! @ *(++bp)
+ sub $np,$np,$tj @ "rewind" np to &np[1]
+ ldr $aj,[$ap,#-4] @ ap[0]
+ ldr $alo,[sp] @ tp[0]
+ ldr $nj,[$np,#-4] @ np[0]
+ ldr $tj,[sp,#4] @ tp[1]
+
+ mov $ahi,#0
+ umlal $alo,$ahi,$aj,$bi @ ap[0]*bp[i]+tp[0]
+ str $tp,[$_bp] @ save bp
+ mul $n0,$alo,$n0
+ mov $nlo,#0
+ umlal $alo,$nlo,$nj,$n0 @ np[0]*n0+"tp[0]"
+ mov $tp,sp
+
+.Linner:
+ ldr $aj,[$ap],#4 @ ap[j],ap++
+ adds $alo,$ahi,$tj @ +=tp[j]
+ ldr $nj,[$np],#4 @ np[j],np++
+ mov $ahi,#0
+ umlal $alo,$ahi,$aj,$bi @ ap[j]*bp[i]
+ mov $nhi,#0
+ umlal $nlo,$nhi,$nj,$n0 @ np[j]*n0
+ adc $ahi,$ahi,#0
+ ldr $tj,[$tp,#8] @ tp[j+1]
+ adds $nlo,$nlo,$alo
+ str $nlo,[$tp],#4 @ tp[j-1]=,tp++
+ adc $nlo,$nhi,#0
+ cmp $tp,$num
+ bne .Linner
+
+ adds $nlo,$nlo,$ahi
+ mov $nhi,#0
+ ldr $tp,[$_bp] @ restore bp
+ adc $nhi,$nhi,#0
+ ldr $n0,[$_n0] @ restore n0
+ adds $nlo,$nlo,$tj
+ ldr $tj,[$_bpend] @ restore &bp[num]
+ adc $nhi,$nhi,#0
+ str $nlo,[$num] @ tp[num-1]=
+ str $nhi,[$num,#4] @ tp[num]=
+
+ cmp $tp,$tj
+#ifdef __thumb2__
+ itt ne
+#endif
+ movne $tj,sp
+ bne .Louter
+
+ ldr $rp,[$_rp] @ pull rp
+ mov $aj,sp
+ add $num,$num,#4 @ $num to point at &tp[num]
+ sub $aj,$num,$aj @ "original" num value
+ mov $tp,sp @ "rewind" $tp
+ mov $ap,$tp @ "borrow" $ap
+ sub $np,$np,$aj @ "rewind" $np to &np[0]
+
+ subs $tj,$tj,$tj @ "clear" carry flag
+.Lsub: ldr $tj,[$tp],#4
+ ldr $nj,[$np],#4
+ sbcs $tj,$tj,$nj @ tp[j]-np[j]
+ str $tj,[$rp],#4 @ rp[j]=
+ teq $tp,$num @ preserve carry
+ bne .Lsub
+ sbcs $nhi,$nhi,#0 @ upmost carry
+ mov $tp,sp @ "rewind" $tp
+ sub $rp,$rp,$aj @ "rewind" $rp
+
+.Lcopy: ldr $tj,[$tp] @ conditional copy
+ ldr $aj,[$rp]
+ str sp,[$tp],#4 @ zap tp
+#ifdef __thumb2__
+ it cc
+#endif
+ movcc $aj,$tj
+ str $aj,[$rp],#4
+ teq $tp,$num @ preserve carry
+ bne .Lcopy
+
+ mov sp,$num
+ add sp,sp,#4 @ skip over tp[num+1]
+ ldmia sp!,{r4-r12,lr} @ restore registers
+ add sp,sp,#2*4 @ skip over {r0,r2}
+ mov r0,#1
+.Labrt:
+#if __ARM_ARCH__>=5
+ ret @ bx lr
+#else
+ tst lr,#1
+ moveq pc,lr @ be binary compatible with V4, yet
+ bx lr @ interoperable with Thumb ISA:-)
+#endif
+.size bn_mul_mont,.-bn_mul_mont
+___
+{
+my ($A0,$A1,$A2,$A3)=map("d$_",(0..3));
+my ($N0,$N1,$N2,$N3)=map("d$_",(4..7));
+my ($Z,$Temp)=("q4","q5");
+my @ACC=map("q$_",(6..13));
+my ($Bi,$Ni,$M0)=map("d$_",(28..31));
+my $zero="$Z#lo";
+my $temp="$Temp#lo";
+
+my ($rptr,$aptr,$bptr,$nptr,$n0,$num)=map("r$_",(0..5));
+my ($tinptr,$toutptr,$inner,$outer,$bnptr)=map("r$_",(6..11));
+
+$code.=<<___;
+#if __ARM_MAX_ARCH__>=7
+.arch armv7-a
+.fpu neon
+
+.type bn_mul8x_mont_neon,%function
+.align 5
+bn_mul8x_mont_neon:
+ mov ip,sp
+ stmdb sp!,{r4-r11}
+ vstmdb sp!,{d8-d15} @ ABI specification says so
+ ldmia ip,{r4-r5} @ load rest of parameter block
+ mov ip,sp
+
+ cmp $num,#8
+ bhi .LNEON_8n
+
+ @ special case for $num==8, everything is in register bank...
+
+ vld1.32 {${Bi}[0]}, [$bptr,:32]!
+ veor $zero,$zero,$zero
+ sub $toutptr,sp,$num,lsl#4
+ vld1.32 {$A0-$A3}, [$aptr]! @ can't specify :32 :-(
+ and $toutptr,$toutptr,#-64
+ vld1.32 {${M0}[0]}, [$n0,:32]
+ mov sp,$toutptr @ alloca
+ vzip.16 $Bi,$zero
+
+ vmull.u32 @ACC[0],$Bi,${A0}[0]
+ vmull.u32 @ACC[1],$Bi,${A0}[1]
+ vmull.u32 @ACC[2],$Bi,${A1}[0]
+ vshl.i64 $Ni,@ACC[0]#hi,#16
+ vmull.u32 @ACC[3],$Bi,${A1}[1]
+
+ vadd.u64 $Ni,$Ni,@ACC[0]#lo
+ veor $zero,$zero,$zero
+ vmul.u32 $Ni,$Ni,$M0
+
+ vmull.u32 @ACC[4],$Bi,${A2}[0]
+ vld1.32 {$N0-$N3}, [$nptr]!
+ vmull.u32 @ACC[5],$Bi,${A2}[1]
+ vmull.u32 @ACC[6],$Bi,${A3}[0]
+ vzip.16 $Ni,$zero
+ vmull.u32 @ACC[7],$Bi,${A3}[1]
+
+ vmlal.u32 @ACC[0],$Ni,${N0}[0]
+ sub $outer,$num,#1
+ vmlal.u32 @ACC[1],$Ni,${N0}[1]
+ vmlal.u32 @ACC[2],$Ni,${N1}[0]
+ vmlal.u32 @ACC[3],$Ni,${N1}[1]
+
+ vmlal.u32 @ACC[4],$Ni,${N2}[0]
+ vmov $Temp,@ACC[0]
+ vmlal.u32 @ACC[5],$Ni,${N2}[1]
+ vmov @ACC[0],@ACC[1]
+ vmlal.u32 @ACC[6],$Ni,${N3}[0]
+ vmov @ACC[1],@ACC[2]
+ vmlal.u32 @ACC[7],$Ni,${N3}[1]
+ vmov @ACC[2],@ACC[3]
+ vmov @ACC[3],@ACC[4]
+ vshr.u64 $temp,$temp,#16
+ vmov @ACC[4],@ACC[5]
+ vmov @ACC[5],@ACC[6]
+ vadd.u64 $temp,$temp,$Temp#hi
+ vmov @ACC[6],@ACC[7]
+ veor @ACC[7],@ACC[7]
+ vshr.u64 $temp,$temp,#16
+
+ b .LNEON_outer8
+
+.align 4
+.LNEON_outer8:
+ vld1.32 {${Bi}[0]}, [$bptr,:32]!
+ veor $zero,$zero,$zero
+ vzip.16 $Bi,$zero
+ vadd.u64 @ACC[0]#lo,@ACC[0]#lo,$temp
+
+ vmlal.u32 @ACC[0],$Bi,${A0}[0]
+ vmlal.u32 @ACC[1],$Bi,${A0}[1]
+ vmlal.u32 @ACC[2],$Bi,${A1}[0]
+ vshl.i64 $Ni,@ACC[0]#hi,#16
+ vmlal.u32 @ACC[3],$Bi,${A1}[1]
+
+ vadd.u64 $Ni,$Ni,@ACC[0]#lo
+ veor $zero,$zero,$zero
+ subs $outer,$outer,#1
+ vmul.u32 $Ni,$Ni,$M0
+
+ vmlal.u32 @ACC[4],$Bi,${A2}[0]
+ vmlal.u32 @ACC[5],$Bi,${A2}[1]
+ vmlal.u32 @ACC[6],$Bi,${A3}[0]
+ vzip.16 $Ni,$zero
+ vmlal.u32 @ACC[7],$Bi,${A3}[1]
+
+ vmlal.u32 @ACC[0],$Ni,${N0}[0]
+ vmlal.u32 @ACC[1],$Ni,${N0}[1]
+ vmlal.u32 @ACC[2],$Ni,${N1}[0]
+ vmlal.u32 @ACC[3],$Ni,${N1}[1]
+
+ vmlal.u32 @ACC[4],$Ni,${N2}[0]
+ vmov $Temp,@ACC[0]
+ vmlal.u32 @ACC[5],$Ni,${N2}[1]
+ vmov @ACC[0],@ACC[1]
+ vmlal.u32 @ACC[6],$Ni,${N3}[0]
+ vmov @ACC[1],@ACC[2]
+ vmlal.u32 @ACC[7],$Ni,${N3}[1]
+ vmov @ACC[2],@ACC[3]
+ vmov @ACC[3],@ACC[4]
+ vshr.u64 $temp,$temp,#16
+ vmov @ACC[4],@ACC[5]
+ vmov @ACC[5],@ACC[6]
+ vadd.u64 $temp,$temp,$Temp#hi
+ vmov @ACC[6],@ACC[7]
+ veor @ACC[7],@ACC[7]
+ vshr.u64 $temp,$temp,#16
+
+ bne .LNEON_outer8
+
+ vadd.u64 @ACC[0]#lo,@ACC[0]#lo,$temp
+ mov $toutptr,sp
+ vshr.u64 $temp,@ACC[0]#lo,#16
+ mov $inner,$num
+ vadd.u64 @ACC[0]#hi,@ACC[0]#hi,$temp
+ add $tinptr,sp,#96
+ vshr.u64 $temp,@ACC[0]#hi,#16
+ vzip.16 @ACC[0]#lo,@ACC[0]#hi
+
+ b .LNEON_tail_entry
+
+.align 4
+.LNEON_8n:
+ veor @ACC[0],@ACC[0],@ACC[0]
+ sub $toutptr,sp,#128
+ veor @ACC[1],@ACC[1],@ACC[1]
+ sub $toutptr,$toutptr,$num,lsl#4
+ veor @ACC[2],@ACC[2],@ACC[2]
+ and $toutptr,$toutptr,#-64
+ veor @ACC[3],@ACC[3],@ACC[3]
+ mov sp,$toutptr @ alloca
+ veor @ACC[4],@ACC[4],@ACC[4]
+ add $toutptr,$toutptr,#256
+ veor @ACC[5],@ACC[5],@ACC[5]
+ sub $inner,$num,#8
+ veor @ACC[6],@ACC[6],@ACC[6]
+ veor @ACC[7],@ACC[7],@ACC[7]
+
+.LNEON_8n_init:
+ vst1.64 {@ACC[0]-@ACC[1]},[$toutptr,:256]!
+ subs $inner,$inner,#8
+ vst1.64 {@ACC[2]-@ACC[3]},[$toutptr,:256]!
+ vst1.64 {@ACC[4]-@ACC[5]},[$toutptr,:256]!
+ vst1.64 {@ACC[6]-@ACC[7]},[$toutptr,:256]!
+ bne .LNEON_8n_init
+
+ add $tinptr,sp,#256
+ vld1.32 {$A0-$A3},[$aptr]!
+ add $bnptr,sp,#8
+ vld1.32 {${M0}[0]},[$n0,:32]
+ mov $outer,$num
+ b .LNEON_8n_outer
+
+.align 4
+.LNEON_8n_outer:
+ vld1.32 {${Bi}[0]},[$bptr,:32]! @ *b++
+ veor $zero,$zero,$zero
+ vzip.16 $Bi,$zero
+ add $toutptr,sp,#128
+ vld1.32 {$N0-$N3},[$nptr]!
+
+ vmlal.u32 @ACC[0],$Bi,${A0}[0]
+ vmlal.u32 @ACC[1],$Bi,${A0}[1]
+ veor $zero,$zero,$zero
+ vmlal.u32 @ACC[2],$Bi,${A1}[0]
+ vshl.i64 $Ni,@ACC[0]#hi,#16
+ vmlal.u32 @ACC[3],$Bi,${A1}[1]
+ vadd.u64 $Ni,$Ni,@ACC[0]#lo
+ vmlal.u32 @ACC[4],$Bi,${A2}[0]
+ vmul.u32 $Ni,$Ni,$M0
+ vmlal.u32 @ACC[5],$Bi,${A2}[1]
+ vst1.32 {$Bi},[sp,:64] @ put aside smashed b[8*i+0]
+ vmlal.u32 @ACC[6],$Bi,${A3}[0]
+ vzip.16 $Ni,$zero
+ vmlal.u32 @ACC[7],$Bi,${A3}[1]
+___
+for ($i=0; $i<7;) {
+$code.=<<___;
+ vld1.32 {${Bi}[0]},[$bptr,:32]! @ *b++
+ vmlal.u32 @ACC[0],$Ni,${N0}[0]
+ veor $temp,$temp,$temp
+ vmlal.u32 @ACC[1],$Ni,${N0}[1]
+ vzip.16 $Bi,$temp
+ vmlal.u32 @ACC[2],$Ni,${N1}[0]
+ vshr.u64 @ACC[0]#lo,@ACC[0]#lo,#16
+ vmlal.u32 @ACC[3],$Ni,${N1}[1]
+ vmlal.u32 @ACC[4],$Ni,${N2}[0]
+ vadd.u64 @ACC[0]#lo,@ACC[0]#lo,@ACC[0]#hi
+ vmlal.u32 @ACC[5],$Ni,${N2}[1]
+ vshr.u64 @ACC[0]#lo,@ACC[0]#lo,#16
+ vmlal.u32 @ACC[6],$Ni,${N3}[0]
+ vmlal.u32 @ACC[7],$Ni,${N3}[1]
+ vadd.u64 @ACC[1]#lo,@ACC[1]#lo,@ACC[0]#lo
+ vst1.32 {$Ni},[$bnptr,:64]! @ put aside smashed m[8*i+$i]
+___
+ push(@ACC,shift(@ACC)); $i++;
+$code.=<<___;
+ vmlal.u32 @ACC[0],$Bi,${A0}[0]
+ vld1.64 {@ACC[7]},[$tinptr,:128]!
+ vmlal.u32 @ACC[1],$Bi,${A0}[1]
+ veor $zero,$zero,$zero
+ vmlal.u32 @ACC[2],$Bi,${A1}[0]
+ vshl.i64 $Ni,@ACC[0]#hi,#16
+ vmlal.u32 @ACC[3],$Bi,${A1}[1]
+ vadd.u64 $Ni,$Ni,@ACC[0]#lo
+ vmlal.u32 @ACC[4],$Bi,${A2}[0]
+ vmul.u32 $Ni,$Ni,$M0
+ vmlal.u32 @ACC[5],$Bi,${A2}[1]
+ vst1.32 {$Bi},[$bnptr,:64]! @ put aside smashed b[8*i+$i]
+ vmlal.u32 @ACC[6],$Bi,${A3}[0]
+ vzip.16 $Ni,$zero
+ vmlal.u32 @ACC[7],$Bi,${A3}[1]
+___
+}
+$code.=<<___;
+ vld1.32 {$Bi},[sp,:64] @ pull smashed b[8*i+0]
+ vmlal.u32 @ACC[0],$Ni,${N0}[0]
+ vld1.32 {$A0-$A3},[$aptr]!
+ vmlal.u32 @ACC[1],$Ni,${N0}[1]
+ vmlal.u32 @ACC[2],$Ni,${N1}[0]
+ vshr.u64 @ACC[0]#lo,@ACC[0]#lo,#16
+ vmlal.u32 @ACC[3],$Ni,${N1}[1]
+ vmlal.u32 @ACC[4],$Ni,${N2}[0]
+ vadd.u64 @ACC[0]#lo,@ACC[0]#lo,@ACC[0]#hi
+ vmlal.u32 @ACC[5],$Ni,${N2}[1]
+ vshr.u64 @ACC[0]#lo,@ACC[0]#lo,#16
+ vmlal.u32 @ACC[6],$Ni,${N3}[0]
+ vmlal.u32 @ACC[7],$Ni,${N3}[1]
+ vadd.u64 @ACC[1]#lo,@ACC[1]#lo,@ACC[0]#lo
+ vst1.32 {$Ni},[$bnptr,:64] @ put aside smashed m[8*i+$i]
+ add $bnptr,sp,#8 @ rewind
+___
+ push(@ACC,shift(@ACC));
+$code.=<<___;
+ sub $inner,$num,#8
+ b .LNEON_8n_inner
+
+.align 4
+.LNEON_8n_inner:
+ subs $inner,$inner,#8
+ vmlal.u32 @ACC[0],$Bi,${A0}[0]
+ vld1.64 {@ACC[7]},[$tinptr,:128]
+ vmlal.u32 @ACC[1],$Bi,${A0}[1]
+ vld1.32 {$Ni},[$bnptr,:64]! @ pull smashed m[8*i+0]
+ vmlal.u32 @ACC[2],$Bi,${A1}[0]
+ vld1.32 {$N0-$N3},[$nptr]!
+ vmlal.u32 @ACC[3],$Bi,${A1}[1]
+ it ne
+ addne $tinptr,$tinptr,#16 @ don't advance in last iteration
+ vmlal.u32 @ACC[4],$Bi,${A2}[0]
+ vmlal.u32 @ACC[5],$Bi,${A2}[1]
+ vmlal.u32 @ACC[6],$Bi,${A3}[0]
+ vmlal.u32 @ACC[7],$Bi,${A3}[1]
+___
+for ($i=1; $i<8; $i++) {
+$code.=<<___;
+ vld1.32 {$Bi},[$bnptr,:64]! @ pull smashed b[8*i+$i]
+ vmlal.u32 @ACC[0],$Ni,${N0}[0]
+ vmlal.u32 @ACC[1],$Ni,${N0}[1]
+ vmlal.u32 @ACC[2],$Ni,${N1}[0]
+ vmlal.u32 @ACC[3],$Ni,${N1}[1]
+ vmlal.u32 @ACC[4],$Ni,${N2}[0]
+ vmlal.u32 @ACC[5],$Ni,${N2}[1]
+ vmlal.u32 @ACC[6],$Ni,${N3}[0]
+ vmlal.u32 @ACC[7],$Ni,${N3}[1]
+ vst1.64 {@ACC[0]},[$toutptr,:128]!
+___
+ push(@ACC,shift(@ACC));
+$code.=<<___;
+ vmlal.u32 @ACC[0],$Bi,${A0}[0]
+ vld1.64 {@ACC[7]},[$tinptr,:128]
+ vmlal.u32 @ACC[1],$Bi,${A0}[1]
+ vld1.32 {$Ni},[$bnptr,:64]! @ pull smashed m[8*i+$i]
+ vmlal.u32 @ACC[2],$Bi,${A1}[0]
+ it ne
+ addne $tinptr,$tinptr,#16 @ don't advance in last iteration
+ vmlal.u32 @ACC[3],$Bi,${A1}[1]
+ vmlal.u32 @ACC[4],$Bi,${A2}[0]
+ vmlal.u32 @ACC[5],$Bi,${A2}[1]
+ vmlal.u32 @ACC[6],$Bi,${A3}[0]
+ vmlal.u32 @ACC[7],$Bi,${A3}[1]
+___
+}
+$code.=<<___;
+ it eq
+ subeq $aptr,$aptr,$num,lsl#2 @ rewind
+ vmlal.u32 @ACC[0],$Ni,${N0}[0]
+ vld1.32 {$Bi},[sp,:64] @ pull smashed b[8*i+0]
+ vmlal.u32 @ACC[1],$Ni,${N0}[1]
+ vld1.32 {$A0-$A3},[$aptr]!
+ vmlal.u32 @ACC[2],$Ni,${N1}[0]
+ add $bnptr,sp,#8 @ rewind
+ vmlal.u32 @ACC[3],$Ni,${N1}[1]
+ vmlal.u32 @ACC[4],$Ni,${N2}[0]
+ vmlal.u32 @ACC[5],$Ni,${N2}[1]
+ vmlal.u32 @ACC[6],$Ni,${N3}[0]
+ vst1.64 {@ACC[0]},[$toutptr,:128]!
+ vmlal.u32 @ACC[7],$Ni,${N3}[1]
+
+ bne .LNEON_8n_inner
+___
+ push(@ACC,shift(@ACC));
+$code.=<<___;
+ add $tinptr,sp,#128
+ vst1.64 {@ACC[0]-@ACC[1]},[$toutptr,:256]!
+ veor q2,q2,q2 @ $N0-$N1
+ vst1.64 {@ACC[2]-@ACC[3]},[$toutptr,:256]!
+ veor q3,q3,q3 @ $N2-$N3
+ vst1.64 {@ACC[4]-@ACC[5]},[$toutptr,:256]!
+ vst1.64 {@ACC[6]},[$toutptr,:128]
+
+ subs $outer,$outer,#8
+ vld1.64 {@ACC[0]-@ACC[1]},[$tinptr,:256]!
+ vld1.64 {@ACC[2]-@ACC[3]},[$tinptr,:256]!
+ vld1.64 {@ACC[4]-@ACC[5]},[$tinptr,:256]!
+ vld1.64 {@ACC[6]-@ACC[7]},[$tinptr,:256]!
+
+ itt ne
+ subne $nptr,$nptr,$num,lsl#2 @ rewind
+ bne .LNEON_8n_outer
+
+ add $toutptr,sp,#128
+ vst1.64 {q2-q3}, [sp,:256]! @ start wiping stack frame
+ vshr.u64 $temp,@ACC[0]#lo,#16
+ vst1.64 {q2-q3},[sp,:256]!
+ vadd.u64 @ACC[0]#hi,@ACC[0]#hi,$temp
+ vst1.64 {q2-q3}, [sp,:256]!
+ vshr.u64 $temp,@ACC[0]#hi,#16
+ vst1.64 {q2-q3}, [sp,:256]!
+ vzip.16 @ACC[0]#lo,@ACC[0]#hi
+
+ mov $inner,$num
+ b .LNEON_tail_entry
+
+.align 4
+.LNEON_tail:
+ vadd.u64 @ACC[0]#lo,@ACC[0]#lo,$temp
+ vshr.u64 $temp,@ACC[0]#lo,#16
+ vld1.64 {@ACC[2]-@ACC[3]}, [$tinptr, :256]!
+ vadd.u64 @ACC[0]#hi,@ACC[0]#hi,$temp
+ vld1.64 {@ACC[4]-@ACC[5]}, [$tinptr, :256]!
+ vshr.u64 $temp,@ACC[0]#hi,#16
+ vld1.64 {@ACC[6]-@ACC[7]}, [$tinptr, :256]!
+ vzip.16 @ACC[0]#lo,@ACC[0]#hi
+
+.LNEON_tail_entry:
+___
+for ($i=1; $i<8; $i++) {
+$code.=<<___;
+ vadd.u64 @ACC[1]#lo,@ACC[1]#lo,$temp
+ vst1.32 {@ACC[0]#lo[0]}, [$toutptr, :32]!
+ vshr.u64 $temp,@ACC[1]#lo,#16
+ vadd.u64 @ACC[1]#hi,@ACC[1]#hi,$temp
+ vshr.u64 $temp,@ACC[1]#hi,#16
+ vzip.16 @ACC[1]#lo,@ACC[1]#hi
+___
+ push(@ACC,shift(@ACC));
+}
+ push(@ACC,shift(@ACC));
+$code.=<<___;
+ vld1.64 {@ACC[0]-@ACC[1]}, [$tinptr, :256]!
+ subs $inner,$inner,#8
+ vst1.32 {@ACC[7]#lo[0]}, [$toutptr, :32]!
+ bne .LNEON_tail
+
+ vst1.32 {${temp}[0]}, [$toutptr, :32] @ top-most bit
+ sub $nptr,$nptr,$num,lsl#2 @ rewind $nptr
+ subs $aptr,sp,#0 @ clear carry flag
+ add $bptr,sp,$num,lsl#2
+
+.LNEON_sub:
+ ldmia $aptr!, {r4-r7}
+ ldmia $nptr!, {r8-r11}
+ sbcs r8, r4,r8
+ sbcs r9, r5,r9
+ sbcs r10,r6,r10
+ sbcs r11,r7,r11
+ teq $aptr,$bptr @ preserves carry
+ stmia $rptr!, {r8-r11}
+ bne .LNEON_sub
+
+ ldr r10, [$aptr] @ load top-most bit
+ mov r11,sp
+ veor q0,q0,q0
+ sub r11,$bptr,r11 @ this is num*4
+ veor q1,q1,q1
+ mov $aptr,sp
+ sub $rptr,$rptr,r11 @ rewind $rptr
+ mov $nptr,$bptr @ second 3/4th of frame
+ sbcs r10,r10,#0 @ result is carry flag
+
+.LNEON_copy_n_zap:
+ ldmia $aptr!, {r4-r7}
+ ldmia $rptr, {r8-r11}
+ it cc
+ movcc r8, r4
+ vst1.64 {q0-q1}, [$nptr,:256]! @ wipe
+ itt cc
+ movcc r9, r5
+ movcc r10,r6
+ vst1.64 {q0-q1}, [$nptr,:256]! @ wipe
+ it cc
+ movcc r11,r7
+ ldmia $aptr, {r4-r7}
+ stmia $rptr!, {r8-r11}
+ sub $aptr,$aptr,#16
+ ldmia $rptr, {r8-r11}
+ it cc
+ movcc r8, r4
+ vst1.64 {q0-q1}, [$aptr,:256]! @ wipe
+ itt cc
+ movcc r9, r5
+ movcc r10,r6
+ vst1.64 {q0-q1}, [$nptr,:256]! @ wipe
+ it cc
+ movcc r11,r7
+ teq $aptr,$bptr @ preserves carry
+ stmia $rptr!, {r8-r11}
+ bne .LNEON_copy_n_zap
+
+ mov sp,ip
+ vldmia sp!,{d8-d15}
+ ldmia sp!,{r4-r11}
+ ret @ bx lr
+.size bn_mul8x_mont_neon,.-bn_mul8x_mont_neon
+#endif
+___
+}
+$code.=<<___;
+.asciz "Montgomery multiplication for ARMv4/NEON, CRYPTOGAMS by <appro\@openssl.org>"
+.align 2
+#if __ARM_MAX_ARCH__>=7
+.comm OPENSSL_armcap_P,4,4
+#endif
+___
+
+foreach (split("\n",$code)) {
+ s/\`([^\`]*)\`/eval $1/ge;
+
+ s/\bq([0-9]+)#(lo|hi)/sprintf "d%d",2*$1+($2 eq "hi")/ge or
+ s/\bret\b/bx lr/g or
+ s/\bbx\s+lr\b/.word\t0xe12fff1e/g; # make it possible to compile with -march=armv4
+
+ print $_,"\n";
+}
+
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/armv8-mont.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/armv8-mont.pl
new file mode 100755
index 0000000..bc9a18d
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/armv8-mont.pl
@@ -0,0 +1,1514 @@
+#! /usr/bin/env perl
+# Copyright 2015-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+
+# March 2015
+#
+# "Teaser" Montgomery multiplication module for ARMv8. Needs more
+# work. While it does improve RSA sign performance by 20-30% (less for
+# longer keys) on most processors, for some reason RSA2048 is not
+# faster and RSA4096 goes 15-20% slower on Cortex-A57. Multiplication
+# instruction issue rate is limited on processor in question, meaning
+# that dedicated squaring procedure is a must. Well, actually all
+# contemporary AArch64 processors seem to have limited multiplication
+# issue rate, i.e. they can't issue multiplication every cycle, which
+# explains moderate improvement coefficients in comparison to
+# compiler-generated code. Recall that compiler is instructed to use
+# umulh and therefore uses same amount of multiplication instructions
+# to do the job. Assembly's edge is to minimize number of "collateral"
+# instructions and of course instruction scheduling.
+#
+# April 2015
+#
+# Squaring procedure that handles lengths divisible by 8 improves
+# RSA/DSA performance by 25-40-60% depending on processor and key
+# length. Overall improvement coefficients are always positive in
+# comparison to compiler-generated code. On Cortex-A57 improvement
+# is still modest on longest key lengths, while others exhibit e.g.
+# 50-70% improvement for RSA4096 sign. RSA2048 sign is ~25% faster
+# on Cortex-A57 and ~60-100% faster on others.
+
+$flavour = shift;
+$output = shift;
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+( $xlate="${dir}arm-xlate.pl" and -f $xlate ) or
+( $xlate="${dir}../../perlasm/arm-xlate.pl" and -f $xlate) or
+die "can't locate arm-xlate.pl";
+
+open OUT,"| \"$^X\" $xlate $flavour $output";
+*STDOUT=*OUT;
+
+($lo0,$hi0,$aj,$m0,$alo,$ahi,
+ $lo1,$hi1,$nj,$m1,$nlo,$nhi,
+ $ovf, $i,$j,$tp,$tj) = map("x$_",6..17,19..24);
+
+# int bn_mul_mont(
+$rp="x0"; # BN_ULONG *rp,
+$ap="x1"; # const BN_ULONG *ap,
+$bp="x2"; # const BN_ULONG *bp,
+$np="x3"; # const BN_ULONG *np,
+$n0="x4"; # const BN_ULONG *n0,
+$num="x5"; # int num);
+
+$code.=<<___;
+.text
+
+.globl bn_mul_mont
+.type bn_mul_mont,%function
+.align 5
+bn_mul_mont:
+ tst $num,#7
+ b.eq __bn_sqr8x_mont
+ tst $num,#3
+ b.eq __bn_mul4x_mont
+.Lmul_mont:
+ stp x29,x30,[sp,#-64]!
+ add x29,sp,#0
+ stp x19,x20,[sp,#16]
+ stp x21,x22,[sp,#32]
+ stp x23,x24,[sp,#48]
+
+ ldr $m0,[$bp],#8 // bp[0]
+ sub $tp,sp,$num,lsl#3
+ ldp $hi0,$aj,[$ap],#16 // ap[0..1]
+ lsl $num,$num,#3
+ ldr $n0,[$n0] // *n0
+ and $tp,$tp,#-16 // ABI says so
+ ldp $hi1,$nj,[$np],#16 // np[0..1]
+
+ mul $lo0,$hi0,$m0 // ap[0]*bp[0]
+ sub $j,$num,#16 // j=num-2
+ umulh $hi0,$hi0,$m0
+ mul $alo,$aj,$m0 // ap[1]*bp[0]
+ umulh $ahi,$aj,$m0
+
+ mul $m1,$lo0,$n0 // "tp[0]"*n0
+ mov sp,$tp // alloca
+
+ // (*) mul $lo1,$hi1,$m1 // np[0]*m1
+ umulh $hi1,$hi1,$m1
+ mul $nlo,$nj,$m1 // np[1]*m1
+ // (*) adds $lo1,$lo1,$lo0 // discarded
+ // (*) As for removal of first multiplication and addition
+ // instructions. The outcome of first addition is
+ // guaranteed to be zero, which leaves two computationally
+ // significant outcomes: it either carries or not. Then
+ // question is when does it carry? Is there alternative
+ // way to deduce it? If you follow operations, you can
+ // observe that condition for carry is quite simple:
+ // $lo0 being non-zero. So that carry can be calculated
+ // by adding -1 to $lo0. That's what next instruction does.
+ subs xzr,$lo0,#1 // (*)
+ umulh $nhi,$nj,$m1
+ adc $hi1,$hi1,xzr
+ cbz $j,.L1st_skip
+
+.L1st:
+ ldr $aj,[$ap],#8
+ adds $lo0,$alo,$hi0
+ sub $j,$j,#8 // j--
+ adc $hi0,$ahi,xzr
+
+ ldr $nj,[$np],#8
+ adds $lo1,$nlo,$hi1
+ mul $alo,$aj,$m0 // ap[j]*bp[0]
+ adc $hi1,$nhi,xzr
+ umulh $ahi,$aj,$m0
+
+ adds $lo1,$lo1,$lo0
+ mul $nlo,$nj,$m1 // np[j]*m1
+ adc $hi1,$hi1,xzr
+ umulh $nhi,$nj,$m1
+ str $lo1,[$tp],#8 // tp[j-1]
+ cbnz $j,.L1st
+
+.L1st_skip:
+ adds $lo0,$alo,$hi0
+ sub $ap,$ap,$num // rewind $ap
+ adc $hi0,$ahi,xzr
+
+ adds $lo1,$nlo,$hi1
+ sub $np,$np,$num // rewind $np
+ adc $hi1,$nhi,xzr
+
+ adds $lo1,$lo1,$lo0
+ sub $i,$num,#8 // i=num-1
+ adcs $hi1,$hi1,$hi0
+
+ adc $ovf,xzr,xzr // upmost overflow bit
+ stp $lo1,$hi1,[$tp]
+
+.Louter:
+ ldr $m0,[$bp],#8 // bp[i]
+ ldp $hi0,$aj,[$ap],#16
+ ldr $tj,[sp] // tp[0]
+ add $tp,sp,#8
+
+ mul $lo0,$hi0,$m0 // ap[0]*bp[i]
+ sub $j,$num,#16 // j=num-2
+ umulh $hi0,$hi0,$m0
+ ldp $hi1,$nj,[$np],#16
+ mul $alo,$aj,$m0 // ap[1]*bp[i]
+ adds $lo0,$lo0,$tj
+ umulh $ahi,$aj,$m0
+ adc $hi0,$hi0,xzr
+
+ mul $m1,$lo0,$n0
+ sub $i,$i,#8 // i--
+
+ // (*) mul $lo1,$hi1,$m1 // np[0]*m1
+ umulh $hi1,$hi1,$m1
+ mul $nlo,$nj,$m1 // np[1]*m1
+ // (*) adds $lo1,$lo1,$lo0
+ subs xzr,$lo0,#1 // (*)
+ umulh $nhi,$nj,$m1
+ cbz $j,.Linner_skip
+
+.Linner:
+ ldr $aj,[$ap],#8
+ adc $hi1,$hi1,xzr
+ ldr $tj,[$tp],#8 // tp[j]
+ adds $lo0,$alo,$hi0
+ sub $j,$j,#8 // j--
+ adc $hi0,$ahi,xzr
+
+ adds $lo1,$nlo,$hi1
+ ldr $nj,[$np],#8
+ adc $hi1,$nhi,xzr
+
+ mul $alo,$aj,$m0 // ap[j]*bp[i]
+ adds $lo0,$lo0,$tj
+ umulh $ahi,$aj,$m0
+ adc $hi0,$hi0,xzr
+
+ mul $nlo,$nj,$m1 // np[j]*m1
+ adds $lo1,$lo1,$lo0
+ umulh $nhi,$nj,$m1
+ str $lo1,[$tp,#-16] // tp[j-1]
+ cbnz $j,.Linner
+
+.Linner_skip:
+ ldr $tj,[$tp],#8 // tp[j]
+ adc $hi1,$hi1,xzr
+ adds $lo0,$alo,$hi0
+ sub $ap,$ap,$num // rewind $ap
+ adc $hi0,$ahi,xzr
+
+ adds $lo1,$nlo,$hi1
+ sub $np,$np,$num // rewind $np
+ adcs $hi1,$nhi,$ovf
+ adc $ovf,xzr,xzr
+
+ adds $lo0,$lo0,$tj
+ adc $hi0,$hi0,xzr
+
+ adds $lo1,$lo1,$lo0
+ adcs $hi1,$hi1,$hi0
+ adc $ovf,$ovf,xzr // upmost overflow bit
+ stp $lo1,$hi1,[$tp,#-16]
+
+ cbnz $i,.Louter
+
+ // Final step. We see if result is larger than modulus, and
+ // if it is, subtract the modulus. But comparison implies
+ // subtraction. So we subtract modulus, see if it borrowed,
+ // and conditionally copy original value.
+ ldr $tj,[sp] // tp[0]
+ add $tp,sp,#8
+ ldr $nj,[$np],#8 // np[0]
+ subs $j,$num,#8 // j=num-1 and clear borrow
+ mov $ap,$rp
+.Lsub:
+ sbcs $aj,$tj,$nj // tp[j]-np[j]
+ ldr $tj,[$tp],#8
+ sub $j,$j,#8 // j--
+ ldr $nj,[$np],#8
+ str $aj,[$ap],#8 // rp[j]=tp[j]-np[j]
+ cbnz $j,.Lsub
+
+ sbcs $aj,$tj,$nj
+ sbcs $ovf,$ovf,xzr // did it borrow?
+ str $aj,[$ap],#8 // rp[num-1]
+
+ ldr $tj,[sp] // tp[0]
+ add $tp,sp,#8
+ ldr $aj,[$rp],#8 // rp[0]
+ sub $num,$num,#8 // num--
+ nop
+.Lcond_copy:
+ sub $num,$num,#8 // num--
+ csel $nj,$tj,$aj,lo // did it borrow?
+ ldr $tj,[$tp],#8
+ ldr $aj,[$rp],#8
+ str xzr,[$tp,#-16] // wipe tp
+ str $nj,[$rp,#-16]
+ cbnz $num,.Lcond_copy
+
+ csel $nj,$tj,$aj,lo
+ str xzr,[$tp,#-8] // wipe tp
+ str $nj,[$rp,#-8]
+
+ ldp x19,x20,[x29,#16]
+ mov sp,x29
+ ldp x21,x22,[x29,#32]
+ mov x0,#1
+ ldp x23,x24,[x29,#48]
+ ldr x29,[sp],#64
+ ret
+.size bn_mul_mont,.-bn_mul_mont
+___
+{
+########################################################################
+# Following is ARMv8 adaptation of sqrx8x_mont from x86_64-mont5 module.
+
+my ($a0,$a1,$a2,$a3,$a4,$a5,$a6,$a7)=map("x$_",(6..13));
+my ($t0,$t1,$t2,$t3)=map("x$_",(14..17));
+my ($acc0,$acc1,$acc2,$acc3,$acc4,$acc5,$acc6,$acc7)=map("x$_",(19..26));
+my ($cnt,$carry,$topmost)=("x27","x28","x30");
+my ($tp,$ap_end,$na0)=($bp,$np,$carry);
+
+$code.=<<___;
+.type __bn_sqr8x_mont,%function
+.align 5
+__bn_sqr8x_mont:
+ cmp $ap,$bp
+ b.ne __bn_mul4x_mont
+.Lsqr8x_mont:
+ .inst 0xd503233f // paciasp
+ stp x29,x30,[sp,#-128]!
+ add x29,sp,#0
+ stp x19,x20,[sp,#16]
+ stp x21,x22,[sp,#32]
+ stp x23,x24,[sp,#48]
+ stp x25,x26,[sp,#64]
+ stp x27,x28,[sp,#80]
+ stp $rp,$np,[sp,#96] // offload rp and np
+
+ ldp $a0,$a1,[$ap,#8*0]
+ ldp $a2,$a3,[$ap,#8*2]
+ ldp $a4,$a5,[$ap,#8*4]
+ ldp $a6,$a7,[$ap,#8*6]
+
+ sub $tp,sp,$num,lsl#4
+ lsl $num,$num,#3
+ ldr $n0,[$n0] // *n0
+ mov sp,$tp // alloca
+ sub $cnt,$num,#8*8
+ b .Lsqr8x_zero_start
+
+.Lsqr8x_zero:
+ sub $cnt,$cnt,#8*8
+ stp xzr,xzr,[$tp,#8*0]
+ stp xzr,xzr,[$tp,#8*2]
+ stp xzr,xzr,[$tp,#8*4]
+ stp xzr,xzr,[$tp,#8*6]
+.Lsqr8x_zero_start:
+ stp xzr,xzr,[$tp,#8*8]
+ stp xzr,xzr,[$tp,#8*10]
+ stp xzr,xzr,[$tp,#8*12]
+ stp xzr,xzr,[$tp,#8*14]
+ add $tp,$tp,#8*16
+ cbnz $cnt,.Lsqr8x_zero
+
+ add $ap_end,$ap,$num
+ add $ap,$ap,#8*8
+ mov $acc0,xzr
+ mov $acc1,xzr
+ mov $acc2,xzr
+ mov $acc3,xzr
+ mov $acc4,xzr
+ mov $acc5,xzr
+ mov $acc6,xzr
+ mov $acc7,xzr
+ mov $tp,sp
+ str $n0,[x29,#112] // offload n0
+
+ // Multiply everything but a[i]*a[i]
+.align 4
+.Lsqr8x_outer_loop:
+ // a[1]a[0] (i)
+ // a[2]a[0]
+ // a[3]a[0]
+ // a[4]a[0]
+ // a[5]a[0]
+ // a[6]a[0]
+ // a[7]a[0]
+ // a[2]a[1] (ii)
+ // a[3]a[1]
+ // a[4]a[1]
+ // a[5]a[1]
+ // a[6]a[1]
+ // a[7]a[1]
+ // a[3]a[2] (iii)
+ // a[4]a[2]
+ // a[5]a[2]
+ // a[6]a[2]
+ // a[7]a[2]
+ // a[4]a[3] (iv)
+ // a[5]a[3]
+ // a[6]a[3]
+ // a[7]a[3]
+ // a[5]a[4] (v)
+ // a[6]a[4]
+ // a[7]a[4]
+ // a[6]a[5] (vi)
+ // a[7]a[5]
+ // a[7]a[6] (vii)
+
+ mul $t0,$a1,$a0 // lo(a[1..7]*a[0]) (i)
+ mul $t1,$a2,$a0
+ mul $t2,$a3,$a0
+ mul $t3,$a4,$a0
+ adds $acc1,$acc1,$t0 // t[1]+lo(a[1]*a[0])
+ mul $t0,$a5,$a0
+ adcs $acc2,$acc2,$t1
+ mul $t1,$a6,$a0
+ adcs $acc3,$acc3,$t2
+ mul $t2,$a7,$a0
+ adcs $acc4,$acc4,$t3
+ umulh $t3,$a1,$a0 // hi(a[1..7]*a[0])
+ adcs $acc5,$acc5,$t0
+ umulh $t0,$a2,$a0
+ adcs $acc6,$acc6,$t1
+ umulh $t1,$a3,$a0
+ adcs $acc7,$acc7,$t2
+ umulh $t2,$a4,$a0
+ stp $acc0,$acc1,[$tp],#8*2 // t[0..1]
+ adc $acc0,xzr,xzr // t[8]
+ adds $acc2,$acc2,$t3 // t[2]+lo(a[1]*a[0])
+ umulh $t3,$a5,$a0
+ adcs $acc3,$acc3,$t0
+ umulh $t0,$a6,$a0
+ adcs $acc4,$acc4,$t1
+ umulh $t1,$a7,$a0
+ adcs $acc5,$acc5,$t2
+ mul $t2,$a2,$a1 // lo(a[2..7]*a[1]) (ii)
+ adcs $acc6,$acc6,$t3
+ mul $t3,$a3,$a1
+ adcs $acc7,$acc7,$t0
+ mul $t0,$a4,$a1
+ adc $acc0,$acc0,$t1
+
+ mul $t1,$a5,$a1
+ adds $acc3,$acc3,$t2
+ mul $t2,$a6,$a1
+ adcs $acc4,$acc4,$t3
+ mul $t3,$a7,$a1
+ adcs $acc5,$acc5,$t0
+ umulh $t0,$a2,$a1 // hi(a[2..7]*a[1])
+ adcs $acc6,$acc6,$t1
+ umulh $t1,$a3,$a1
+ adcs $acc7,$acc7,$t2
+ umulh $t2,$a4,$a1
+ adcs $acc0,$acc0,$t3
+ umulh $t3,$a5,$a1
+ stp $acc2,$acc3,[$tp],#8*2 // t[2..3]
+ adc $acc1,xzr,xzr // t[9]
+ adds $acc4,$acc4,$t0
+ umulh $t0,$a6,$a1
+ adcs $acc5,$acc5,$t1
+ umulh $t1,$a7,$a1
+ adcs $acc6,$acc6,$t2
+ mul $t2,$a3,$a2 // lo(a[3..7]*a[2]) (iii)
+ adcs $acc7,$acc7,$t3
+ mul $t3,$a4,$a2
+ adcs $acc0,$acc0,$t0
+ mul $t0,$a5,$a2
+ adc $acc1,$acc1,$t1
+
+ mul $t1,$a6,$a2
+ adds $acc5,$acc5,$t2
+ mul $t2,$a7,$a2
+ adcs $acc6,$acc6,$t3
+ umulh $t3,$a3,$a2 // hi(a[3..7]*a[2])
+ adcs $acc7,$acc7,$t0
+ umulh $t0,$a4,$a2
+ adcs $acc0,$acc0,$t1
+ umulh $t1,$a5,$a2
+ adcs $acc1,$acc1,$t2
+ umulh $t2,$a6,$a2
+ stp $acc4,$acc5,[$tp],#8*2 // t[4..5]
+ adc $acc2,xzr,xzr // t[10]
+ adds $acc6,$acc6,$t3
+ umulh $t3,$a7,$a2
+ adcs $acc7,$acc7,$t0
+ mul $t0,$a4,$a3 // lo(a[4..7]*a[3]) (iv)
+ adcs $acc0,$acc0,$t1
+ mul $t1,$a5,$a3
+ adcs $acc1,$acc1,$t2
+ mul $t2,$a6,$a3
+ adc $acc2,$acc2,$t3
+
+ mul $t3,$a7,$a3
+ adds $acc7,$acc7,$t0
+ umulh $t0,$a4,$a3 // hi(a[4..7]*a[3])
+ adcs $acc0,$acc0,$t1
+ umulh $t1,$a5,$a3
+ adcs $acc1,$acc1,$t2
+ umulh $t2,$a6,$a3
+ adcs $acc2,$acc2,$t3
+ umulh $t3,$a7,$a3
+ stp $acc6,$acc7,[$tp],#8*2 // t[6..7]
+ adc $acc3,xzr,xzr // t[11]
+ adds $acc0,$acc0,$t0
+ mul $t0,$a5,$a4 // lo(a[5..7]*a[4]) (v)
+ adcs $acc1,$acc1,$t1
+ mul $t1,$a6,$a4
+ adcs $acc2,$acc2,$t2
+ mul $t2,$a7,$a4
+ adc $acc3,$acc3,$t3
+
+ umulh $t3,$a5,$a4 // hi(a[5..7]*a[4])
+ adds $acc1,$acc1,$t0
+ umulh $t0,$a6,$a4
+ adcs $acc2,$acc2,$t1
+ umulh $t1,$a7,$a4
+ adcs $acc3,$acc3,$t2
+ mul $t2,$a6,$a5 // lo(a[6..7]*a[5]) (vi)
+ adc $acc4,xzr,xzr // t[12]
+ adds $acc2,$acc2,$t3
+ mul $t3,$a7,$a5
+ adcs $acc3,$acc3,$t0
+ umulh $t0,$a6,$a5 // hi(a[6..7]*a[5])
+ adc $acc4,$acc4,$t1
+
+ umulh $t1,$a7,$a5
+ adds $acc3,$acc3,$t2
+ mul $t2,$a7,$a6 // lo(a[7]*a[6]) (vii)
+ adcs $acc4,$acc4,$t3
+ umulh $t3,$a7,$a6 // hi(a[7]*a[6])
+ adc $acc5,xzr,xzr // t[13]
+ adds $acc4,$acc4,$t0
+ sub $cnt,$ap_end,$ap // done yet?
+ adc $acc5,$acc5,$t1
+
+ adds $acc5,$acc5,$t2
+ sub $t0,$ap_end,$num // rewinded ap
+ adc $acc6,xzr,xzr // t[14]
+ add $acc6,$acc6,$t3
+
+ cbz $cnt,.Lsqr8x_outer_break
+
+ mov $n0,$a0
+ ldp $a0,$a1,[$tp,#8*0]
+ ldp $a2,$a3,[$tp,#8*2]
+ ldp $a4,$a5,[$tp,#8*4]
+ ldp $a6,$a7,[$tp,#8*6]
+ adds $acc0,$acc0,$a0
+ adcs $acc1,$acc1,$a1
+ ldp $a0,$a1,[$ap,#8*0]
+ adcs $acc2,$acc2,$a2
+ adcs $acc3,$acc3,$a3
+ ldp $a2,$a3,[$ap,#8*2]
+ adcs $acc4,$acc4,$a4
+ adcs $acc5,$acc5,$a5
+ ldp $a4,$a5,[$ap,#8*4]
+ adcs $acc6,$acc6,$a6
+ mov $rp,$ap
+ adcs $acc7,xzr,$a7
+ ldp $a6,$a7,[$ap,#8*6]
+ add $ap,$ap,#8*8
+ //adc $carry,xzr,xzr // moved below
+ mov $cnt,#-8*8
+
+ // a[8]a[0]
+ // a[9]a[0]
+ // a[a]a[0]
+ // a[b]a[0]
+ // a[c]a[0]
+ // a[d]a[0]
+ // a[e]a[0]
+ // a[f]a[0]
+ // a[8]a[1]
+ // a[f]a[1]........................
+ // a[8]a[2]
+ // a[f]a[2]........................
+ // a[8]a[3]
+ // a[f]a[3]........................
+ // a[8]a[4]
+ // a[f]a[4]........................
+ // a[8]a[5]
+ // a[f]a[5]........................
+ // a[8]a[6]
+ // a[f]a[6]........................
+ // a[8]a[7]
+ // a[f]a[7]........................
+.Lsqr8x_mul:
+ mul $t0,$a0,$n0
+ adc $carry,xzr,xzr // carry bit, modulo-scheduled
+ mul $t1,$a1,$n0
+ add $cnt,$cnt,#8
+ mul $t2,$a2,$n0
+ mul $t3,$a3,$n0
+ adds $acc0,$acc0,$t0
+ mul $t0,$a4,$n0
+ adcs $acc1,$acc1,$t1
+ mul $t1,$a5,$n0
+ adcs $acc2,$acc2,$t2
+ mul $t2,$a6,$n0
+ adcs $acc3,$acc3,$t3
+ mul $t3,$a7,$n0
+ adcs $acc4,$acc4,$t0
+ umulh $t0,$a0,$n0
+ adcs $acc5,$acc5,$t1
+ umulh $t1,$a1,$n0
+ adcs $acc6,$acc6,$t2
+ umulh $t2,$a2,$n0
+ adcs $acc7,$acc7,$t3
+ umulh $t3,$a3,$n0
+ adc $carry,$carry,xzr
+ str $acc0,[$tp],#8
+ adds $acc0,$acc1,$t0
+ umulh $t0,$a4,$n0
+ adcs $acc1,$acc2,$t1
+ umulh $t1,$a5,$n0
+ adcs $acc2,$acc3,$t2
+ umulh $t2,$a6,$n0
+ adcs $acc3,$acc4,$t3
+ umulh $t3,$a7,$n0
+ ldr $n0,[$rp,$cnt]
+ adcs $acc4,$acc5,$t0
+ adcs $acc5,$acc6,$t1
+ adcs $acc6,$acc7,$t2
+ adcs $acc7,$carry,$t3
+ //adc $carry,xzr,xzr // moved above
+ cbnz $cnt,.Lsqr8x_mul
+ // note that carry flag is guaranteed
+ // to be zero at this point
+ cmp $ap,$ap_end // done yet?
+ b.eq .Lsqr8x_break
+
+ ldp $a0,$a1,[$tp,#8*0]
+ ldp $a2,$a3,[$tp,#8*2]
+ ldp $a4,$a5,[$tp,#8*4]
+ ldp $a6,$a7,[$tp,#8*6]
+ adds $acc0,$acc0,$a0
+ ldr $n0,[$rp,#-8*8]
+ adcs $acc1,$acc1,$a1
+ ldp $a0,$a1,[$ap,#8*0]
+ adcs $acc2,$acc2,$a2
+ adcs $acc3,$acc3,$a3
+ ldp $a2,$a3,[$ap,#8*2]
+ adcs $acc4,$acc4,$a4
+ adcs $acc5,$acc5,$a5
+ ldp $a4,$a5,[$ap,#8*4]
+ adcs $acc6,$acc6,$a6
+ mov $cnt,#-8*8
+ adcs $acc7,$acc7,$a7
+ ldp $a6,$a7,[$ap,#8*6]
+ add $ap,$ap,#8*8
+ //adc $carry,xzr,xzr // moved above
+ b .Lsqr8x_mul
+
+.align 4
+.Lsqr8x_break:
+ ldp $a0,$a1,[$rp,#8*0]
+ add $ap,$rp,#8*8
+ ldp $a2,$a3,[$rp,#8*2]
+ sub $t0,$ap_end,$ap // is it last iteration?
+ ldp $a4,$a5,[$rp,#8*4]
+ sub $t1,$tp,$t0
+ ldp $a6,$a7,[$rp,#8*6]
+ cbz $t0,.Lsqr8x_outer_loop
+
+ stp $acc0,$acc1,[$tp,#8*0]
+ ldp $acc0,$acc1,[$t1,#8*0]
+ stp $acc2,$acc3,[$tp,#8*2]
+ ldp $acc2,$acc3,[$t1,#8*2]
+ stp $acc4,$acc5,[$tp,#8*4]
+ ldp $acc4,$acc5,[$t1,#8*4]
+ stp $acc6,$acc7,[$tp,#8*6]
+ mov $tp,$t1
+ ldp $acc6,$acc7,[$t1,#8*6]
+ b .Lsqr8x_outer_loop
+
+.align 4
+.Lsqr8x_outer_break:
+ // Now multiply above result by 2 and add a[n-1]*a[n-1]|...|a[0]*a[0]
+ ldp $a1,$a3,[$t0,#8*0] // recall that $t0 is &a[0]
+ ldp $t1,$t2,[sp,#8*1]
+ ldp $a5,$a7,[$t0,#8*2]
+ add $ap,$t0,#8*4
+ ldp $t3,$t0,[sp,#8*3]
+
+ stp $acc0,$acc1,[$tp,#8*0]
+ mul $acc0,$a1,$a1
+ stp $acc2,$acc3,[$tp,#8*2]
+ umulh $a1,$a1,$a1
+ stp $acc4,$acc5,[$tp,#8*4]
+ mul $a2,$a3,$a3
+ stp $acc6,$acc7,[$tp,#8*6]
+ mov $tp,sp
+ umulh $a3,$a3,$a3
+ adds $acc1,$a1,$t1,lsl#1
+ extr $t1,$t2,$t1,#63
+ sub $cnt,$num,#8*4
+
+.Lsqr4x_shift_n_add:
+ adcs $acc2,$a2,$t1
+ extr $t2,$t3,$t2,#63
+ sub $cnt,$cnt,#8*4
+ adcs $acc3,$a3,$t2
+ ldp $t1,$t2,[$tp,#8*5]
+ mul $a4,$a5,$a5
+ ldp $a1,$a3,[$ap],#8*2
+ umulh $a5,$a5,$a5
+ mul $a6,$a7,$a7
+ umulh $a7,$a7,$a7
+ extr $t3,$t0,$t3,#63
+ stp $acc0,$acc1,[$tp,#8*0]
+ adcs $acc4,$a4,$t3
+ extr $t0,$t1,$t0,#63
+ stp $acc2,$acc3,[$tp,#8*2]
+ adcs $acc5,$a5,$t0
+ ldp $t3,$t0,[$tp,#8*7]
+ extr $t1,$t2,$t1,#63
+ adcs $acc6,$a6,$t1
+ extr $t2,$t3,$t2,#63
+ adcs $acc7,$a7,$t2
+ ldp $t1,$t2,[$tp,#8*9]
+ mul $a0,$a1,$a1
+ ldp $a5,$a7,[$ap],#8*2
+ umulh $a1,$a1,$a1
+ mul $a2,$a3,$a3
+ umulh $a3,$a3,$a3
+ stp $acc4,$acc5,[$tp,#8*4]
+ extr $t3,$t0,$t3,#63
+ stp $acc6,$acc7,[$tp,#8*6]
+ add $tp,$tp,#8*8
+ adcs $acc0,$a0,$t3
+ extr $t0,$t1,$t0,#63
+ adcs $acc1,$a1,$t0
+ ldp $t3,$t0,[$tp,#8*3]
+ extr $t1,$t2,$t1,#63
+ cbnz $cnt,.Lsqr4x_shift_n_add
+___
+my ($np,$np_end)=($ap,$ap_end);
+$code.=<<___;
+ ldp $np,$n0,[x29,#104] // pull np and n0
+
+ adcs $acc2,$a2,$t1
+ extr $t2,$t3,$t2,#63
+ adcs $acc3,$a3,$t2
+ ldp $t1,$t2,[$tp,#8*5]
+ mul $a4,$a5,$a5
+ umulh $a5,$a5,$a5
+ stp $acc0,$acc1,[$tp,#8*0]
+ mul $a6,$a7,$a7
+ umulh $a7,$a7,$a7
+ stp $acc2,$acc3,[$tp,#8*2]
+ extr $t3,$t0,$t3,#63
+ adcs $acc4,$a4,$t3
+ extr $t0,$t1,$t0,#63
+ ldp $acc0,$acc1,[sp,#8*0]
+ adcs $acc5,$a5,$t0
+ extr $t1,$t2,$t1,#63
+ ldp $a0,$a1,[$np,#8*0]
+ adcs $acc6,$a6,$t1
+ extr $t2,xzr,$t2,#63
+ ldp $a2,$a3,[$np,#8*2]
+ adc $acc7,$a7,$t2
+ ldp $a4,$a5,[$np,#8*4]
+
+ // Reduce by 512 bits per iteration
+ mul $na0,$n0,$acc0 // t[0]*n0
+ ldp $a6,$a7,[$np,#8*6]
+ add $np_end,$np,$num
+ ldp $acc2,$acc3,[sp,#8*2]
+ stp $acc4,$acc5,[$tp,#8*4]
+ ldp $acc4,$acc5,[sp,#8*4]
+ stp $acc6,$acc7,[$tp,#8*6]
+ ldp $acc6,$acc7,[sp,#8*6]
+ add $np,$np,#8*8
+ mov $topmost,xzr // initial top-most carry
+ mov $tp,sp
+ mov $cnt,#8
+
+.Lsqr8x_reduction:
+ // (*) mul $t0,$a0,$na0 // lo(n[0-7])*lo(t[0]*n0)
+ mul $t1,$a1,$na0
+ sub $cnt,$cnt,#1
+ mul $t2,$a2,$na0
+ str $na0,[$tp],#8 // put aside t[0]*n0 for tail processing
+ mul $t3,$a3,$na0
+ // (*) adds xzr,$acc0,$t0
+ subs xzr,$acc0,#1 // (*)
+ mul $t0,$a4,$na0
+ adcs $acc0,$acc1,$t1
+ mul $t1,$a5,$na0
+ adcs $acc1,$acc2,$t2
+ mul $t2,$a6,$na0
+ adcs $acc2,$acc3,$t3
+ mul $t3,$a7,$na0
+ adcs $acc3,$acc4,$t0
+ umulh $t0,$a0,$na0 // hi(n[0-7])*lo(t[0]*n0)
+ adcs $acc4,$acc5,$t1
+ umulh $t1,$a1,$na0
+ adcs $acc5,$acc6,$t2
+ umulh $t2,$a2,$na0
+ adcs $acc6,$acc7,$t3
+ umulh $t3,$a3,$na0
+ adc $acc7,xzr,xzr
+ adds $acc0,$acc0,$t0
+ umulh $t0,$a4,$na0
+ adcs $acc1,$acc1,$t1
+ umulh $t1,$a5,$na0
+ adcs $acc2,$acc2,$t2
+ umulh $t2,$a6,$na0
+ adcs $acc3,$acc3,$t3
+ umulh $t3,$a7,$na0
+ mul $na0,$n0,$acc0 // next t[0]*n0
+ adcs $acc4,$acc4,$t0
+ adcs $acc5,$acc5,$t1
+ adcs $acc6,$acc6,$t2
+ adc $acc7,$acc7,$t3
+ cbnz $cnt,.Lsqr8x_reduction
+
+ ldp $t0,$t1,[$tp,#8*0]
+ ldp $t2,$t3,[$tp,#8*2]
+ mov $rp,$tp
+ sub $cnt,$np_end,$np // done yet?
+ adds $acc0,$acc0,$t0
+ adcs $acc1,$acc1,$t1
+ ldp $t0,$t1,[$tp,#8*4]
+ adcs $acc2,$acc2,$t2
+ adcs $acc3,$acc3,$t3
+ ldp $t2,$t3,[$tp,#8*6]
+ adcs $acc4,$acc4,$t0
+ adcs $acc5,$acc5,$t1
+ adcs $acc6,$acc6,$t2
+ adcs $acc7,$acc7,$t3
+ //adc $carry,xzr,xzr // moved below
+ cbz $cnt,.Lsqr8x8_post_condition
+
+ ldr $n0,[$tp,#-8*8]
+ ldp $a0,$a1,[$np,#8*0]
+ ldp $a2,$a3,[$np,#8*2]
+ ldp $a4,$a5,[$np,#8*4]
+ mov $cnt,#-8*8
+ ldp $a6,$a7,[$np,#8*6]
+ add $np,$np,#8*8
+
+.Lsqr8x_tail:
+ mul $t0,$a0,$n0
+ adc $carry,xzr,xzr // carry bit, modulo-scheduled
+ mul $t1,$a1,$n0
+ add $cnt,$cnt,#8
+ mul $t2,$a2,$n0
+ mul $t3,$a3,$n0
+ adds $acc0,$acc0,$t0
+ mul $t0,$a4,$n0
+ adcs $acc1,$acc1,$t1
+ mul $t1,$a5,$n0
+ adcs $acc2,$acc2,$t2
+ mul $t2,$a6,$n0
+ adcs $acc3,$acc3,$t3
+ mul $t3,$a7,$n0
+ adcs $acc4,$acc4,$t0
+ umulh $t0,$a0,$n0
+ adcs $acc5,$acc5,$t1
+ umulh $t1,$a1,$n0
+ adcs $acc6,$acc6,$t2
+ umulh $t2,$a2,$n0
+ adcs $acc7,$acc7,$t3
+ umulh $t3,$a3,$n0
+ adc $carry,$carry,xzr
+ str $acc0,[$tp],#8
+ adds $acc0,$acc1,$t0
+ umulh $t0,$a4,$n0
+ adcs $acc1,$acc2,$t1
+ umulh $t1,$a5,$n0
+ adcs $acc2,$acc3,$t2
+ umulh $t2,$a6,$n0
+ adcs $acc3,$acc4,$t3
+ umulh $t3,$a7,$n0
+ ldr $n0,[$rp,$cnt]
+ adcs $acc4,$acc5,$t0
+ adcs $acc5,$acc6,$t1
+ adcs $acc6,$acc7,$t2
+ adcs $acc7,$carry,$t3
+ //adc $carry,xzr,xzr // moved above
+ cbnz $cnt,.Lsqr8x_tail
+ // note that carry flag is guaranteed
+ // to be zero at this point
+ ldp $a0,$a1,[$tp,#8*0]
+ sub $cnt,$np_end,$np // done yet?
+ sub $t2,$np_end,$num // rewinded np
+ ldp $a2,$a3,[$tp,#8*2]
+ ldp $a4,$a5,[$tp,#8*4]
+ ldp $a6,$a7,[$tp,#8*6]
+ cbz $cnt,.Lsqr8x_tail_break
+
+ ldr $n0,[$rp,#-8*8]
+ adds $acc0,$acc0,$a0
+ adcs $acc1,$acc1,$a1
+ ldp $a0,$a1,[$np,#8*0]
+ adcs $acc2,$acc2,$a2
+ adcs $acc3,$acc3,$a3
+ ldp $a2,$a3,[$np,#8*2]
+ adcs $acc4,$acc4,$a4
+ adcs $acc5,$acc5,$a5
+ ldp $a4,$a5,[$np,#8*4]
+ adcs $acc6,$acc6,$a6
+ mov $cnt,#-8*8
+ adcs $acc7,$acc7,$a7
+ ldp $a6,$a7,[$np,#8*6]
+ add $np,$np,#8*8
+ //adc $carry,xzr,xzr // moved above
+ b .Lsqr8x_tail
+
+.align 4
+.Lsqr8x_tail_break:
+ ldr $n0,[x29,#112] // pull n0
+ add $cnt,$tp,#8*8 // end of current t[num] window
+
+ subs xzr,$topmost,#1 // "move" top-most carry to carry bit
+ adcs $t0,$acc0,$a0
+ adcs $t1,$acc1,$a1
+ ldp $acc0,$acc1,[$rp,#8*0]
+ adcs $acc2,$acc2,$a2
+ ldp $a0,$a1,[$t2,#8*0] // recall that $t2 is &n[0]
+ adcs $acc3,$acc3,$a3
+ ldp $a2,$a3,[$t2,#8*2]
+ adcs $acc4,$acc4,$a4
+ adcs $acc5,$acc5,$a5
+ ldp $a4,$a5,[$t2,#8*4]
+ adcs $acc6,$acc6,$a6
+ adcs $acc7,$acc7,$a7
+ ldp $a6,$a7,[$t2,#8*6]
+ add $np,$t2,#8*8
+ adc $topmost,xzr,xzr // top-most carry
+ mul $na0,$n0,$acc0
+ stp $t0,$t1,[$tp,#8*0]
+ stp $acc2,$acc3,[$tp,#8*2]
+ ldp $acc2,$acc3,[$rp,#8*2]
+ stp $acc4,$acc5,[$tp,#8*4]
+ ldp $acc4,$acc5,[$rp,#8*4]
+ cmp $cnt,x29 // did we hit the bottom?
+ stp $acc6,$acc7,[$tp,#8*6]
+ mov $tp,$rp // slide the window
+ ldp $acc6,$acc7,[$rp,#8*6]
+ mov $cnt,#8
+ b.ne .Lsqr8x_reduction
+
+ // Final step. We see if result is larger than modulus, and
+ // if it is, subtract the modulus. But comparison implies
+ // subtraction. So we subtract modulus, see if it borrowed,
+ // and conditionally copy original value.
+ ldr $rp,[x29,#96] // pull rp
+ add $tp,$tp,#8*8
+ subs $t0,$acc0,$a0
+ sbcs $t1,$acc1,$a1
+ sub $cnt,$num,#8*8
+ mov $ap_end,$rp // $rp copy
+
+.Lsqr8x_sub:
+ sbcs $t2,$acc2,$a2
+ ldp $a0,$a1,[$np,#8*0]
+ sbcs $t3,$acc3,$a3
+ stp $t0,$t1,[$rp,#8*0]
+ sbcs $t0,$acc4,$a4
+ ldp $a2,$a3,[$np,#8*2]
+ sbcs $t1,$acc5,$a5
+ stp $t2,$t3,[$rp,#8*2]
+ sbcs $t2,$acc6,$a6
+ ldp $a4,$a5,[$np,#8*4]
+ sbcs $t3,$acc7,$a7
+ ldp $a6,$a7,[$np,#8*6]
+ add $np,$np,#8*8
+ ldp $acc0,$acc1,[$tp,#8*0]
+ sub $cnt,$cnt,#8*8
+ ldp $acc2,$acc3,[$tp,#8*2]
+ ldp $acc4,$acc5,[$tp,#8*4]
+ ldp $acc6,$acc7,[$tp,#8*6]
+ add $tp,$tp,#8*8
+ stp $t0,$t1,[$rp,#8*4]
+ sbcs $t0,$acc0,$a0
+ stp $t2,$t3,[$rp,#8*6]
+ add $rp,$rp,#8*8
+ sbcs $t1,$acc1,$a1
+ cbnz $cnt,.Lsqr8x_sub
+
+ sbcs $t2,$acc2,$a2
+ mov $tp,sp
+ add $ap,sp,$num
+ ldp $a0,$a1,[$ap_end,#8*0]
+ sbcs $t3,$acc3,$a3
+ stp $t0,$t1,[$rp,#8*0]
+ sbcs $t0,$acc4,$a4
+ ldp $a2,$a3,[$ap_end,#8*2]
+ sbcs $t1,$acc5,$a5
+ stp $t2,$t3,[$rp,#8*2]
+ sbcs $t2,$acc6,$a6
+ ldp $acc0,$acc1,[$ap,#8*0]
+ sbcs $t3,$acc7,$a7
+ ldp $acc2,$acc3,[$ap,#8*2]
+ sbcs xzr,$topmost,xzr // did it borrow?
+ ldr x30,[x29,#8] // pull return address
+ stp $t0,$t1,[$rp,#8*4]
+ stp $t2,$t3,[$rp,#8*6]
+
+ sub $cnt,$num,#8*4
+.Lsqr4x_cond_copy:
+ sub $cnt,$cnt,#8*4
+ csel $t0,$acc0,$a0,lo
+ stp xzr,xzr,[$tp,#8*0]
+ csel $t1,$acc1,$a1,lo
+ ldp $a0,$a1,[$ap_end,#8*4]
+ ldp $acc0,$acc1,[$ap,#8*4]
+ csel $t2,$acc2,$a2,lo
+ stp xzr,xzr,[$tp,#8*2]
+ add $tp,$tp,#8*4
+ csel $t3,$acc3,$a3,lo
+ ldp $a2,$a3,[$ap_end,#8*6]
+ ldp $acc2,$acc3,[$ap,#8*6]
+ add $ap,$ap,#8*4
+ stp $t0,$t1,[$ap_end,#8*0]
+ stp $t2,$t3,[$ap_end,#8*2]
+ add $ap_end,$ap_end,#8*4
+ stp xzr,xzr,[$ap,#8*0]
+ stp xzr,xzr,[$ap,#8*2]
+ cbnz $cnt,.Lsqr4x_cond_copy
+
+ csel $t0,$acc0,$a0,lo
+ stp xzr,xzr,[$tp,#8*0]
+ csel $t1,$acc1,$a1,lo
+ stp xzr,xzr,[$tp,#8*2]
+ csel $t2,$acc2,$a2,lo
+ csel $t3,$acc3,$a3,lo
+ stp $t0,$t1,[$ap_end,#8*0]
+ stp $t2,$t3,[$ap_end,#8*2]
+
+ b .Lsqr8x_done
+
+.align 4
+.Lsqr8x8_post_condition:
+ adc $carry,xzr,xzr
+ ldr x30,[x29,#8] // pull return address
+ // $acc0-7,$carry hold result, $a0-7 hold modulus
+ subs $a0,$acc0,$a0
+ ldr $ap,[x29,#96] // pull rp
+ sbcs $a1,$acc1,$a1
+ stp xzr,xzr,[sp,#8*0]
+ sbcs $a2,$acc2,$a2
+ stp xzr,xzr,[sp,#8*2]
+ sbcs $a3,$acc3,$a3
+ stp xzr,xzr,[sp,#8*4]
+ sbcs $a4,$acc4,$a4
+ stp xzr,xzr,[sp,#8*6]
+ sbcs $a5,$acc5,$a5
+ stp xzr,xzr,[sp,#8*8]
+ sbcs $a6,$acc6,$a6
+ stp xzr,xzr,[sp,#8*10]
+ sbcs $a7,$acc7,$a7
+ stp xzr,xzr,[sp,#8*12]
+ sbcs $carry,$carry,xzr // did it borrow?
+ stp xzr,xzr,[sp,#8*14]
+
+ // $a0-7 hold result-modulus
+ csel $a0,$acc0,$a0,lo
+ csel $a1,$acc1,$a1,lo
+ csel $a2,$acc2,$a2,lo
+ csel $a3,$acc3,$a3,lo
+ stp $a0,$a1,[$ap,#8*0]
+ csel $a4,$acc4,$a4,lo
+ csel $a5,$acc5,$a5,lo
+ stp $a2,$a3,[$ap,#8*2]
+ csel $a6,$acc6,$a6,lo
+ csel $a7,$acc7,$a7,lo
+ stp $a4,$a5,[$ap,#8*4]
+ stp $a6,$a7,[$ap,#8*6]
+
+.Lsqr8x_done:
+ ldp x19,x20,[x29,#16]
+ mov sp,x29
+ ldp x21,x22,[x29,#32]
+ mov x0,#1
+ ldp x23,x24,[x29,#48]
+ ldp x25,x26,[x29,#64]
+ ldp x27,x28,[x29,#80]
+ ldr x29,[sp],#128
+ .inst 0xd50323bf // autiasp
+ ret
+.size __bn_sqr8x_mont,.-__bn_sqr8x_mont
+___
+}
+
+{
+########################################################################
+# Even though this might look as ARMv8 adaptation of mulx4x_mont from
+# x86_64-mont5 module, it's different in sense that it performs
+# reduction 256 bits at a time.
+
+my ($a0,$a1,$a2,$a3,
+ $t0,$t1,$t2,$t3,
+ $m0,$m1,$m2,$m3,
+ $acc0,$acc1,$acc2,$acc3,$acc4,
+ $bi,$mi,$tp,$ap_end,$cnt) = map("x$_",(6..17,19..28));
+my $bp_end=$rp;
+my ($carry,$topmost) = ($rp,"x30");
+
+$code.=<<___;
+.type __bn_mul4x_mont,%function
+.align 5
+__bn_mul4x_mont:
+ .inst 0xd503233f // paciasp
+ stp x29,x30,[sp,#-128]!
+ add x29,sp,#0
+ stp x19,x20,[sp,#16]
+ stp x21,x22,[sp,#32]
+ stp x23,x24,[sp,#48]
+ stp x25,x26,[sp,#64]
+ stp x27,x28,[sp,#80]
+
+ sub $tp,sp,$num,lsl#3
+ lsl $num,$num,#3
+ ldr $n0,[$n0] // *n0
+ sub sp,$tp,#8*4 // alloca
+
+ add $t0,$bp,$num
+ add $ap_end,$ap,$num
+ stp $rp,$t0,[x29,#96] // offload rp and &b[num]
+
+ ldr $bi,[$bp,#8*0] // b[0]
+ ldp $a0,$a1,[$ap,#8*0] // a[0..3]
+ ldp $a2,$a3,[$ap,#8*2]
+ add $ap,$ap,#8*4
+ mov $acc0,xzr
+ mov $acc1,xzr
+ mov $acc2,xzr
+ mov $acc3,xzr
+ ldp $m0,$m1,[$np,#8*0] // n[0..3]
+ ldp $m2,$m3,[$np,#8*2]
+ adds $np,$np,#8*4 // clear carry bit
+ mov $carry,xzr
+ mov $cnt,#0
+ mov $tp,sp
+
+.Loop_mul4x_1st_reduction:
+ mul $t0,$a0,$bi // lo(a[0..3]*b[0])
+ adc $carry,$carry,xzr // modulo-scheduled
+ mul $t1,$a1,$bi
+ add $cnt,$cnt,#8
+ mul $t2,$a2,$bi
+ and $cnt,$cnt,#31
+ mul $t3,$a3,$bi
+ adds $acc0,$acc0,$t0
+ umulh $t0,$a0,$bi // hi(a[0..3]*b[0])
+ adcs $acc1,$acc1,$t1
+ mul $mi,$acc0,$n0 // t[0]*n0
+ adcs $acc2,$acc2,$t2
+ umulh $t1,$a1,$bi
+ adcs $acc3,$acc3,$t3
+ umulh $t2,$a2,$bi
+ adc $acc4,xzr,xzr
+ umulh $t3,$a3,$bi
+ ldr $bi,[$bp,$cnt] // next b[i] (or b[0])
+ adds $acc1,$acc1,$t0
+ // (*) mul $t0,$m0,$mi // lo(n[0..3]*t[0]*n0)
+ str $mi,[$tp],#8 // put aside t[0]*n0 for tail processing
+ adcs $acc2,$acc2,$t1
+ mul $t1,$m1,$mi
+ adcs $acc3,$acc3,$t2
+ mul $t2,$m2,$mi
+ adc $acc4,$acc4,$t3 // can't overflow
+ mul $t3,$m3,$mi
+ // (*) adds xzr,$acc0,$t0
+ subs xzr,$acc0,#1 // (*)
+ umulh $t0,$m0,$mi // hi(n[0..3]*t[0]*n0)
+ adcs $acc0,$acc1,$t1
+ umulh $t1,$m1,$mi
+ adcs $acc1,$acc2,$t2
+ umulh $t2,$m2,$mi
+ adcs $acc2,$acc3,$t3
+ umulh $t3,$m3,$mi
+ adcs $acc3,$acc4,$carry
+ adc $carry,xzr,xzr
+ adds $acc0,$acc0,$t0
+ sub $t0,$ap_end,$ap
+ adcs $acc1,$acc1,$t1
+ adcs $acc2,$acc2,$t2
+ adcs $acc3,$acc3,$t3
+ //adc $carry,$carry,xzr
+ cbnz $cnt,.Loop_mul4x_1st_reduction
+
+ cbz $t0,.Lmul4x4_post_condition
+
+ ldp $a0,$a1,[$ap,#8*0] // a[4..7]
+ ldp $a2,$a3,[$ap,#8*2]
+ add $ap,$ap,#8*4
+ ldr $mi,[sp] // a[0]*n0
+ ldp $m0,$m1,[$np,#8*0] // n[4..7]
+ ldp $m2,$m3,[$np,#8*2]
+ add $np,$np,#8*4
+
+.Loop_mul4x_1st_tail:
+ mul $t0,$a0,$bi // lo(a[4..7]*b[i])
+ adc $carry,$carry,xzr // modulo-scheduled
+ mul $t1,$a1,$bi
+ add $cnt,$cnt,#8
+ mul $t2,$a2,$bi
+ and $cnt,$cnt,#31
+ mul $t3,$a3,$bi
+ adds $acc0,$acc0,$t0
+ umulh $t0,$a0,$bi // hi(a[4..7]*b[i])
+ adcs $acc1,$acc1,$t1
+ umulh $t1,$a1,$bi
+ adcs $acc2,$acc2,$t2
+ umulh $t2,$a2,$bi
+ adcs $acc3,$acc3,$t3
+ umulh $t3,$a3,$bi
+ adc $acc4,xzr,xzr
+ ldr $bi,[$bp,$cnt] // next b[i] (or b[0])
+ adds $acc1,$acc1,$t0
+ mul $t0,$m0,$mi // lo(n[4..7]*a[0]*n0)
+ adcs $acc2,$acc2,$t1
+ mul $t1,$m1,$mi
+ adcs $acc3,$acc3,$t2
+ mul $t2,$m2,$mi
+ adc $acc4,$acc4,$t3 // can't overflow
+ mul $t3,$m3,$mi
+ adds $acc0,$acc0,$t0
+ umulh $t0,$m0,$mi // hi(n[4..7]*a[0]*n0)
+ adcs $acc1,$acc1,$t1
+ umulh $t1,$m1,$mi
+ adcs $acc2,$acc2,$t2
+ umulh $t2,$m2,$mi
+ adcs $acc3,$acc3,$t3
+ adcs $acc4,$acc4,$carry
+ umulh $t3,$m3,$mi
+ adc $carry,xzr,xzr
+ ldr $mi,[sp,$cnt] // next t[0]*n0
+ str $acc0,[$tp],#8 // result!!!
+ adds $acc0,$acc1,$t0
+ sub $t0,$ap_end,$ap // done yet?
+ adcs $acc1,$acc2,$t1
+ adcs $acc2,$acc3,$t2
+ adcs $acc3,$acc4,$t3
+ //adc $carry,$carry,xzr
+ cbnz $cnt,.Loop_mul4x_1st_tail
+
+ sub $t1,$ap_end,$num // rewinded $ap
+ cbz $t0,.Lmul4x_proceed
+
+ ldp $a0,$a1,[$ap,#8*0]
+ ldp $a2,$a3,[$ap,#8*2]
+ add $ap,$ap,#8*4
+ ldp $m0,$m1,[$np,#8*0]
+ ldp $m2,$m3,[$np,#8*2]
+ add $np,$np,#8*4
+ b .Loop_mul4x_1st_tail
+
+.align 5
+.Lmul4x_proceed:
+ ldr $bi,[$bp,#8*4]! // *++b
+ adc $topmost,$carry,xzr
+ ldp $a0,$a1,[$t1,#8*0] // a[0..3]
+ sub $np,$np,$num // rewind np
+ ldp $a2,$a3,[$t1,#8*2]
+ add $ap,$t1,#8*4
+
+ stp $acc0,$acc1,[$tp,#8*0] // result!!!
+ ldp $acc0,$acc1,[sp,#8*4] // t[0..3]
+ stp $acc2,$acc3,[$tp,#8*2] // result!!!
+ ldp $acc2,$acc3,[sp,#8*6]
+
+ ldp $m0,$m1,[$np,#8*0] // n[0..3]
+ mov $tp,sp
+ ldp $m2,$m3,[$np,#8*2]
+ adds $np,$np,#8*4 // clear carry bit
+ mov $carry,xzr
+
+.align 4
+.Loop_mul4x_reduction:
+ mul $t0,$a0,$bi // lo(a[0..3]*b[4])
+ adc $carry,$carry,xzr // modulo-scheduled
+ mul $t1,$a1,$bi
+ add $cnt,$cnt,#8
+ mul $t2,$a2,$bi
+ and $cnt,$cnt,#31
+ mul $t3,$a3,$bi
+ adds $acc0,$acc0,$t0
+ umulh $t0,$a0,$bi // hi(a[0..3]*b[4])
+ adcs $acc1,$acc1,$t1
+ mul $mi,$acc0,$n0 // t[0]*n0
+ adcs $acc2,$acc2,$t2
+ umulh $t1,$a1,$bi
+ adcs $acc3,$acc3,$t3
+ umulh $t2,$a2,$bi
+ adc $acc4,xzr,xzr
+ umulh $t3,$a3,$bi
+ ldr $bi,[$bp,$cnt] // next b[i]
+ adds $acc1,$acc1,$t0
+ // (*) mul $t0,$m0,$mi
+ str $mi,[$tp],#8 // put aside t[0]*n0 for tail processing
+ adcs $acc2,$acc2,$t1
+ mul $t1,$m1,$mi // lo(n[0..3]*t[0]*n0
+ adcs $acc3,$acc3,$t2
+ mul $t2,$m2,$mi
+ adc $acc4,$acc4,$t3 // can't overflow
+ mul $t3,$m3,$mi
+ // (*) adds xzr,$acc0,$t0
+ subs xzr,$acc0,#1 // (*)
+ umulh $t0,$m0,$mi // hi(n[0..3]*t[0]*n0
+ adcs $acc0,$acc1,$t1
+ umulh $t1,$m1,$mi
+ adcs $acc1,$acc2,$t2
+ umulh $t2,$m2,$mi
+ adcs $acc2,$acc3,$t3
+ umulh $t3,$m3,$mi
+ adcs $acc3,$acc4,$carry
+ adc $carry,xzr,xzr
+ adds $acc0,$acc0,$t0
+ adcs $acc1,$acc1,$t1
+ adcs $acc2,$acc2,$t2
+ adcs $acc3,$acc3,$t3
+ //adc $carry,$carry,xzr
+ cbnz $cnt,.Loop_mul4x_reduction
+
+ adc $carry,$carry,xzr
+ ldp $t0,$t1,[$tp,#8*4] // t[4..7]
+ ldp $t2,$t3,[$tp,#8*6]
+ ldp $a0,$a1,[$ap,#8*0] // a[4..7]
+ ldp $a2,$a3,[$ap,#8*2]
+ add $ap,$ap,#8*4
+ adds $acc0,$acc0,$t0
+ adcs $acc1,$acc1,$t1
+ adcs $acc2,$acc2,$t2
+ adcs $acc3,$acc3,$t3
+ //adc $carry,$carry,xzr
+
+ ldr $mi,[sp] // t[0]*n0
+ ldp $m0,$m1,[$np,#8*0] // n[4..7]
+ ldp $m2,$m3,[$np,#8*2]
+ add $np,$np,#8*4
+
+.align 4
+.Loop_mul4x_tail:
+ mul $t0,$a0,$bi // lo(a[4..7]*b[4])
+ adc $carry,$carry,xzr // modulo-scheduled
+ mul $t1,$a1,$bi
+ add $cnt,$cnt,#8
+ mul $t2,$a2,$bi
+ and $cnt,$cnt,#31
+ mul $t3,$a3,$bi
+ adds $acc0,$acc0,$t0
+ umulh $t0,$a0,$bi // hi(a[4..7]*b[4])
+ adcs $acc1,$acc1,$t1
+ umulh $t1,$a1,$bi
+ adcs $acc2,$acc2,$t2
+ umulh $t2,$a2,$bi
+ adcs $acc3,$acc3,$t3
+ umulh $t3,$a3,$bi
+ adc $acc4,xzr,xzr
+ ldr $bi,[$bp,$cnt] // next b[i]
+ adds $acc1,$acc1,$t0
+ mul $t0,$m0,$mi // lo(n[4..7]*t[0]*n0)
+ adcs $acc2,$acc2,$t1
+ mul $t1,$m1,$mi
+ adcs $acc3,$acc3,$t2
+ mul $t2,$m2,$mi
+ adc $acc4,$acc4,$t3 // can't overflow
+ mul $t3,$m3,$mi
+ adds $acc0,$acc0,$t0
+ umulh $t0,$m0,$mi // hi(n[4..7]*t[0]*n0)
+ adcs $acc1,$acc1,$t1
+ umulh $t1,$m1,$mi
+ adcs $acc2,$acc2,$t2
+ umulh $t2,$m2,$mi
+ adcs $acc3,$acc3,$t3
+ umulh $t3,$m3,$mi
+ adcs $acc4,$acc4,$carry
+ ldr $mi,[sp,$cnt] // next a[0]*n0
+ adc $carry,xzr,xzr
+ str $acc0,[$tp],#8 // result!!!
+ adds $acc0,$acc1,$t0
+ sub $t0,$ap_end,$ap // done yet?
+ adcs $acc1,$acc2,$t1
+ adcs $acc2,$acc3,$t2
+ adcs $acc3,$acc4,$t3
+ //adc $carry,$carry,xzr
+ cbnz $cnt,.Loop_mul4x_tail
+
+ sub $t1,$np,$num // rewinded np?
+ adc $carry,$carry,xzr
+ cbz $t0,.Loop_mul4x_break
+
+ ldp $t0,$t1,[$tp,#8*4]
+ ldp $t2,$t3,[$tp,#8*6]
+ ldp $a0,$a1,[$ap,#8*0]
+ ldp $a2,$a3,[$ap,#8*2]
+ add $ap,$ap,#8*4
+ adds $acc0,$acc0,$t0
+ adcs $acc1,$acc1,$t1
+ adcs $acc2,$acc2,$t2
+ adcs $acc3,$acc3,$t3
+ //adc $carry,$carry,xzr
+ ldp $m0,$m1,[$np,#8*0]
+ ldp $m2,$m3,[$np,#8*2]
+ add $np,$np,#8*4
+ b .Loop_mul4x_tail
+
+.align 4
+.Loop_mul4x_break:
+ ldp $t2,$t3,[x29,#96] // pull rp and &b[num]
+ adds $acc0,$acc0,$topmost
+ add $bp,$bp,#8*4 // bp++
+ adcs $acc1,$acc1,xzr
+ sub $ap,$ap,$num // rewind ap
+ adcs $acc2,$acc2,xzr
+ stp $acc0,$acc1,[$tp,#8*0] // result!!!
+ adcs $acc3,$acc3,xzr
+ ldp $acc0,$acc1,[sp,#8*4] // t[0..3]
+ adc $topmost,$carry,xzr
+ stp $acc2,$acc3,[$tp,#8*2] // result!!!
+ cmp $bp,$t3 // done yet?
+ ldp $acc2,$acc3,[sp,#8*6]
+ ldp $m0,$m1,[$t1,#8*0] // n[0..3]
+ ldp $m2,$m3,[$t1,#8*2]
+ add $np,$t1,#8*4
+ b.eq .Lmul4x_post
+
+ ldr $bi,[$bp]
+ ldp $a0,$a1,[$ap,#8*0] // a[0..3]
+ ldp $a2,$a3,[$ap,#8*2]
+ adds $ap,$ap,#8*4 // clear carry bit
+ mov $carry,xzr
+ mov $tp,sp
+ b .Loop_mul4x_reduction
+
+.align 4
+.Lmul4x_post:
+ // Final step. We see if result is larger than modulus, and
+ // if it is, subtract the modulus. But comparison implies
+ // subtraction. So we subtract modulus, see if it borrowed,
+ // and conditionally copy original value.
+ mov $rp,$t2
+ mov $ap_end,$t2 // $rp copy
+ subs $t0,$acc0,$m0
+ add $tp,sp,#8*8
+ sbcs $t1,$acc1,$m1
+ sub $cnt,$num,#8*4
+
+.Lmul4x_sub:
+ sbcs $t2,$acc2,$m2
+ ldp $m0,$m1,[$np,#8*0]
+ sub $cnt,$cnt,#8*4
+ ldp $acc0,$acc1,[$tp,#8*0]
+ sbcs $t3,$acc3,$m3
+ ldp $m2,$m3,[$np,#8*2]
+ add $np,$np,#8*4
+ ldp $acc2,$acc3,[$tp,#8*2]
+ add $tp,$tp,#8*4
+ stp $t0,$t1,[$rp,#8*0]
+ sbcs $t0,$acc0,$m0
+ stp $t2,$t3,[$rp,#8*2]
+ add $rp,$rp,#8*4
+ sbcs $t1,$acc1,$m1
+ cbnz $cnt,.Lmul4x_sub
+
+ sbcs $t2,$acc2,$m2
+ mov $tp,sp
+ add $ap,sp,#8*4
+ ldp $a0,$a1,[$ap_end,#8*0]
+ sbcs $t3,$acc3,$m3
+ stp $t0,$t1,[$rp,#8*0]
+ ldp $a2,$a3,[$ap_end,#8*2]
+ stp $t2,$t3,[$rp,#8*2]
+ ldp $acc0,$acc1,[$ap,#8*0]
+ ldp $acc2,$acc3,[$ap,#8*2]
+ sbcs xzr,$topmost,xzr // did it borrow?
+ ldr x30,[x29,#8] // pull return address
+
+ sub $cnt,$num,#8*4
+.Lmul4x_cond_copy:
+ sub $cnt,$cnt,#8*4
+ csel $t0,$acc0,$a0,lo
+ stp xzr,xzr,[$tp,#8*0]
+ csel $t1,$acc1,$a1,lo
+ ldp $a0,$a1,[$ap_end,#8*4]
+ ldp $acc0,$acc1,[$ap,#8*4]
+ csel $t2,$acc2,$a2,lo
+ stp xzr,xzr,[$tp,#8*2]
+ add $tp,$tp,#8*4
+ csel $t3,$acc3,$a3,lo
+ ldp $a2,$a3,[$ap_end,#8*6]
+ ldp $acc2,$acc3,[$ap,#8*6]
+ add $ap,$ap,#8*4
+ stp $t0,$t1,[$ap_end,#8*0]
+ stp $t2,$t3,[$ap_end,#8*2]
+ add $ap_end,$ap_end,#8*4
+ cbnz $cnt,.Lmul4x_cond_copy
+
+ csel $t0,$acc0,$a0,lo
+ stp xzr,xzr,[$tp,#8*0]
+ csel $t1,$acc1,$a1,lo
+ stp xzr,xzr,[$tp,#8*2]
+ csel $t2,$acc2,$a2,lo
+ stp xzr,xzr,[$tp,#8*3]
+ csel $t3,$acc3,$a3,lo
+ stp xzr,xzr,[$tp,#8*4]
+ stp $t0,$t1,[$ap_end,#8*0]
+ stp $t2,$t3,[$ap_end,#8*2]
+
+ b .Lmul4x_done
+
+.align 4
+.Lmul4x4_post_condition:
+ adc $carry,$carry,xzr
+ ldr $ap,[x29,#96] // pull rp
+ // $acc0-3,$carry hold result, $m0-7 hold modulus
+ subs $a0,$acc0,$m0
+ ldr x30,[x29,#8] // pull return address
+ sbcs $a1,$acc1,$m1
+ stp xzr,xzr,[sp,#8*0]
+ sbcs $a2,$acc2,$m2
+ stp xzr,xzr,[sp,#8*2]
+ sbcs $a3,$acc3,$m3
+ stp xzr,xzr,[sp,#8*4]
+ sbcs xzr,$carry,xzr // did it borrow?
+ stp xzr,xzr,[sp,#8*6]
+
+ // $a0-3 hold result-modulus
+ csel $a0,$acc0,$a0,lo
+ csel $a1,$acc1,$a1,lo
+ csel $a2,$acc2,$a2,lo
+ csel $a3,$acc3,$a3,lo
+ stp $a0,$a1,[$ap,#8*0]
+ stp $a2,$a3,[$ap,#8*2]
+
+.Lmul4x_done:
+ ldp x19,x20,[x29,#16]
+ mov sp,x29
+ ldp x21,x22,[x29,#32]
+ mov x0,#1
+ ldp x23,x24,[x29,#48]
+ ldp x25,x26,[x29,#64]
+ ldp x27,x28,[x29,#80]
+ ldr x29,[sp],#128
+ .inst 0xd50323bf // autiasp
+ ret
+.size __bn_mul4x_mont,.-__bn_mul4x_mont
+___
+}
+$code.=<<___;
+.asciz "Montgomery Multiplication for ARMv8, CRYPTOGAMS by <appro\@openssl.org>"
+.align 4
+___
+
+print $code;
+
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/bn-586.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/bn-586.pl
new file mode 100644
index 0000000..e042240
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/bn-586.pl
@@ -0,0 +1,785 @@
+#! /usr/bin/env perl
+# Copyright 1995-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+push(@INC,"${dir}","${dir}../../perlasm");
+require "x86asm.pl";
+
+$output = pop;
+open STDOUT,">$output";
+
+&asm_init($ARGV[0]);
+
+$sse2=0;
+for (@ARGV) { $sse2=1 if (/-DOPENSSL_IA32_SSE2/); }
+
+&external_label("OPENSSL_ia32cap_P") if ($sse2);
+
+&bn_mul_add_words("bn_mul_add_words");
+&bn_mul_words("bn_mul_words");
+&bn_sqr_words("bn_sqr_words");
+&bn_div_words("bn_div_words");
+&bn_add_words("bn_add_words");
+&bn_sub_words("bn_sub_words");
+&bn_sub_part_words("bn_sub_part_words");
+
+&asm_finish();
+
+close STDOUT or die "error closing STDOUT: $!";
+
+sub bn_mul_add_words
+ {
+ local($name)=@_;
+
+ &function_begin_B($name,$sse2?"EXTRN\t_OPENSSL_ia32cap_P:DWORD":"");
+
+ $r="eax";
+ $a="edx";
+ $c="ecx";
+
+ if ($sse2) {
+ &picmeup("eax","OPENSSL_ia32cap_P");
+ &bt(&DWP(0,"eax"),26);
+ &jnc(&label("maw_non_sse2"));
+
+ &mov($r,&wparam(0));
+ &mov($a,&wparam(1));
+ &mov($c,&wparam(2));
+ &movd("mm0",&wparam(3)); # mm0 = w
+ &pxor("mm1","mm1"); # mm1 = carry_in
+ &jmp(&label("maw_sse2_entry"));
+
+ &set_label("maw_sse2_unrolled",16);
+ &movd("mm3",&DWP(0,$r,"",0)); # mm3 = r[0]
+ &paddq("mm1","mm3"); # mm1 = carry_in + r[0]
+ &movd("mm2",&DWP(0,$a,"",0)); # mm2 = a[0]
+ &pmuludq("mm2","mm0"); # mm2 = w*a[0]
+ &movd("mm4",&DWP(4,$a,"",0)); # mm4 = a[1]
+ &pmuludq("mm4","mm0"); # mm4 = w*a[1]
+ &movd("mm6",&DWP(8,$a,"",0)); # mm6 = a[2]
+ &pmuludq("mm6","mm0"); # mm6 = w*a[2]
+ &movd("mm7",&DWP(12,$a,"",0)); # mm7 = a[3]
+ &pmuludq("mm7","mm0"); # mm7 = w*a[3]
+ &paddq("mm1","mm2"); # mm1 = carry_in + r[0] + w*a[0]
+ &movd("mm3",&DWP(4,$r,"",0)); # mm3 = r[1]
+ &paddq("mm3","mm4"); # mm3 = r[1] + w*a[1]
+ &movd("mm5",&DWP(8,$r,"",0)); # mm5 = r[2]
+ &paddq("mm5","mm6"); # mm5 = r[2] + w*a[2]
+ &movd("mm4",&DWP(12,$r,"",0)); # mm4 = r[3]
+ &paddq("mm7","mm4"); # mm7 = r[3] + w*a[3]
+ &movd(&DWP(0,$r,"",0),"mm1");
+ &movd("mm2",&DWP(16,$a,"",0)); # mm2 = a[4]
+ &pmuludq("mm2","mm0"); # mm2 = w*a[4]
+ &psrlq("mm1",32); # mm1 = carry0
+ &movd("mm4",&DWP(20,$a,"",0)); # mm4 = a[5]
+ &pmuludq("mm4","mm0"); # mm4 = w*a[5]
+ &paddq("mm1","mm3"); # mm1 = carry0 + r[1] + w*a[1]
+ &movd("mm6",&DWP(24,$a,"",0)); # mm6 = a[6]
+ &pmuludq("mm6","mm0"); # mm6 = w*a[6]
+ &movd(&DWP(4,$r,"",0),"mm1");
+ &psrlq("mm1",32); # mm1 = carry1
+ &movd("mm3",&DWP(28,$a,"",0)); # mm3 = a[7]
+ &add($a,32);
+ &pmuludq("mm3","mm0"); # mm3 = w*a[7]
+ &paddq("mm1","mm5"); # mm1 = carry1 + r[2] + w*a[2]
+ &movd("mm5",&DWP(16,$r,"",0)); # mm5 = r[4]
+ &paddq("mm2","mm5"); # mm2 = r[4] + w*a[4]
+ &movd(&DWP(8,$r,"",0),"mm1");
+ &psrlq("mm1",32); # mm1 = carry2
+ &paddq("mm1","mm7"); # mm1 = carry2 + r[3] + w*a[3]
+ &movd("mm5",&DWP(20,$r,"",0)); # mm5 = r[5]
+ &paddq("mm4","mm5"); # mm4 = r[5] + w*a[5]
+ &movd(&DWP(12,$r,"",0),"mm1");
+ &psrlq("mm1",32); # mm1 = carry3
+ &paddq("mm1","mm2"); # mm1 = carry3 + r[4] + w*a[4]
+ &movd("mm5",&DWP(24,$r,"",0)); # mm5 = r[6]
+ &paddq("mm6","mm5"); # mm6 = r[6] + w*a[6]
+ &movd(&DWP(16,$r,"",0),"mm1");
+ &psrlq("mm1",32); # mm1 = carry4
+ &paddq("mm1","mm4"); # mm1 = carry4 + r[5] + w*a[5]
+ &movd("mm5",&DWP(28,$r,"",0)); # mm5 = r[7]
+ &paddq("mm3","mm5"); # mm3 = r[7] + w*a[7]
+ &movd(&DWP(20,$r,"",0),"mm1");
+ &psrlq("mm1",32); # mm1 = carry5
+ &paddq("mm1","mm6"); # mm1 = carry5 + r[6] + w*a[6]
+ &movd(&DWP(24,$r,"",0),"mm1");
+ &psrlq("mm1",32); # mm1 = carry6
+ &paddq("mm1","mm3"); # mm1 = carry6 + r[7] + w*a[7]
+ &movd(&DWP(28,$r,"",0),"mm1");
+ &lea($r,&DWP(32,$r));
+ &psrlq("mm1",32); # mm1 = carry_out
+
+ &sub($c,8);
+ &jz(&label("maw_sse2_exit"));
+ &set_label("maw_sse2_entry");
+ &test($c,0xfffffff8);
+ &jnz(&label("maw_sse2_unrolled"));
+
+ &set_label("maw_sse2_loop",4);
+ &movd("mm2",&DWP(0,$a)); # mm2 = a[i]
+ &movd("mm3",&DWP(0,$r)); # mm3 = r[i]
+ &pmuludq("mm2","mm0"); # a[i] *= w
+ &lea($a,&DWP(4,$a));
+ &paddq("mm1","mm3"); # carry += r[i]
+ &paddq("mm1","mm2"); # carry += a[i]*w
+ &movd(&DWP(0,$r),"mm1"); # r[i] = carry_low
+ &sub($c,1);
+ &psrlq("mm1",32); # carry = carry_high
+ &lea($r,&DWP(4,$r));
+ &jnz(&label("maw_sse2_loop"));
+ &set_label("maw_sse2_exit");
+ &movd("eax","mm1"); # c = carry_out
+ &emms();
+ &ret();
+
+ &set_label("maw_non_sse2",16);
+ }
+
+ # function_begin prologue
+ &push("ebp");
+ &push("ebx");
+ &push("esi");
+ &push("edi");
+
+ &comment("");
+ $Low="eax";
+ $High="edx";
+ $a="ebx";
+ $w="ebp";
+ $r="edi";
+ $c="esi";
+
+ &xor($c,$c); # clear carry
+ &mov($r,&wparam(0)); #
+
+ &mov("ecx",&wparam(2)); #
+ &mov($a,&wparam(1)); #
+
+ &and("ecx",0xfffffff8); # num / 8
+ &mov($w,&wparam(3)); #
+
+ &push("ecx"); # Up the stack for a tmp variable
+
+ &jz(&label("maw_finish"));
+
+ &set_label("maw_loop",16);
+
+ for ($i=0; $i<32; $i+=4)
+ {
+ &comment("Round $i");
+
+ &mov("eax",&DWP($i,$a)); # *a
+ &mul($w); # *a * w
+ &add("eax",$c); # L(t)+= c
+ &adc("edx",0); # H(t)+=carry
+ &add("eax",&DWP($i,$r)); # L(t)+= *r
+ &adc("edx",0); # H(t)+=carry
+ &mov(&DWP($i,$r),"eax"); # *r= L(t);
+ &mov($c,"edx"); # c= H(t);
+ }
+
+ &comment("");
+ &sub("ecx",8);
+ &lea($a,&DWP(32,$a));
+ &lea($r,&DWP(32,$r));
+ &jnz(&label("maw_loop"));
+
+ &set_label("maw_finish",0);
+ &mov("ecx",&wparam(2)); # get num
+ &and("ecx",7);
+ &jnz(&label("maw_finish2")); # helps branch prediction
+ &jmp(&label("maw_end"));
+
+ &set_label("maw_finish2",1);
+ for ($i=0; $i<7; $i++)
+ {
+ &comment("Tail Round $i");
+ &mov("eax",&DWP($i*4,$a)); # *a
+ &mul($w); # *a * w
+ &add("eax",$c); # L(t)+=c
+ &adc("edx",0); # H(t)+=carry
+ &add("eax",&DWP($i*4,$r)); # L(t)+= *r
+ &adc("edx",0); # H(t)+=carry
+ &dec("ecx") if ($i != 7-1);
+ &mov(&DWP($i*4,$r),"eax"); # *r= L(t);
+ &mov($c,"edx"); # c= H(t);
+ &jz(&label("maw_end")) if ($i != 7-1);
+ }
+ &set_label("maw_end",0);
+ &mov("eax",$c);
+
+ &pop("ecx"); # clear variable from
+
+ &function_end($name);
+ }
+
+sub bn_mul_words
+ {
+ local($name)=@_;
+
+ &function_begin_B($name,$sse2?"EXTRN\t_OPENSSL_ia32cap_P:DWORD":"");
+
+ $r="eax";
+ $a="edx";
+ $c="ecx";
+
+ if ($sse2) {
+ &picmeup("eax","OPENSSL_ia32cap_P");
+ &bt(&DWP(0,"eax"),26);
+ &jnc(&label("mw_non_sse2"));
+
+ &mov($r,&wparam(0));
+ &mov($a,&wparam(1));
+ &mov($c,&wparam(2));
+ &movd("mm0",&wparam(3)); # mm0 = w
+ &pxor("mm1","mm1"); # mm1 = carry = 0
+
+ &set_label("mw_sse2_loop",16);
+ &movd("mm2",&DWP(0,$a)); # mm2 = a[i]
+ &pmuludq("mm2","mm0"); # a[i] *= w
+ &lea($a,&DWP(4,$a));
+ &paddq("mm1","mm2"); # carry += a[i]*w
+ &movd(&DWP(0,$r),"mm1"); # r[i] = carry_low
+ &sub($c,1);
+ &psrlq("mm1",32); # carry = carry_high
+ &lea($r,&DWP(4,$r));
+ &jnz(&label("mw_sse2_loop"));
+
+ &movd("eax","mm1"); # return carry
+ &emms();
+ &ret();
+ &set_label("mw_non_sse2",16);
+ }
+
+ # function_begin prologue
+ &push("ebp");
+ &push("ebx");
+ &push("esi");
+ &push("edi");
+
+ &comment("");
+ $Low="eax";
+ $High="edx";
+ $a="ebx";
+ $w="ecx";
+ $r="edi";
+ $c="esi";
+ $num="ebp";
+
+ &xor($c,$c); # clear carry
+ &mov($r,&wparam(0)); #
+ &mov($a,&wparam(1)); #
+ &mov($num,&wparam(2)); #
+ &mov($w,&wparam(3)); #
+
+ &and($num,0xfffffff8); # num / 8
+ &jz(&label("mw_finish"));
+
+ &set_label("mw_loop",0);
+ for ($i=0; $i<32; $i+=4)
+ {
+ &comment("Round $i");
+
+ &mov("eax",&DWP($i,$a,"",0)); # *a
+ &mul($w); # *a * w
+ &add("eax",$c); # L(t)+=c
+ # XXX
+
+ &adc("edx",0); # H(t)+=carry
+ &mov(&DWP($i,$r,"",0),"eax"); # *r= L(t);
+
+ &mov($c,"edx"); # c= H(t);
+ }
+
+ &comment("");
+ &add($a,32);
+ &add($r,32);
+ &sub($num,8);
+ &jz(&label("mw_finish"));
+ &jmp(&label("mw_loop"));
+
+ &set_label("mw_finish",0);
+ &mov($num,&wparam(2)); # get num
+ &and($num,7);
+ &jnz(&label("mw_finish2"));
+ &jmp(&label("mw_end"));
+
+ &set_label("mw_finish2",1);
+ for ($i=0; $i<7; $i++)
+ {
+ &comment("Tail Round $i");
+ &mov("eax",&DWP($i*4,$a,"",0));# *a
+ &mul($w); # *a * w
+ &add("eax",$c); # L(t)+=c
+ # XXX
+ &adc("edx",0); # H(t)+=carry
+ &mov(&DWP($i*4,$r,"",0),"eax");# *r= L(t);
+ &mov($c,"edx"); # c= H(t);
+ &dec($num) if ($i != 7-1);
+ &jz(&label("mw_end")) if ($i != 7-1);
+ }
+ &set_label("mw_end",0);
+ &mov("eax",$c);
+
+ &function_end($name);
+ }
+
+sub bn_sqr_words
+ {
+ local($name)=@_;
+
+ &function_begin_B($name,$sse2?"EXTRN\t_OPENSSL_ia32cap_P:DWORD":"");
+
+ $r="eax";
+ $a="edx";
+ $c="ecx";
+
+ if ($sse2) {
+ &picmeup("eax","OPENSSL_ia32cap_P");
+ &bt(&DWP(0,"eax"),26);
+ &jnc(&label("sqr_non_sse2"));
+
+ &mov($r,&wparam(0));
+ &mov($a,&wparam(1));
+ &mov($c,&wparam(2));
+
+ &set_label("sqr_sse2_loop",16);
+ &movd("mm0",&DWP(0,$a)); # mm0 = a[i]
+ &pmuludq("mm0","mm0"); # a[i] *= a[i]
+ &lea($a,&DWP(4,$a)); # a++
+ &movq(&QWP(0,$r),"mm0"); # r[i] = a[i]*a[i]
+ &sub($c,1);
+ &lea($r,&DWP(8,$r)); # r += 2
+ &jnz(&label("sqr_sse2_loop"));
+
+ &emms();
+ &ret();
+ &set_label("sqr_non_sse2",16);
+ }
+
+ # function_begin prologue
+ &push("ebp");
+ &push("ebx");
+ &push("esi");
+ &push("edi");
+
+ &comment("");
+ $r="esi";
+ $a="edi";
+ $num="ebx";
+
+ &mov($r,&wparam(0)); #
+ &mov($a,&wparam(1)); #
+ &mov($num,&wparam(2)); #
+
+ &and($num,0xfffffff8); # num / 8
+ &jz(&label("sw_finish"));
+
+ &set_label("sw_loop",0);
+ for ($i=0; $i<32; $i+=4)
+ {
+ &comment("Round $i");
+ &mov("eax",&DWP($i,$a,"",0)); # *a
+ # XXX
+ &mul("eax"); # *a * *a
+ &mov(&DWP($i*2,$r,"",0),"eax"); #
+ &mov(&DWP($i*2+4,$r,"",0),"edx");#
+ }
+
+ &comment("");
+ &add($a,32);
+ &add($r,64);
+ &sub($num,8);
+ &jnz(&label("sw_loop"));
+
+ &set_label("sw_finish",0);
+ &mov($num,&wparam(2)); # get num
+ &and($num,7);
+ &jz(&label("sw_end"));
+
+ for ($i=0; $i<7; $i++)
+ {
+ &comment("Tail Round $i");
+ &mov("eax",&DWP($i*4,$a,"",0)); # *a
+ # XXX
+ &mul("eax"); # *a * *a
+ &mov(&DWP($i*8,$r,"",0),"eax"); #
+ &dec($num) if ($i != 7-1);
+ &mov(&DWP($i*8+4,$r,"",0),"edx");
+ &jz(&label("sw_end")) if ($i != 7-1);
+ }
+ &set_label("sw_end",0);
+
+ &function_end($name);
+ }
+
+sub bn_div_words
+ {
+ local($name)=@_;
+
+ &function_begin_B($name,"");
+ &mov("edx",&wparam(0)); #
+ &mov("eax",&wparam(1)); #
+ &mov("ecx",&wparam(2)); #
+ &div("ecx");
+ &ret();
+ &function_end_B($name);
+ }
+
+sub bn_add_words
+ {
+ local($name)=@_;
+
+ &function_begin($name,"");
+
+ &comment("");
+ $a="esi";
+ $b="edi";
+ $c="eax";
+ $r="ebx";
+ $tmp1="ecx";
+ $tmp2="edx";
+ $num="ebp";
+
+ &mov($r,&wparam(0)); # get r
+ &mov($a,&wparam(1)); # get a
+ &mov($b,&wparam(2)); # get b
+ &mov($num,&wparam(3)); # get num
+ &xor($c,$c); # clear carry
+ &and($num,0xfffffff8); # num / 8
+
+ &jz(&label("aw_finish"));
+
+ &set_label("aw_loop",0);
+ for ($i=0; $i<8; $i++)
+ {
+ &comment("Round $i");
+
+ &mov($tmp1,&DWP($i*4,$a,"",0)); # *a
+ &mov($tmp2,&DWP($i*4,$b,"",0)); # *b
+ &add($tmp1,$c);
+ &mov($c,0);
+ &adc($c,$c);
+ &add($tmp1,$tmp2);
+ &adc($c,0);
+ &mov(&DWP($i*4,$r,"",0),$tmp1); # *r
+ }
+
+ &comment("");
+ &add($a,32);
+ &add($b,32);
+ &add($r,32);
+ &sub($num,8);
+ &jnz(&label("aw_loop"));
+
+ &set_label("aw_finish",0);
+ &mov($num,&wparam(3)); # get num
+ &and($num,7);
+ &jz(&label("aw_end"));
+
+ for ($i=0; $i<7; $i++)
+ {
+ &comment("Tail Round $i");
+ &mov($tmp1,&DWP($i*4,$a,"",0)); # *a
+ &mov($tmp2,&DWP($i*4,$b,"",0));# *b
+ &add($tmp1,$c);
+ &mov($c,0);
+ &adc($c,$c);
+ &add($tmp1,$tmp2);
+ &adc($c,0);
+ &dec($num) if ($i != 6);
+ &mov(&DWP($i*4,$r,"",0),$tmp1); # *r
+ &jz(&label("aw_end")) if ($i != 6);
+ }
+ &set_label("aw_end",0);
+
+# &mov("eax",$c); # $c is "eax"
+
+ &function_end($name);
+ }
+
+sub bn_sub_words
+ {
+ local($name)=@_;
+
+ &function_begin($name,"");
+
+ &comment("");
+ $a="esi";
+ $b="edi";
+ $c="eax";
+ $r="ebx";
+ $tmp1="ecx";
+ $tmp2="edx";
+ $num="ebp";
+
+ &mov($r,&wparam(0)); # get r
+ &mov($a,&wparam(1)); # get a
+ &mov($b,&wparam(2)); # get b
+ &mov($num,&wparam(3)); # get num
+ &xor($c,$c); # clear carry
+ &and($num,0xfffffff8); # num / 8
+
+ &jz(&label("aw_finish"));
+
+ &set_label("aw_loop",0);
+ for ($i=0; $i<8; $i++)
+ {
+ &comment("Round $i");
+
+ &mov($tmp1,&DWP($i*4,$a,"",0)); # *a
+ &mov($tmp2,&DWP($i*4,$b,"",0)); # *b
+ &sub($tmp1,$c);
+ &mov($c,0);
+ &adc($c,$c);
+ &sub($tmp1,$tmp2);
+ &adc($c,0);
+ &mov(&DWP($i*4,$r,"",0),$tmp1); # *r
+ }
+
+ &comment("");
+ &add($a,32);
+ &add($b,32);
+ &add($r,32);
+ &sub($num,8);
+ &jnz(&label("aw_loop"));
+
+ &set_label("aw_finish",0);
+ &mov($num,&wparam(3)); # get num
+ &and($num,7);
+ &jz(&label("aw_end"));
+
+ for ($i=0; $i<7; $i++)
+ {
+ &comment("Tail Round $i");
+ &mov($tmp1,&DWP($i*4,$a,"",0)); # *a
+ &mov($tmp2,&DWP($i*4,$b,"",0));# *b
+ &sub($tmp1,$c);
+ &mov($c,0);
+ &adc($c,$c);
+ &sub($tmp1,$tmp2);
+ &adc($c,0);
+ &dec($num) if ($i != 6);
+ &mov(&DWP($i*4,$r,"",0),$tmp1); # *r
+ &jz(&label("aw_end")) if ($i != 6);
+ }
+ &set_label("aw_end",0);
+
+# &mov("eax",$c); # $c is "eax"
+
+ &function_end($name);
+ }
+
+sub bn_sub_part_words
+ {
+ local($name)=@_;
+
+ &function_begin($name,"");
+
+ &comment("");
+ $a="esi";
+ $b="edi";
+ $c="eax";
+ $r="ebx";
+ $tmp1="ecx";
+ $tmp2="edx";
+ $num="ebp";
+
+ &mov($r,&wparam(0)); # get r
+ &mov($a,&wparam(1)); # get a
+ &mov($b,&wparam(2)); # get b
+ &mov($num,&wparam(3)); # get num
+ &xor($c,$c); # clear carry
+ &and($num,0xfffffff8); # num / 8
+
+ &jz(&label("aw_finish"));
+
+ &set_label("aw_loop",0);
+ for ($i=0; $i<8; $i++)
+ {
+ &comment("Round $i");
+
+ &mov($tmp1,&DWP($i*4,$a,"",0)); # *a
+ &mov($tmp2,&DWP($i*4,$b,"",0)); # *b
+ &sub($tmp1,$c);
+ &mov($c,0);
+ &adc($c,$c);
+ &sub($tmp1,$tmp2);
+ &adc($c,0);
+ &mov(&DWP($i*4,$r,"",0),$tmp1); # *r
+ }
+
+ &comment("");
+ &add($a,32);
+ &add($b,32);
+ &add($r,32);
+ &sub($num,8);
+ &jnz(&label("aw_loop"));
+
+ &set_label("aw_finish",0);
+ &mov($num,&wparam(3)); # get num
+ &and($num,7);
+ &jz(&label("aw_end"));
+
+ for ($i=0; $i<7; $i++)
+ {
+ &comment("Tail Round $i");
+ &mov($tmp1,&DWP(0,$a,"",0)); # *a
+ &mov($tmp2,&DWP(0,$b,"",0));# *b
+ &sub($tmp1,$c);
+ &mov($c,0);
+ &adc($c,$c);
+ &sub($tmp1,$tmp2);
+ &adc($c,0);
+ &mov(&DWP(0,$r,"",0),$tmp1); # *r
+ &add($a, 4);
+ &add($b, 4);
+ &add($r, 4);
+ &dec($num) if ($i != 6);
+ &jz(&label("aw_end")) if ($i != 6);
+ }
+ &set_label("aw_end",0);
+
+ &cmp(&wparam(4),0);
+ &je(&label("pw_end"));
+
+ &mov($num,&wparam(4)); # get dl
+ &cmp($num,0);
+ &je(&label("pw_end"));
+ &jge(&label("pw_pos"));
+
+ &comment("pw_neg");
+ &mov($tmp2,0);
+ &sub($tmp2,$num);
+ &mov($num,$tmp2);
+ &and($num,0xfffffff8); # num / 8
+ &jz(&label("pw_neg_finish"));
+
+ &set_label("pw_neg_loop",0);
+ for ($i=0; $i<8; $i++)
+ {
+ &comment("dl<0 Round $i");
+
+ &mov($tmp1,0);
+ &mov($tmp2,&DWP($i*4,$b,"",0)); # *b
+ &sub($tmp1,$c);
+ &mov($c,0);
+ &adc($c,$c);
+ &sub($tmp1,$tmp2);
+ &adc($c,0);
+ &mov(&DWP($i*4,$r,"",0),$tmp1); # *r
+ }
+
+ &comment("");
+ &add($b,32);
+ &add($r,32);
+ &sub($num,8);
+ &jnz(&label("pw_neg_loop"));
+
+ &set_label("pw_neg_finish",0);
+ &mov($tmp2,&wparam(4)); # get dl
+ &mov($num,0);
+ &sub($num,$tmp2);
+ &and($num,7);
+ &jz(&label("pw_end"));
+
+ for ($i=0; $i<7; $i++)
+ {
+ &comment("dl<0 Tail Round $i");
+ &mov($tmp1,0);
+ &mov($tmp2,&DWP($i*4,$b,"",0));# *b
+ &sub($tmp1,$c);
+ &mov($c,0);
+ &adc($c,$c);
+ &sub($tmp1,$tmp2);
+ &adc($c,0);
+ &dec($num) if ($i != 6);
+ &mov(&DWP($i*4,$r,"",0),$tmp1); # *r
+ &jz(&label("pw_end")) if ($i != 6);
+ }
+
+ &jmp(&label("pw_end"));
+
+ &set_label("pw_pos",0);
+
+ &and($num,0xfffffff8); # num / 8
+ &jz(&label("pw_pos_finish"));
+
+ &set_label("pw_pos_loop",0);
+
+ for ($i=0; $i<8; $i++)
+ {
+ &comment("dl>0 Round $i");
+
+ &mov($tmp1,&DWP($i*4,$a,"",0)); # *a
+ &sub($tmp1,$c);
+ &mov(&DWP($i*4,$r,"",0),$tmp1); # *r
+ &jnc(&label("pw_nc".$i));
+ }
+
+ &comment("");
+ &add($a,32);
+ &add($r,32);
+ &sub($num,8);
+ &jnz(&label("pw_pos_loop"));
+
+ &set_label("pw_pos_finish",0);
+ &mov($num,&wparam(4)); # get dl
+ &and($num,7);
+ &jz(&label("pw_end"));
+
+ for ($i=0; $i<7; $i++)
+ {
+ &comment("dl>0 Tail Round $i");
+ &mov($tmp1,&DWP($i*4,$a,"",0)); # *a
+ &sub($tmp1,$c);
+ &mov(&DWP($i*4,$r,"",0),$tmp1); # *r
+ &jnc(&label("pw_tail_nc".$i));
+ &dec($num) if ($i != 6);
+ &jz(&label("pw_end")) if ($i != 6);
+ }
+ &mov($c,1);
+ &jmp(&label("pw_end"));
+
+ &set_label("pw_nc_loop",0);
+ for ($i=0; $i<8; $i++)
+ {
+ &mov($tmp1,&DWP($i*4,$a,"",0)); # *a
+ &mov(&DWP($i*4,$r,"",0),$tmp1); # *r
+ &set_label("pw_nc".$i,0);
+ }
+
+ &comment("");
+ &add($a,32);
+ &add($r,32);
+ &sub($num,8);
+ &jnz(&label("pw_nc_loop"));
+
+ &mov($num,&wparam(4)); # get dl
+ &and($num,7);
+ &jz(&label("pw_nc_end"));
+
+ for ($i=0; $i<7; $i++)
+ {
+ &mov($tmp1,&DWP($i*4,$a,"",0)); # *a
+ &mov(&DWP($i*4,$r,"",0),$tmp1); # *r
+ &set_label("pw_tail_nc".$i,0);
+ &dec($num) if ($i != 6);
+ &jz(&label("pw_nc_end")) if ($i != 6);
+ }
+
+ &set_label("pw_nc_end",0);
+ &mov($c,0);
+
+ &set_label("pw_end",0);
+
+# &mov("eax",$c); # $c is "eax"
+
+ &function_end($name);
+ }
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/bn-c64xplus.asm b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/bn-c64xplus.asm
new file mode 100644
index 0000000..de6d377
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/bn-c64xplus.asm
@@ -0,0 +1,382 @@
+;; Copyright 2012-2016 The OpenSSL Project Authors. All Rights Reserved.
+;;
+;; Licensed under the OpenSSL license (the "License"). You may not use
+;; this file except in compliance with the License. You can obtain a copy
+;; in the file LICENSE in the source distribution or at
+;; https://www.openssl.org/source/license.html
+;;
+;;====================================================================
+;; Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+;; project.
+;;
+;; Rights for redistribution and usage in source and binary forms are
+;; granted according to the OpenSSL license. Warranty of any kind is
+;; disclaimed.
+;;====================================================================
+;; Compiler-generated multiply-n-add SPLOOP runs at 12*n cycles, n
+;; being the number of 32-bit words, addition - 8*n. Corresponding 4x
+;; unrolled SPLOOP-free loops - at ~8*n and ~5*n. Below assembler
+;; SPLOOPs spin at ... 2*n cycles [plus epilogue].
+;;====================================================================
+ .text
+
+ .if .ASSEMBLER_VERSION<7000000
+ .asg 0,__TI_EABI__
+ .endif
+ .if __TI_EABI__
+ .asg bn_mul_add_words,_bn_mul_add_words
+ .asg bn_mul_words,_bn_mul_words
+ .asg bn_sqr_words,_bn_sqr_words
+ .asg bn_add_words,_bn_add_words
+ .asg bn_sub_words,_bn_sub_words
+ .asg bn_div_words,_bn_div_words
+ .asg bn_sqr_comba8,_bn_sqr_comba8
+ .asg bn_mul_comba8,_bn_mul_comba8
+ .asg bn_sqr_comba4,_bn_sqr_comba4
+ .asg bn_mul_comba4,_bn_mul_comba4
+ .endif
+
+ .asg B3,RA
+ .asg A4,ARG0
+ .asg B4,ARG1
+ .asg A6,ARG2
+ .asg B6,ARG3
+ .asg A8,ARG4
+ .asg B8,ARG5
+ .asg A4,RET
+ .asg A15,FP
+ .asg B14,DP
+ .asg B15,SP
+
+ .global _bn_mul_add_words
+_bn_mul_add_words:
+ .asmfunc
+ MV ARG2,B0
+ [!B0] BNOP RA
+||[!B0] MVK 0,RET
+ [B0] MVC B0,ILC
+ [B0] ZERO A19 ; high part of accumulator
+|| [B0] MV ARG0,A2
+|| [B0] MV ARG3,A3
+ NOP 3
+
+ SPLOOP 2 ; 2*n+10
+;;====================================================================
+ LDW *ARG1++,B7 ; ap[i]
+ NOP 3
+ LDW *ARG0++,A7 ; rp[i]
+ MPY32U B7,A3,A17:A16
+ NOP 3 ; [2,0] in epilogue
+ ADDU A16,A7,A21:A20
+ ADDU A19,A21:A20,A19:A18
+|| MV.S A17,A23
+ SPKERNEL 2,1 ; leave slot for "return value"
+|| STW A18,*A2++ ; rp[i]
+|| ADD A19,A23,A19
+;;====================================================================
+ BNOP RA,4
+ MV A19,RET ; return value
+ .endasmfunc
+
+ .global _bn_mul_words
+_bn_mul_words:
+ .asmfunc
+ MV ARG2,B0
+ [!B0] BNOP RA
+||[!B0] MVK 0,RET
+ [B0] MVC B0,ILC
+ [B0] ZERO A19 ; high part of accumulator
+ NOP 3
+
+ SPLOOP 2 ; 2*n+10
+;;====================================================================
+ LDW *ARG1++,A7 ; ap[i]
+ NOP 4
+ MPY32U A7,ARG3,A17:A16
+ NOP 4 ; [2,0] in epiloque
+ ADDU A19,A16,A19:A18
+|| MV.S A17,A21
+ SPKERNEL 2,1 ; leave slot for "return value"
+|| STW A18,*ARG0++ ; rp[i]
+|| ADD.L A19,A21,A19
+;;====================================================================
+ BNOP RA,4
+ MV A19,RET ; return value
+ .endasmfunc
+
+ .global _bn_sqr_words
+_bn_sqr_words:
+ .asmfunc
+ MV ARG2,B0
+ [!B0] BNOP RA
+||[!B0] MVK 0,RET
+ [B0] MVC B0,ILC
+ [B0] MV ARG0,B2
+|| [B0] ADD 4,ARG0,ARG0
+ NOP 3
+
+ SPLOOP 2 ; 2*n+10
+;;====================================================================
+ LDW *ARG1++,B7 ; ap[i]
+ NOP 4
+ MPY32U B7,B7,B1:B0
+ NOP 3 ; [2,0] in epilogue
+ STW B0,*B2++(8) ; rp[2*i]
+ MV B1,A1
+ SPKERNEL 2,0 ; fully overlap BNOP RA,5
+|| STW A1,*ARG0++(8) ; rp[2*i+1]
+;;====================================================================
+ BNOP RA,5
+ .endasmfunc
+
+ .global _bn_add_words
+_bn_add_words:
+ .asmfunc
+ MV ARG3,B0
+ [!B0] BNOP RA
+||[!B0] MVK 0,RET
+ [B0] MVC B0,ILC
+ [B0] ZERO A1 ; carry flag
+|| [B0] MV ARG0,A3
+ NOP 3
+
+ SPLOOP 2 ; 2*n+6
+;;====================================================================
+ LDW *ARG2++,A7 ; bp[i]
+|| LDW *ARG1++,B7 ; ap[i]
+ NOP 4
+ ADDU A7,B7,A9:A8
+ ADDU A1,A9:A8,A1:A0
+ SPKERNEL 0,0 ; fully overlap BNOP RA,5
+|| STW A0,*A3++ ; write result
+|| MV A1,RET ; keep carry flag in RET
+;;====================================================================
+ BNOP RA,5
+ .endasmfunc
+
+ .global _bn_sub_words
+_bn_sub_words:
+ .asmfunc
+ MV ARG3,B0
+ [!B0] BNOP RA
+||[!B0] MVK 0,RET
+ [B0] MVC B0,ILC
+ [B0] ZERO A2 ; borrow flag
+|| [B0] MV ARG0,A3
+ NOP 3
+
+ SPLOOP 2 ; 2*n+6
+;;====================================================================
+ LDW *ARG2++,A7 ; bp[i]
+|| LDW *ARG1++,B7 ; ap[i]
+ NOP 4
+ SUBU B7,A7,A1:A0
+ [A2] SUB A1:A0,1,A1:A0
+ SPKERNEL 0,1 ; leave slot for "return borrow flag"
+|| STW A0,*A3++ ; write result
+|| AND 1,A1,A2 ; pass on borrow flag
+;;====================================================================
+ BNOP RA,4
+ AND 1,A1,RET ; return borrow flag
+ .endasmfunc
+
+ .global _bn_div_words
+_bn_div_words:
+ .asmfunc
+ LMBD 1,A6,A0 ; leading zero bits in dv
+ LMBD 1,A4,A1 ; leading zero bits in hi
+|| MVK 32,B0
+ CMPLTU A1,A0,A2
+|| ADD A0,B0,B0
+ [ A2] BNOP RA
+||[ A2] MVK -1,A4 ; return overflow
+||[!A2] MV A4,A3 ; reassign hi
+ [!A2] MV B4,A4 ; reassign lo, will be quotient
+||[!A2] MVC B0,ILC
+ [!A2] SHL A6,A0,A6 ; normalize dv
+|| MVK 1,A1
+
+ [!A2] CMPLTU A3,A6,A1 ; hi<dv?
+||[!A2] SHL A4,1,A5:A4 ; lo<<1
+ [!A1] SUB A3,A6,A3 ; hi-=dv
+||[!A1] OR 1,A4,A4
+ [!A2] SHRU A3,31,A1 ; upper bit
+||[!A2] ADDAH A5,A3,A3 ; hi<<1|lo>>31
+
+ SPLOOP 3
+ [!A1] CMPLTU A3,A6,A1 ; hi<dv?
+||[ A1] ZERO A1
+|| SHL A4,1,A5:A4 ; lo<<1
+ [!A1] SUB A3,A6,A3 ; hi-=dv
+||[!A1] OR 1,A4,A4 ; quotient
+ SHRU A3,31,A1 ; upper bit
+|| ADDAH A5,A3,A3 ; hi<<1|lo>>31
+ SPKERNEL
+
+ BNOP RA,5
+ .endasmfunc
+
+;;====================================================================
+;; Not really Comba algorithm, just straightforward NxM... Dedicated
+;; fully unrolled real Comba implementations are asymptotically 2x
+;; faster, but naturally larger undertaking. Purpose of this exercise
+;; was rather to learn to master nested SPLOOPs...
+;;====================================================================
+ .global _bn_sqr_comba8
+ .global _bn_mul_comba8
+_bn_sqr_comba8:
+ MV ARG1,ARG2
+_bn_mul_comba8:
+ .asmfunc
+ MVK 8,B0 ; N, RILC
+|| MVK 8,A0 ; M, outer loop counter
+|| MV ARG1,A5 ; copy ap
+|| MV ARG0,B4 ; copy rp
+|| ZERO B19 ; high part of accumulator
+ MVC B0,RILC
+|| SUB B0,2,B1 ; N-2, initial ILC
+|| SUB B0,1,B2 ; const B2=N-1
+|| LDW *A5++,B6 ; ap[0]
+|| MV A0,A3 ; const A3=M
+sploopNxM?: ; for best performance arrange M<=N
+ [A0] SPLOOPD 2 ; 2*n+10
+|| MVC B1,ILC
+|| ADDAW B4,B0,B5
+|| ZERO B7
+|| LDW *A5++,A9 ; pre-fetch ap[1]
+|| ZERO A1
+|| SUB A0,1,A0
+;;====================================================================
+;; SPLOOP from bn_mul_add_words, but with flipped A<>B register files.
+;; This is because of Advisory 15 from TI publication SPRZ247I.
+ LDW *ARG2++,A7 ; bp[i]
+ NOP 3
+ [A1] LDW *B5++,B7 ; rp[i]
+ MPY32U A7,B6,B17:B16
+ NOP 3
+ ADDU B16,B7,B21:B20
+ ADDU B19,B21:B20,B19:B18
+|| MV.S B17,B23
+ SPKERNEL
+|| STW B18,*B4++ ; rp[i]
+|| ADD.S B19,B23,B19
+;;====================================================================
+outer?: ; m*2*(n+1)+10
+ SUBAW ARG2,A3,ARG2 ; rewind bp to bp[0]
+ SPMASKR
+|| CMPGT A0,1,A2 ; done pre-fetching ap[i+1]?
+ MVD A9,B6 ; move through .M unit(*)
+ [A2] LDW *A5++,A9 ; pre-fetch ap[i+1]
+ SUBAW B5,B2,B5 ; rewind rp to rp[1]
+ MVK 1,A1
+ [A0] BNOP.S1 outer?,4
+|| [A0] SUB.L A0,1,A0
+ STW B19,*B4--[B2] ; rewind rp tp rp[1]
+|| ZERO.S B19 ; high part of accumulator
+;; end of outer?
+ BNOP RA,5 ; return
+ .endasmfunc
+;; (*) It should be noted that B6 is used as input to MPY32U in
+;; chronologically next cycle in *preceding* SPLOOP iteration.
+;; Normally such arrangement would require DINT, but at this
+;; point SPLOOP is draining and interrupts are disabled
+;; implicitly.
+
+ .global _bn_sqr_comba4
+ .global _bn_mul_comba4
+_bn_sqr_comba4:
+ MV ARG1,ARG2
+_bn_mul_comba4:
+ .asmfunc
+ .if 0
+ BNOP sploopNxM?,3
+ ;; Above mentioned m*2*(n+1)+10 does not apply in n=m=4 case,
+ ;; because of low-counter effect, when prologue phase finishes
+ ;; before SPKERNEL instruction is reached. As result it's 25%
+ ;; slower than expected...
+ MVK 4,B0 ; N, RILC
+|| MVK 4,A0 ; M, outer loop counter
+|| MV ARG1,A5 ; copy ap
+|| MV ARG0,B4 ; copy rp
+|| ZERO B19 ; high part of accumulator
+ MVC B0,RILC
+|| SUB B0,2,B1 ; first ILC
+|| SUB B0,1,B2 ; const B2=N-1
+|| LDW *A5++,B6 ; ap[0]
+|| MV A0,A3 ; const A3=M
+ .else
+ ;; This alternative is an exercise in fully unrolled Comba
+ ;; algorithm implementation that operates at n*(n+1)+12, or
+ ;; as little as 32 cycles...
+ LDW *ARG1[0],B16 ; a[0]
+|| LDW *ARG2[0],A16 ; b[0]
+ LDW *ARG1[1],B17 ; a[1]
+|| LDW *ARG2[1],A17 ; b[1]
+ LDW *ARG1[2],B18 ; a[2]
+|| LDW *ARG2[2],A18 ; b[2]
+ LDW *ARG1[3],B19 ; a[3]
+|| LDW *ARG2[3],A19 ; b[3]
+ NOP
+ MPY32U A16,B16,A1:A0 ; a[0]*b[0]
+ MPY32U A17,B16,A23:A22 ; a[0]*b[1]
+ MPY32U A16,B17,A25:A24 ; a[1]*b[0]
+ MPY32U A16,B18,A27:A26 ; a[2]*b[0]
+ STW A0,*ARG0[0]
+|| MPY32U A17,B17,A29:A28 ; a[1]*b[1]
+ MPY32U A18,B16,A31:A30 ; a[0]*b[2]
+|| ADDU A22,A1,A1:A0
+ MV A23,B0
+|| MPY32U A19,B16,A21:A20 ; a[3]*b[0]
+|| ADDU A24,A1:A0,A1:A0
+ ADDU A25,B0,B1:B0
+|| STW A0,*ARG0[1]
+|| MPY32U A18,B17,A23:A22 ; a[2]*b[1]
+|| ADDU A26,A1,A9:A8
+ ADDU A27,B1,B9:B8
+|| MPY32U A17,B18,A25:A24 ; a[1]*b[2]
+|| ADDU A28,A9:A8,A9:A8
+ ADDU A29,B9:B8,B9:B8
+|| MPY32U A16,B19,A27:A26 ; a[0]*b[3]
+|| ADDU A30,A9:A8,A9:A8
+ ADDU A31,B9:B8,B9:B8
+|| ADDU B0,A9:A8,A9:A8
+ STW A8,*ARG0[2]
+|| ADDU A20,A9,A1:A0
+ ADDU A21,B9,B1:B0
+|| MPY32U A19,B17,A21:A20 ; a[3]*b[1]
+|| ADDU A22,A1:A0,A1:A0
+ ADDU A23,B1:B0,B1:B0
+|| MPY32U A18,B18,A23:A22 ; a[2]*b[2]
+|| ADDU A24,A1:A0,A1:A0
+ ADDU A25,B1:B0,B1:B0
+|| MPY32U A17,B19,A25:A24 ; a[1]*b[3]
+|| ADDU A26,A1:A0,A1:A0
+ ADDU A27,B1:B0,B1:B0
+|| ADDU B8,A1:A0,A1:A0
+ STW A0,*ARG0[3]
+|| MPY32U A19,B18,A27:A26 ; a[3]*b[2]
+|| ADDU A20,A1,A9:A8
+ ADDU A21,B1,B9:B8
+|| MPY32U A18,B19,A29:A28 ; a[2]*b[3]
+|| ADDU A22,A9:A8,A9:A8
+ ADDU A23,B9:B8,B9:B8
+|| MPY32U A19,B19,A31:A30 ; a[3]*b[3]
+|| ADDU A24,A9:A8,A9:A8
+ ADDU A25,B9:B8,B9:B8
+|| ADDU B0,A9:A8,A9:A8
+ STW A8,*ARG0[4]
+|| ADDU A26,A9,A1:A0
+ ADDU A27,B9,B1:B0
+|| ADDU A28,A1:A0,A1:A0
+ ADDU A29,B1:B0,B1:B0
+|| BNOP RA
+|| ADDU B8,A1:A0,A1:A0
+ STW A0,*ARG0[5]
+|| ADDU A30,A1,A9:A8
+ ADD A31,B1,B8
+ ADDU B0,A9:A8,A9:A8 ; removed || to avoid cross-path stall below
+ ADD B8,A9,A9
+|| STW A8,*ARG0[6]
+ STW A9,*ARG0[7]
+ .endif
+ .endasmfunc
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/c64xplus-gf2m.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/c64xplus-gf2m.pl
new file mode 100644
index 0000000..3bb8d12
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/c64xplus-gf2m.pl
@@ -0,0 +1,160 @@
+#! /usr/bin/env perl
+# Copyright 2012-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+#
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+#
+# February 2012
+#
+# The module implements bn_GF2m_mul_2x2 polynomial multiplication
+# used in bn_gf2m.c. It's kind of low-hanging mechanical port from
+# C for the time being... The subroutine runs in 37 cycles, which is
+# 4.5x faster than compiler-generated code. Though comparison is
+# totally unfair, because this module utilizes Galois Field Multiply
+# instruction.
+
+while (($output=shift) && ($output!~/\w[\w\-]*\.\w+$/)) {}
+open STDOUT,">$output";
+
+($rp,$a1,$a0,$b1,$b0)=("A4","B4","A6","B6","A8"); # argument vector
+
+($Alo,$Alox0,$Alox1,$Alox2,$Alox3)=map("A$_",(16..20));
+($Ahi,$Ahix0,$Ahix1,$Ahix2,$Ahix3)=map("B$_",(16..20));
+($B_0,$B_1,$B_2,$B_3)=("B5","A5","A7","B7");
+($A,$B)=($Alo,$B_1);
+$xFF="B1";
+
+sub mul_1x1_upper {
+my ($A,$B)=@_;
+$code.=<<___;
+ EXTU $B,8,24,$B_2 ; smash $B to 4 bytes
+|| AND $B,$xFF,$B_0
+|| SHRU $B,24,$B_3
+ SHRU $A,16, $Ahi ; smash $A to two halfwords
+|| EXTU $A,16,16,$Alo
+
+ XORMPY $Alo,$B_2,$Alox2 ; 16x8 bits multiplication
+|| XORMPY $Ahi,$B_2,$Ahix2
+|| EXTU $B,16,24,$B_1
+ XORMPY $Alo,$B_0,$Alox0
+|| XORMPY $Ahi,$B_0,$Ahix0
+ XORMPY $Alo,$B_3,$Alox3
+|| XORMPY $Ahi,$B_3,$Ahix3
+ XORMPY $Alo,$B_1,$Alox1
+|| XORMPY $Ahi,$B_1,$Ahix1
+___
+}
+sub mul_1x1_merged {
+my ($OUTlo,$OUThi,$A,$B)=@_;
+$code.=<<___;
+ EXTU $B,8,24,$B_2 ; smash $B to 4 bytes
+|| AND $B,$xFF,$B_0
+|| SHRU $B,24,$B_3
+ SHRU $A,16, $Ahi ; smash $A to two halfwords
+|| EXTU $A,16,16,$Alo
+
+ XOR $Ahix0,$Alox2,$Ahix0
+|| MV $Ahix2,$OUThi
+|| XORMPY $Alo,$B_2,$Alox2
+ XORMPY $Ahi,$B_2,$Ahix2
+|| EXTU $B,16,24,$B_1
+|| XORMPY $Alo,$B_0,A1 ; $Alox0
+ XOR $Ahix1,$Alox3,$Ahix1
+|| SHL $Ahix0,16,$OUTlo
+|| SHRU $Ahix0,16,$Ahix0
+ XOR $Alox0,$OUTlo,$OUTlo
+|| XOR $Ahix0,$OUThi,$OUThi
+|| XORMPY $Ahi,$B_0,$Ahix0
+|| XORMPY $Alo,$B_3,$Alox3
+|| SHL $Alox1,8,$Alox1
+|| SHL $Ahix3,8,$Ahix3
+ XOR $Alox1,$OUTlo,$OUTlo
+|| XOR $Ahix3,$OUThi,$OUThi
+|| XORMPY $Ahi,$B_3,$Ahix3
+|| SHL $Ahix1,24,$Alox1
+|| SHRU $Ahix1,8, $Ahix1
+ XOR $Alox1,$OUTlo,$OUTlo
+|| XOR $Ahix1,$OUThi,$OUThi
+|| XORMPY $Alo,$B_1,$Alox1
+|| XORMPY $Ahi,$B_1,$Ahix1
+|| MV A1,$Alox0
+___
+}
+sub mul_1x1_lower {
+my ($OUTlo,$OUThi)=@_;
+$code.=<<___;
+ ;NOP
+ XOR $Ahix0,$Alox2,$Ahix0
+|| MV $Ahix2,$OUThi
+ NOP
+ XOR $Ahix1,$Alox3,$Ahix1
+|| SHL $Ahix0,16,$OUTlo
+|| SHRU $Ahix0,16,$Ahix0
+ XOR $Alox0,$OUTlo,$OUTlo
+|| XOR $Ahix0,$OUThi,$OUThi
+|| SHL $Alox1,8,$Alox1
+|| SHL $Ahix3,8,$Ahix3
+ XOR $Alox1,$OUTlo,$OUTlo
+|| XOR $Ahix3,$OUThi,$OUThi
+|| SHL $Ahix1,24,$Alox1
+|| SHRU $Ahix1,8, $Ahix1
+ XOR $Alox1,$OUTlo,$OUTlo
+|| XOR $Ahix1,$OUThi,$OUThi
+___
+}
+$code.=<<___;
+ .text
+
+ .if .ASSEMBLER_VERSION<7000000
+ .asg 0,__TI_EABI__
+ .endif
+ .if __TI_EABI__
+ .asg bn_GF2m_mul_2x2,_bn_GF2m_mul_2x2
+ .endif
+
+ .global _bn_GF2m_mul_2x2
+_bn_GF2m_mul_2x2:
+ .asmfunc
+ MVK 0xFF,$xFF
+___
+ &mul_1x1_upper($a0,$b0); # a0·b0
+$code.=<<___;
+|| MV $b1,$B
+ MV $a1,$A
+___
+ &mul_1x1_merged("A28","B28",$A,$B); # a0·b0/a1·b1
+$code.=<<___;
+|| XOR $b0,$b1,$B
+ XOR $a0,$a1,$A
+___
+ &mul_1x1_merged("A31","B31",$A,$B); # a1·b1/(a0+a1)·(b0+b1)
+$code.=<<___;
+ XOR A28,A31,A29
+|| XOR B28,B31,B29 ; a0·b0+a1·b1
+___
+ &mul_1x1_lower("A30","B30"); # (a0+a1)·(b0+b1)
+$code.=<<___;
+|| BNOP B3
+ XOR A29,A30,A30
+|| XOR B29,B30,B30 ; (a0+a1)·(b0+b1)-a0·b0-a1·b1
+ XOR B28,A30,A30
+|| STW A28,*${rp}[0]
+ XOR B30,A31,A31
+|| STW A30,*${rp}[1]
+ STW A31,*${rp}[2]
+ STW B31,*${rp}[3]
+ .endasmfunc
+___
+
+print $code;
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/co-586.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/co-586.pl
new file mode 100644
index 0000000..3c34fa8
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/co-586.pl
@@ -0,0 +1,298 @@
+#! /usr/bin/env perl
+# Copyright 1995-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+push(@INC,"${dir}","${dir}../../perlasm");
+require "x86asm.pl";
+
+$output = pop;
+open STDOUT,">$output";
+
+&asm_init($ARGV[0]);
+
+&bn_mul_comba("bn_mul_comba8",8);
+&bn_mul_comba("bn_mul_comba4",4);
+&bn_sqr_comba("bn_sqr_comba8",8);
+&bn_sqr_comba("bn_sqr_comba4",4);
+
+&asm_finish();
+
+close STDOUT or die "error closing STDOUT: $!";
+
+sub mul_add_c
+ {
+ local($a,$ai,$b,$bi,$c0,$c1,$c2,$pos,$i,$na,$nb)=@_;
+
+ # pos == -1 if eax and edx are pre-loaded, 0 to load from next
+ # words, and 1 if load return value
+
+ &comment("mul a[$ai]*b[$bi]");
+
+ # "eax" and "edx" will always be pre-loaded.
+ # &mov("eax",&DWP($ai*4,$a,"",0)) ;
+ # &mov("edx",&DWP($bi*4,$b,"",0));
+
+ &mul("edx");
+ &add($c0,"eax");
+ &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 0; # load next a
+ &mov("eax",&wparam(0)) if $pos > 0; # load r[]
+ ###
+ &adc($c1,"edx");
+ &mov("edx",&DWP(($nb)*4,$b,"",0)) if $pos == 0; # load next b
+ &mov("edx",&DWP(($nb)*4,$b,"",0)) if $pos == 1; # load next b
+ ###
+ &adc($c2,0);
+ # is pos > 1, it means it is the last loop
+ &mov(&DWP($i*4,"eax","",0),$c0) if $pos > 0; # save r[];
+ &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 1; # load next a
+ }
+
+sub sqr_add_c
+ {
+ local($r,$a,$ai,$bi,$c0,$c1,$c2,$pos,$i,$na,$nb)=@_;
+
+ # pos == -1 if eax and edx are pre-loaded, 0 to load from next
+ # words, and 1 if load return value
+
+ &comment("sqr a[$ai]*a[$bi]");
+
+ # "eax" and "edx" will always be pre-loaded.
+ # &mov("eax",&DWP($ai*4,$a,"",0)) ;
+ # &mov("edx",&DWP($bi*4,$b,"",0));
+
+ if ($ai == $bi)
+ { &mul("eax");}
+ else
+ { &mul("edx");}
+ &add($c0,"eax");
+ &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 0; # load next a
+ ###
+ &adc($c1,"edx");
+ &mov("edx",&DWP(($nb)*4,$a,"",0)) if ($pos == 1) && ($na != $nb);
+ ###
+ &adc($c2,0);
+ # is pos > 1, it means it is the last loop
+ &mov(&DWP($i*4,$r,"",0),$c0) if $pos > 0; # save r[];
+ &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 1; # load next b
+ }
+
+sub sqr_add_c2
+ {
+ local($r,$a,$ai,$bi,$c0,$c1,$c2,$pos,$i,$na,$nb)=@_;
+
+ # pos == -1 if eax and edx are pre-loaded, 0 to load from next
+ # words, and 1 if load return value
+
+ &comment("sqr a[$ai]*a[$bi]");
+
+ # "eax" and "edx" will always be pre-loaded.
+ # &mov("eax",&DWP($ai*4,$a,"",0)) ;
+ # &mov("edx",&DWP($bi*4,$a,"",0));
+
+ if ($ai == $bi)
+ { &mul("eax");}
+ else
+ { &mul("edx");}
+ &add("eax","eax");
+ ###
+ &adc("edx","edx");
+ ###
+ &adc($c2,0);
+ &add($c0,"eax");
+ &adc($c1,"edx");
+ &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 0; # load next a
+ &mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 1; # load next b
+ &adc($c2,0);
+ &mov(&DWP($i*4,$r,"",0),$c0) if $pos > 0; # save r[];
+ &mov("edx",&DWP(($nb)*4,$a,"",0)) if ($pos <= 1) && ($na != $nb);
+ ###
+ }
+
+sub bn_mul_comba
+ {
+ local($name,$num)=@_;
+ local($a,$b,$c0,$c1,$c2);
+ local($i,$as,$ae,$bs,$be,$ai,$bi);
+ local($tot,$end);
+
+ &function_begin_B($name,"");
+
+ $c0="ebx";
+ $c1="ecx";
+ $c2="ebp";
+ $a="esi";
+ $b="edi";
+
+ $as=0;
+ $ae=0;
+ $bs=0;
+ $be=0;
+ $tot=$num+$num-1;
+
+ &push("esi");
+ &mov($a,&wparam(1));
+ &push("edi");
+ &mov($b,&wparam(2));
+ &push("ebp");
+ &push("ebx");
+
+ &xor($c0,$c0);
+ &mov("eax",&DWP(0,$a,"",0)); # load the first word
+ &xor($c1,$c1);
+ &mov("edx",&DWP(0,$b,"",0)); # load the first second
+
+ for ($i=0; $i<$tot; $i++)
+ {
+ $ai=$as;
+ $bi=$bs;
+ $end=$be+1;
+
+ &comment("################## Calculate word $i");
+
+ for ($j=$bs; $j<$end; $j++)
+ {
+ &xor($c2,$c2) if ($j == $bs);
+ if (($j+1) == $end)
+ {
+ $v=1;
+ $v=2 if (($i+1) == $tot);
+ }
+ else
+ { $v=0; }
+ if (($j+1) != $end)
+ {
+ $na=($ai-1);
+ $nb=($bi+1);
+ }
+ else
+ {
+ $na=$as+($i < ($num-1));
+ $nb=$bs+($i >= ($num-1));
+ }
+#printf STDERR "[$ai,$bi] -> [$na,$nb]\n";
+ &mul_add_c($a,$ai,$b,$bi,$c0,$c1,$c2,$v,$i,$na,$nb);
+ if ($v)
+ {
+ &comment("saved r[$i]");
+ # &mov("eax",&wparam(0));
+ # &mov(&DWP($i*4,"eax","",0),$c0);
+ ($c0,$c1,$c2)=($c1,$c2,$c0);
+ }
+ $ai--;
+ $bi++;
+ }
+ $as++ if ($i < ($num-1));
+ $ae++ if ($i >= ($num-1));
+
+ $bs++ if ($i >= ($num-1));
+ $be++ if ($i < ($num-1));
+ }
+ &comment("save r[$i]");
+ # &mov("eax",&wparam(0));
+ &mov(&DWP($i*4,"eax","",0),$c0);
+
+ &pop("ebx");
+ &pop("ebp");
+ &pop("edi");
+ &pop("esi");
+ &ret();
+ &function_end_B($name);
+ }
+
+sub bn_sqr_comba
+ {
+ local($name,$num)=@_;
+ local($r,$a,$c0,$c1,$c2)=@_;
+ local($i,$as,$ae,$bs,$be,$ai,$bi);
+ local($b,$tot,$end,$half);
+
+ &function_begin_B($name,"");
+
+ $c0="ebx";
+ $c1="ecx";
+ $c2="ebp";
+ $a="esi";
+ $r="edi";
+
+ &push("esi");
+ &push("edi");
+ &push("ebp");
+ &push("ebx");
+ &mov($r,&wparam(0));
+ &mov($a,&wparam(1));
+ &xor($c0,$c0);
+ &xor($c1,$c1);
+ &mov("eax",&DWP(0,$a,"",0)); # load the first word
+
+ $as=0;
+ $ae=0;
+ $bs=0;
+ $be=0;
+ $tot=$num+$num-1;
+
+ for ($i=0; $i<$tot; $i++)
+ {
+ $ai=$as;
+ $bi=$bs;
+ $end=$be+1;
+
+ &comment("############### Calculate word $i");
+ for ($j=$bs; $j<$end; $j++)
+ {
+ &xor($c2,$c2) if ($j == $bs);
+ if (($ai-1) < ($bi+1))
+ {
+ $v=1;
+ $v=2 if ($i+1) == $tot;
+ }
+ else
+ { $v=0; }
+ if (!$v)
+ {
+ $na=$ai-1;
+ $nb=$bi+1;
+ }
+ else
+ {
+ $na=$as+($i < ($num-1));
+ $nb=$bs+($i >= ($num-1));
+ }
+ if ($ai == $bi)
+ {
+ &sqr_add_c($r,$a,$ai,$bi,
+ $c0,$c1,$c2,$v,$i,$na,$nb);
+ }
+ else
+ {
+ &sqr_add_c2($r,$a,$ai,$bi,
+ $c0,$c1,$c2,$v,$i,$na,$nb);
+ }
+ if ($v)
+ {
+ &comment("saved r[$i]");
+ #&mov(&DWP($i*4,$r,"",0),$c0);
+ ($c0,$c1,$c2)=($c1,$c2,$c0);
+ last;
+ }
+ $ai--;
+ $bi++;
+ }
+ $as++ if ($i < ($num-1));
+ $ae++ if ($i >= ($num-1));
+
+ $bs++ if ($i >= ($num-1));
+ $be++ if ($i < ($num-1));
+ }
+ &mov(&DWP($i*4,$r,"",0),$c0);
+ &pop("ebx");
+ &pop("ebp");
+ &pop("edi");
+ &pop("esi");
+ &ret();
+ &function_end_B($name);
+ }
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/ia64-mont.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/ia64-mont.pl
new file mode 100644
index 0000000..7a4e74d
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/ia64-mont.pl
@@ -0,0 +1,860 @@
+#! /usr/bin/env perl
+# Copyright 2010-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+#
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+
+# January 2010
+#
+# "Teaser" Montgomery multiplication module for IA-64. There are
+# several possibilities for improvement:
+#
+# - modulo-scheduling outer loop would eliminate quite a number of
+# stalls after ldf8, xma and getf.sig outside inner loop and
+# improve shorter key performance;
+# - shorter vector support [with input vectors being fetched only
+# once] should be added;
+# - 2x unroll with help of n0[1] would make the code scalable on
+# "wider" IA-64, "wider" than Itanium 2 that is, which is not of
+# acute interest, because upcoming Tukwila's individual cores are
+# reportedly based on Itanium 2 design;
+# - dedicated squaring procedure(?);
+#
+# January 2010
+#
+# Shorter vector support is implemented by zero-padding ap and np
+# vectors up to 8 elements, or 512 bits. This means that 256-bit
+# inputs will be processed only 2 times faster than 512-bit inputs,
+# not 4 [as one would expect, because algorithm complexity is n^2].
+# The reason for padding is that inputs shorter than 512 bits won't
+# be processed faster anyway, because minimal critical path of the
+# core loop happens to match 512-bit timing. Either way, it resulted
+# in >100% improvement of 512-bit RSA sign benchmark and 50% - of
+# 1024-bit one [in comparison to original version of *this* module].
+#
+# So far 'openssl speed rsa dsa' output on 900MHz Itanium 2 *with*
+# this module is:
+# sign verify sign/s verify/s
+# rsa 512 bits 0.000290s 0.000024s 3452.8 42031.4
+# rsa 1024 bits 0.000793s 0.000058s 1261.7 17172.0
+# rsa 2048 bits 0.005908s 0.000148s 169.3 6754.0
+# rsa 4096 bits 0.033456s 0.000469s 29.9 2133.6
+# dsa 512 bits 0.000253s 0.000198s 3949.9 5057.0
+# dsa 1024 bits 0.000585s 0.000607s 1708.4 1647.4
+# dsa 2048 bits 0.001453s 0.001703s 688.1 587.4
+#
+# ... and *without* (but still with ia64.S):
+#
+# rsa 512 bits 0.000670s 0.000041s 1491.8 24145.5
+# rsa 1024 bits 0.001988s 0.000080s 502.9 12499.3
+# rsa 2048 bits 0.008702s 0.000189s 114.9 5293.9
+# rsa 4096 bits 0.043860s 0.000533s 22.8 1875.9
+# dsa 512 bits 0.000441s 0.000427s 2265.3 2340.6
+# dsa 1024 bits 0.000823s 0.000867s 1215.6 1153.2
+# dsa 2048 bits 0.001894s 0.002179s 528.1 458.9
+#
+# As it can be seen, RSA sign performance improves by 130-30%,
+# hereafter less for longer keys, while verify - by 74-13%.
+# DSA performance improves by 115-30%.
+
+$output=pop;
+
+if ($^O eq "hpux") {
+ $ADDP="addp4";
+ for (@ARGV) { $ADDP="add" if (/[\+DD|\-mlp]64/); }
+} else { $ADDP="add"; }
+
+$code=<<___;
+.explicit
+.text
+
+// int bn_mul_mont (BN_ULONG *rp,const BN_ULONG *ap,
+// const BN_ULONG *bp,const BN_ULONG *np,
+// const BN_ULONG *n0p,int num);
+.align 64
+.global bn_mul_mont#
+.proc bn_mul_mont#
+bn_mul_mont:
+ .prologue
+ .body
+{ .mmi; cmp4.le p6,p7=2,r37;;
+(p6) cmp4.lt.unc p8,p9=8,r37
+ mov ret0=r0 };;
+{ .bbb;
+(p9) br.cond.dptk.many bn_mul_mont_8
+(p8) br.cond.dpnt.many bn_mul_mont_general
+(p7) br.ret.spnt.many b0 };;
+.endp bn_mul_mont#
+
+prevfs=r2; prevpr=r3; prevlc=r10; prevsp=r11;
+
+rptr=r8; aptr=r9; bptr=r14; nptr=r15;
+tptr=r16; // &tp[0]
+tp_1=r17; // &tp[-1]
+num=r18; len=r19; lc=r20;
+topbit=r21; // carry bit from tmp[num]
+
+n0=f6;
+m0=f7;
+bi=f8;
+
+.align 64
+.local bn_mul_mont_general#
+.proc bn_mul_mont_general#
+bn_mul_mont_general:
+ .prologue
+{ .mmi; .save ar.pfs,prevfs
+ alloc prevfs=ar.pfs,6,2,0,8
+ $ADDP aptr=0,in1
+ .save ar.lc,prevlc
+ mov prevlc=ar.lc }
+{ .mmi; .vframe prevsp
+ mov prevsp=sp
+ $ADDP bptr=0,in2
+ .save pr,prevpr
+ mov prevpr=pr };;
+
+ .body
+ .rotf alo[6],nlo[4],ahi[8],nhi[6]
+ .rotr a[3],n[3],t[2]
+
+{ .mmi; ldf8 bi=[bptr],8 // (*bp++)
+ ldf8 alo[4]=[aptr],16 // ap[0]
+ $ADDP r30=8,in1 };;
+{ .mmi; ldf8 alo[3]=[r30],16 // ap[1]
+ ldf8 alo[2]=[aptr],16 // ap[2]
+ $ADDP in4=0,in4 };;
+{ .mmi; ldf8 alo[1]=[r30] // ap[3]
+ ldf8 n0=[in4] // n0
+ $ADDP rptr=0,in0 }
+{ .mmi; $ADDP nptr=0,in3
+ mov r31=16
+ zxt4 num=in5 };;
+{ .mmi; ldf8 nlo[2]=[nptr],8 // np[0]
+ shladd len=num,3,r0
+ shladd r31=num,3,r31 };;
+{ .mmi; ldf8 nlo[1]=[nptr],8 // np[1]
+ add lc=-5,num
+ sub r31=sp,r31 };;
+{ .mfb; and sp=-16,r31 // alloca
+ xmpy.hu ahi[2]=alo[4],bi // ap[0]*bp[0]
+ nop.b 0 }
+{ .mfb; nop.m 0
+ xmpy.lu alo[4]=alo[4],bi
+ brp.loop.imp .L1st_ctop,.L1st_cend-16
+ };;
+{ .mfi; nop.m 0
+ xma.hu ahi[1]=alo[3],bi,ahi[2] // ap[1]*bp[0]
+ add tp_1=8,sp }
+{ .mfi; nop.m 0
+ xma.lu alo[3]=alo[3],bi,ahi[2]
+ mov pr.rot=0x20001f<<16
+ // ------^----- (p40) at first (p23)
+ // ----------^^ p[16:20]=1
+ };;
+{ .mfi; nop.m 0
+ xmpy.lu m0=alo[4],n0 // (ap[0]*bp[0])*n0
+ mov ar.lc=lc }
+{ .mfi; nop.m 0
+ fcvt.fxu.s1 nhi[1]=f0
+ mov ar.ec=8 };;
+
+.align 32
+.L1st_ctop:
+.pred.rel "mutex",p40,p42
+{ .mfi; (p16) ldf8 alo[0]=[aptr],8 // *(aptr++)
+ (p18) xma.hu ahi[0]=alo[2],bi,ahi[1]
+ (p40) add n[2]=n[2],a[2] } // (p23) }
+{ .mfi; (p18) ldf8 nlo[0]=[nptr],8 // *(nptr++)(p16)
+ (p18) xma.lu alo[2]=alo[2],bi,ahi[1]
+ (p42) add n[2]=n[2],a[2],1 };; // (p23)
+{ .mfi; (p21) getf.sig a[0]=alo[5]
+ (p20) xma.hu nhi[0]=nlo[2],m0,nhi[1]
+ (p42) cmp.leu p41,p39=n[2],a[2] } // (p23)
+{ .mfi; (p23) st8 [tp_1]=n[2],8
+ (p20) xma.lu nlo[2]=nlo[2],m0,nhi[1]
+ (p40) cmp.ltu p41,p39=n[2],a[2] } // (p23)
+{ .mmb; (p21) getf.sig n[0]=nlo[3]
+ (p16) nop.m 0
+ br.ctop.sptk .L1st_ctop };;
+.L1st_cend:
+
+{ .mmi; getf.sig a[0]=ahi[6] // (p24)
+ getf.sig n[0]=nhi[4]
+ add num=-1,num };; // num--
+{ .mmi; .pred.rel "mutex",p40,p42
+(p40) add n[0]=n[0],a[0]
+(p42) add n[0]=n[0],a[0],1
+ sub aptr=aptr,len };; // rewind
+{ .mmi; .pred.rel "mutex",p40,p42
+(p40) cmp.ltu p41,p39=n[0],a[0]
+(p42) cmp.leu p41,p39=n[0],a[0]
+ sub nptr=nptr,len };;
+{ .mmi; .pred.rel "mutex",p39,p41
+(p39) add topbit=r0,r0
+(p41) add topbit=r0,r0,1
+ nop.i 0 }
+{ .mmi; st8 [tp_1]=n[0]
+ add tptr=16,sp
+ add tp_1=8,sp };;
+
+.Louter:
+{ .mmi; ldf8 bi=[bptr],8 // (*bp++)
+ ldf8 ahi[3]=[tptr] // tp[0]
+ add r30=8,aptr };;
+{ .mmi; ldf8 alo[4]=[aptr],16 // ap[0]
+ ldf8 alo[3]=[r30],16 // ap[1]
+ add r31=8,nptr };;
+{ .mfb; ldf8 alo[2]=[aptr],16 // ap[2]
+ xma.hu ahi[2]=alo[4],bi,ahi[3] // ap[0]*bp[i]+tp[0]
+ brp.loop.imp .Linner_ctop,.Linner_cend-16
+ }
+{ .mfb; ldf8 alo[1]=[r30] // ap[3]
+ xma.lu alo[4]=alo[4],bi,ahi[3]
+ clrrrb.pr };;
+{ .mfi; ldf8 nlo[2]=[nptr],16 // np[0]
+ xma.hu ahi[1]=alo[3],bi,ahi[2] // ap[1]*bp[i]
+ nop.i 0 }
+{ .mfi; ldf8 nlo[1]=[r31] // np[1]
+ xma.lu alo[3]=alo[3],bi,ahi[2]
+ mov pr.rot=0x20101f<<16
+ // ------^----- (p40) at first (p23)
+ // --------^--- (p30) at first (p22)
+ // ----------^^ p[16:20]=1
+ };;
+{ .mfi; st8 [tptr]=r0 // tp[0] is already accounted
+ xmpy.lu m0=alo[4],n0 // (ap[0]*bp[i]+tp[0])*n0
+ mov ar.lc=lc }
+{ .mfi;
+ fcvt.fxu.s1 nhi[1]=f0
+ mov ar.ec=8 };;
+
+// This loop spins in 4*(n+7) ticks on Itanium 2 and should spin in
+// 7*(n+7) ticks on Itanium (the one codenamed Merced). Factor of 7
+// in latter case accounts for two-tick pipeline stall, which means
+// that its performance would be ~20% lower than optimal one. No
+// attempt was made to address this, because original Itanium is
+// hardly represented out in the wild...
+.align 32
+.Linner_ctop:
+.pred.rel "mutex",p40,p42
+.pred.rel "mutex",p30,p32
+{ .mfi; (p16) ldf8 alo[0]=[aptr],8 // *(aptr++)
+ (p18) xma.hu ahi[0]=alo[2],bi,ahi[1]
+ (p40) add n[2]=n[2],a[2] } // (p23)
+{ .mfi; (p16) nop.m 0
+ (p18) xma.lu alo[2]=alo[2],bi,ahi[1]
+ (p42) add n[2]=n[2],a[2],1 };; // (p23)
+{ .mfi; (p21) getf.sig a[0]=alo[5]
+ (p16) nop.f 0
+ (p40) cmp.ltu p41,p39=n[2],a[2] } // (p23)
+{ .mfi; (p21) ld8 t[0]=[tptr],8
+ (p16) nop.f 0
+ (p42) cmp.leu p41,p39=n[2],a[2] };; // (p23)
+{ .mfi; (p18) ldf8 nlo[0]=[nptr],8 // *(nptr++)
+ (p20) xma.hu nhi[0]=nlo[2],m0,nhi[1]
+ (p30) add a[1]=a[1],t[1] } // (p22)
+{ .mfi; (p16) nop.m 0
+ (p20) xma.lu nlo[2]=nlo[2],m0,nhi[1]
+ (p32) add a[1]=a[1],t[1],1 };; // (p22)
+{ .mmi; (p21) getf.sig n[0]=nlo[3]
+ (p16) nop.m 0
+ (p30) cmp.ltu p31,p29=a[1],t[1] } // (p22)
+{ .mmb; (p23) st8 [tp_1]=n[2],8
+ (p32) cmp.leu p31,p29=a[1],t[1] // (p22)
+ br.ctop.sptk .Linner_ctop };;
+.Linner_cend:
+
+{ .mmi; getf.sig a[0]=ahi[6] // (p24)
+ getf.sig n[0]=nhi[4]
+ nop.i 0 };;
+
+{ .mmi; .pred.rel "mutex",p31,p33
+(p31) add a[0]=a[0],topbit
+(p33) add a[0]=a[0],topbit,1
+ mov topbit=r0 };;
+{ .mfi; .pred.rel "mutex",p31,p33
+(p31) cmp.ltu p32,p30=a[0],topbit
+(p33) cmp.leu p32,p30=a[0],topbit
+ }
+{ .mfi; .pred.rel "mutex",p40,p42
+(p40) add n[0]=n[0],a[0]
+(p42) add n[0]=n[0],a[0],1
+ };;
+{ .mmi; .pred.rel "mutex",p44,p46
+(p40) cmp.ltu p41,p39=n[0],a[0]
+(p42) cmp.leu p41,p39=n[0],a[0]
+(p32) add topbit=r0,r0,1 }
+
+{ .mmi; st8 [tp_1]=n[0],8
+ cmp4.ne p6,p0=1,num
+ sub aptr=aptr,len };; // rewind
+{ .mmi; sub nptr=nptr,len
+(p41) add topbit=r0,r0,1
+ add tptr=16,sp }
+{ .mmb; add tp_1=8,sp
+ add num=-1,num // num--
+(p6) br.cond.sptk.many .Louter };;
+
+{ .mbb; add lc=4,lc
+ brp.loop.imp .Lsub_ctop,.Lsub_cend-16
+ clrrrb.pr };;
+{ .mii; nop.m 0
+ mov pr.rot=0x10001<<16
+ // ------^---- (p33) at first (p17)
+ mov ar.lc=lc }
+{ .mii; nop.m 0
+ mov ar.ec=3
+ nop.i 0 };;
+
+.Lsub_ctop:
+.pred.rel "mutex",p33,p35
+{ .mfi; (p16) ld8 t[0]=[tptr],8 // t=*(tp++)
+ (p16) nop.f 0
+ (p33) sub n[1]=t[1],n[1] } // (p17)
+{ .mfi; (p16) ld8 n[0]=[nptr],8 // n=*(np++)
+ (p16) nop.f 0
+ (p35) sub n[1]=t[1],n[1],1 };; // (p17)
+{ .mib; (p18) st8 [rptr]=n[2],8 // *(rp++)=r
+ (p33) cmp.gtu p34,p32=n[1],t[1] // (p17)
+ (p18) nop.b 0 }
+{ .mib; (p18) nop.m 0
+ (p35) cmp.geu p34,p32=n[1],t[1] // (p17)
+ br.ctop.sptk .Lsub_ctop };;
+.Lsub_cend:
+
+{ .mmb; .pred.rel "mutex",p34,p36
+(p34) sub topbit=topbit,r0 // (p19)
+(p36) sub topbit=topbit,r0,1
+ brp.loop.imp .Lcopy_ctop,.Lcopy_cend-16
+ }
+{ .mmb; sub rptr=rptr,len // rewind
+ sub tptr=tptr,len
+ clrrrb.pr };;
+{ .mmi; mov aptr=rptr
+ mov bptr=tptr
+ mov pr.rot=1<<16 };;
+{ .mii; cmp.eq p0,p6=topbit,r0
+ mov ar.lc=lc
+ mov ar.ec=2 };;
+
+.Lcopy_ctop:
+{ .mmi; (p16) ld8 a[0]=[aptr],8
+ (p16) ld8 t[0]=[bptr],8
+ (p6) mov a[1]=t[1] };; // (p17)
+{ .mmb; (p17) st8 [rptr]=a[1],8
+ (p17) st8 [tptr]=r0,8
+ br.ctop.sptk .Lcopy_ctop };;
+.Lcopy_cend:
+
+{ .mmi; mov ret0=1 // signal "handled"
+ rum 1<<5 // clear um.mfh
+ mov ar.lc=prevlc }
+{ .mib; .restore sp
+ mov sp=prevsp
+ mov pr=prevpr,0x1ffff
+ br.ret.sptk.many b0 };;
+.endp bn_mul_mont_general#
+
+a1=r16; a2=r17; a3=r18; a4=r19; a5=r20; a6=r21; a7=r22; a8=r23;
+n1=r24; n2=r25; n3=r26; n4=r27; n5=r28; n6=r29; n7=r30; n8=r31;
+t0=r15;
+
+ai0=f8; ai1=f9; ai2=f10; ai3=f11; ai4=f12; ai5=f13; ai6=f14; ai7=f15;
+ni0=f16; ni1=f17; ni2=f18; ni3=f19; ni4=f20; ni5=f21; ni6=f22; ni7=f23;
+
+.align 64
+.skip 48 // aligns loop body
+.local bn_mul_mont_8#
+.proc bn_mul_mont_8#
+bn_mul_mont_8:
+ .prologue
+{ .mmi; .save ar.pfs,prevfs
+ alloc prevfs=ar.pfs,6,2,0,8
+ .vframe prevsp
+ mov prevsp=sp
+ .save ar.lc,prevlc
+ mov prevlc=ar.lc }
+{ .mmi; add r17=-6*16,sp
+ add sp=-7*16,sp
+ .save pr,prevpr
+ mov prevpr=pr };;
+
+{ .mmi; .save.gf 0,0x10
+ stf.spill [sp]=f16,-16
+ .save.gf 0,0x20
+ stf.spill [r17]=f17,32
+ add r16=-5*16,prevsp};;
+{ .mmi; .save.gf 0,0x40
+ stf.spill [r16]=f18,32
+ .save.gf 0,0x80
+ stf.spill [r17]=f19,32
+ $ADDP aptr=0,in1 };;
+{ .mmi; .save.gf 0,0x100
+ stf.spill [r16]=f20,32
+ .save.gf 0,0x200
+ stf.spill [r17]=f21,32
+ $ADDP r29=8,in1 };;
+{ .mmi; .save.gf 0,0x400
+ stf.spill [r16]=f22
+ .save.gf 0,0x800
+ stf.spill [r17]=f23
+ $ADDP rptr=0,in0 };;
+
+ .body
+ .rotf bj[8],mj[2],tf[2],alo[10],ahi[10],nlo[10],nhi[10]
+ .rotr t[8]
+
+// load input vectors padding them to 8 elements
+{ .mmi; ldf8 ai0=[aptr],16 // ap[0]
+ ldf8 ai1=[r29],16 // ap[1]
+ $ADDP bptr=0,in2 }
+{ .mmi; $ADDP r30=8,in2
+ $ADDP nptr=0,in3
+ $ADDP r31=8,in3 };;
+{ .mmi; ldf8 bj[7]=[bptr],16 // bp[0]
+ ldf8 bj[6]=[r30],16 // bp[1]
+ cmp4.le p4,p5=3,in5 }
+{ .mmi; ldf8 ni0=[nptr],16 // np[0]
+ ldf8 ni1=[r31],16 // np[1]
+ cmp4.le p6,p7=4,in5 };;
+
+{ .mfi; (p4)ldf8 ai2=[aptr],16 // ap[2]
+ (p5)fcvt.fxu ai2=f0
+ cmp4.le p8,p9=5,in5 }
+{ .mfi; (p6)ldf8 ai3=[r29],16 // ap[3]
+ (p7)fcvt.fxu ai3=f0
+ cmp4.le p10,p11=6,in5 }
+{ .mfi; (p4)ldf8 bj[5]=[bptr],16 // bp[2]
+ (p5)fcvt.fxu bj[5]=f0
+ cmp4.le p12,p13=7,in5 }
+{ .mfi; (p6)ldf8 bj[4]=[r30],16 // bp[3]
+ (p7)fcvt.fxu bj[4]=f0
+ cmp4.le p14,p15=8,in5 }
+{ .mfi; (p4)ldf8 ni2=[nptr],16 // np[2]
+ (p5)fcvt.fxu ni2=f0
+ addp4 r28=-1,in5 }
+{ .mfi; (p6)ldf8 ni3=[r31],16 // np[3]
+ (p7)fcvt.fxu ni3=f0
+ $ADDP in4=0,in4 };;
+
+{ .mfi; ldf8 n0=[in4]
+ fcvt.fxu tf[1]=f0
+ nop.i 0 }
+
+{ .mfi; (p8)ldf8 ai4=[aptr],16 // ap[4]
+ (p9)fcvt.fxu ai4=f0
+ mov t[0]=r0 }
+{ .mfi; (p10)ldf8 ai5=[r29],16 // ap[5]
+ (p11)fcvt.fxu ai5=f0
+ mov t[1]=r0 }
+{ .mfi; (p8)ldf8 bj[3]=[bptr],16 // bp[4]
+ (p9)fcvt.fxu bj[3]=f0
+ mov t[2]=r0 }
+{ .mfi; (p10)ldf8 bj[2]=[r30],16 // bp[5]
+ (p11)fcvt.fxu bj[2]=f0
+ mov t[3]=r0 }
+{ .mfi; (p8)ldf8 ni4=[nptr],16 // np[4]
+ (p9)fcvt.fxu ni4=f0
+ mov t[4]=r0 }
+{ .mfi; (p10)ldf8 ni5=[r31],16 // np[5]
+ (p11)fcvt.fxu ni5=f0
+ mov t[5]=r0 };;
+
+{ .mfi; (p12)ldf8 ai6=[aptr],16 // ap[6]
+ (p13)fcvt.fxu ai6=f0
+ mov t[6]=r0 }
+{ .mfi; (p14)ldf8 ai7=[r29],16 // ap[7]
+ (p15)fcvt.fxu ai7=f0
+ mov t[7]=r0 }
+{ .mfi; (p12)ldf8 bj[1]=[bptr],16 // bp[6]
+ (p13)fcvt.fxu bj[1]=f0
+ mov ar.lc=r28 }
+{ .mfi; (p14)ldf8 bj[0]=[r30],16 // bp[7]
+ (p15)fcvt.fxu bj[0]=f0
+ mov ar.ec=1 }
+{ .mfi; (p12)ldf8 ni6=[nptr],16 // np[6]
+ (p13)fcvt.fxu ni6=f0
+ mov pr.rot=1<<16 }
+{ .mfb; (p14)ldf8 ni7=[r31],16 // np[7]
+ (p15)fcvt.fxu ni7=f0
+ brp.loop.imp .Louter_8_ctop,.Louter_8_cend-16
+ };;
+
+// The loop is scheduled for 32*n ticks on Itanium 2. Actual attempt
+// to measure with help of Interval Time Counter indicated that the
+// factor is a tad higher: 33 or 34, if not 35. Exact measurement and
+// addressing the issue is problematic, because I don't have access
+// to platform-specific instruction-level profiler. On Itanium it
+// should run in 56*n ticks, because of higher xma latency...
+.Louter_8_ctop:
+ .pred.rel "mutex",p40,p42
+ .pred.rel "mutex",p48,p50
+{ .mfi; (p16) nop.m 0 // 0:
+ (p16) xma.hu ahi[0]=ai0,bj[7],tf[1] // ap[0]*b[i]+t[0]
+ (p40) add a3=a3,n3 } // (p17) a3+=n3
+{ .mfi; (p42) add a3=a3,n3,1
+ (p16) xma.lu alo[0]=ai0,bj[7],tf[1]
+ (p16) nop.i 0 };;
+{ .mii; (p17) getf.sig a7=alo[8] // 1:
+ (p48) add t[6]=t[6],a3 // (p17) t[6]+=a3
+ (p50) add t[6]=t[6],a3,1 };;
+{ .mfi; (p17) getf.sig a8=ahi[8] // 2:
+ (p17) xma.hu nhi[7]=ni6,mj[1],nhi[6] // np[6]*m0
+ (p40) cmp.ltu p43,p41=a3,n3 }
+{ .mfi; (p42) cmp.leu p43,p41=a3,n3
+ (p17) xma.lu nlo[7]=ni6,mj[1],nhi[6]
+ (p16) nop.i 0 };;
+{ .mii; (p17) getf.sig n5=nlo[6] // 3:
+ (p48) cmp.ltu p51,p49=t[6],a3
+ (p50) cmp.leu p51,p49=t[6],a3 };;
+ .pred.rel "mutex",p41,p43
+ .pred.rel "mutex",p49,p51
+{ .mfi; (p16) nop.m 0 // 4:
+ (p16) xma.hu ahi[1]=ai1,bj[7],ahi[0] // ap[1]*b[i]
+ (p41) add a4=a4,n4 } // (p17) a4+=n4
+{ .mfi; (p43) add a4=a4,n4,1
+ (p16) xma.lu alo[1]=ai1,bj[7],ahi[0]
+ (p16) nop.i 0 };;
+{ .mfi; (p49) add t[5]=t[5],a4 // 5: (p17) t[5]+=a4
+ (p16) xmpy.lu mj[0]=alo[0],n0 // (ap[0]*b[i]+t[0])*n0
+ (p51) add t[5]=t[5],a4,1 };;
+{ .mfi; (p16) nop.m 0 // 6:
+ (p17) xma.hu nhi[8]=ni7,mj[1],nhi[7] // np[7]*m0
+ (p41) cmp.ltu p42,p40=a4,n4 }
+{ .mfi; (p43) cmp.leu p42,p40=a4,n4
+ (p17) xma.lu nlo[8]=ni7,mj[1],nhi[7]
+ (p16) nop.i 0 };;
+{ .mii; (p17) getf.sig n6=nlo[7] // 7:
+ (p49) cmp.ltu p50,p48=t[5],a4
+ (p51) cmp.leu p50,p48=t[5],a4 };;
+ .pred.rel "mutex",p40,p42
+ .pred.rel "mutex",p48,p50
+{ .mfi; (p16) nop.m 0 // 8:
+ (p16) xma.hu ahi[2]=ai2,bj[7],ahi[1] // ap[2]*b[i]
+ (p40) add a5=a5,n5 } // (p17) a5+=n5
+{ .mfi; (p42) add a5=a5,n5,1
+ (p16) xma.lu alo[2]=ai2,bj[7],ahi[1]
+ (p16) nop.i 0 };;
+{ .mii; (p16) getf.sig a1=alo[1] // 9:
+ (p48) add t[4]=t[4],a5 // p(17) t[4]+=a5
+ (p50) add t[4]=t[4],a5,1 };;
+{ .mfi; (p16) nop.m 0 // 10:
+ (p16) xma.hu nhi[0]=ni0,mj[0],alo[0] // np[0]*m0
+ (p40) cmp.ltu p43,p41=a5,n5 }
+{ .mfi; (p42) cmp.leu p43,p41=a5,n5
+ (p16) xma.lu nlo[0]=ni0,mj[0],alo[0]
+ (p16) nop.i 0 };;
+{ .mii; (p17) getf.sig n7=nlo[8] // 11:
+ (p48) cmp.ltu p51,p49=t[4],a5
+ (p50) cmp.leu p51,p49=t[4],a5 };;
+ .pred.rel "mutex",p41,p43
+ .pred.rel "mutex",p49,p51
+{ .mfi; (p17) getf.sig n8=nhi[8] // 12:
+ (p16) xma.hu ahi[3]=ai3,bj[7],ahi[2] // ap[3]*b[i]
+ (p41) add a6=a6,n6 } // (p17) a6+=n6
+{ .mfi; (p43) add a6=a6,n6,1
+ (p16) xma.lu alo[3]=ai3,bj[7],ahi[2]
+ (p16) nop.i 0 };;
+{ .mii; (p16) getf.sig a2=alo[2] // 13:
+ (p49) add t[3]=t[3],a6 // (p17) t[3]+=a6
+ (p51) add t[3]=t[3],a6,1 };;
+{ .mfi; (p16) nop.m 0 // 14:
+ (p16) xma.hu nhi[1]=ni1,mj[0],nhi[0] // np[1]*m0
+ (p41) cmp.ltu p42,p40=a6,n6 }
+{ .mfi; (p43) cmp.leu p42,p40=a6,n6
+ (p16) xma.lu nlo[1]=ni1,mj[0],nhi[0]
+ (p16) nop.i 0 };;
+{ .mii; (p16) nop.m 0 // 15:
+ (p49) cmp.ltu p50,p48=t[3],a6
+ (p51) cmp.leu p50,p48=t[3],a6 };;
+ .pred.rel "mutex",p40,p42
+ .pred.rel "mutex",p48,p50
+{ .mfi; (p16) nop.m 0 // 16:
+ (p16) xma.hu ahi[4]=ai4,bj[7],ahi[3] // ap[4]*b[i]
+ (p40) add a7=a7,n7 } // (p17) a7+=n7
+{ .mfi; (p42) add a7=a7,n7,1
+ (p16) xma.lu alo[4]=ai4,bj[7],ahi[3]
+ (p16) nop.i 0 };;
+{ .mii; (p16) getf.sig a3=alo[3] // 17:
+ (p48) add t[2]=t[2],a7 // (p17) t[2]+=a7
+ (p50) add t[2]=t[2],a7,1 };;
+{ .mfi; (p16) nop.m 0 // 18:
+ (p16) xma.hu nhi[2]=ni2,mj[0],nhi[1] // np[2]*m0
+ (p40) cmp.ltu p43,p41=a7,n7 }
+{ .mfi; (p42) cmp.leu p43,p41=a7,n7
+ (p16) xma.lu nlo[2]=ni2,mj[0],nhi[1]
+ (p16) nop.i 0 };;
+{ .mii; (p16) getf.sig n1=nlo[1] // 19:
+ (p48) cmp.ltu p51,p49=t[2],a7
+ (p50) cmp.leu p51,p49=t[2],a7 };;
+ .pred.rel "mutex",p41,p43
+ .pred.rel "mutex",p49,p51
+{ .mfi; (p16) nop.m 0 // 20:
+ (p16) xma.hu ahi[5]=ai5,bj[7],ahi[4] // ap[5]*b[i]
+ (p41) add a8=a8,n8 } // (p17) a8+=n8
+{ .mfi; (p43) add a8=a8,n8,1
+ (p16) xma.lu alo[5]=ai5,bj[7],ahi[4]
+ (p16) nop.i 0 };;
+{ .mii; (p16) getf.sig a4=alo[4] // 21:
+ (p49) add t[1]=t[1],a8 // (p17) t[1]+=a8
+ (p51) add t[1]=t[1],a8,1 };;
+{ .mfi; (p16) nop.m 0 // 22:
+ (p16) xma.hu nhi[3]=ni3,mj[0],nhi[2] // np[3]*m0
+ (p41) cmp.ltu p42,p40=a8,n8 }
+{ .mfi; (p43) cmp.leu p42,p40=a8,n8
+ (p16) xma.lu nlo[3]=ni3,mj[0],nhi[2]
+ (p16) nop.i 0 };;
+{ .mii; (p16) getf.sig n2=nlo[2] // 23:
+ (p49) cmp.ltu p50,p48=t[1],a8
+ (p51) cmp.leu p50,p48=t[1],a8 };;
+{ .mfi; (p16) nop.m 0 // 24:
+ (p16) xma.hu ahi[6]=ai6,bj[7],ahi[5] // ap[6]*b[i]
+ (p16) add a1=a1,n1 } // (p16) a1+=n1
+{ .mfi; (p16) nop.m 0
+ (p16) xma.lu alo[6]=ai6,bj[7],ahi[5]
+ (p17) mov t[0]=r0 };;
+{ .mii; (p16) getf.sig a5=alo[5] // 25:
+ (p16) add t0=t[7],a1 // (p16) t[7]+=a1
+ (p42) add t[0]=t[0],r0,1 };;
+{ .mfi; (p16) setf.sig tf[0]=t0 // 26:
+ (p16) xma.hu nhi[4]=ni4,mj[0],nhi[3] // np[4]*m0
+ (p50) add t[0]=t[0],r0,1 }
+{ .mfi; (p16) cmp.ltu.unc p42,p40=a1,n1
+ (p16) xma.lu nlo[4]=ni4,mj[0],nhi[3]
+ (p16) nop.i 0 };;
+{ .mii; (p16) getf.sig n3=nlo[3] // 27:
+ (p16) cmp.ltu.unc p50,p48=t0,a1
+ (p16) nop.i 0 };;
+ .pred.rel "mutex",p40,p42
+ .pred.rel "mutex",p48,p50
+{ .mfi; (p16) nop.m 0 // 28:
+ (p16) xma.hu ahi[7]=ai7,bj[7],ahi[6] // ap[7]*b[i]
+ (p40) add a2=a2,n2 } // (p16) a2+=n2
+{ .mfi; (p42) add a2=a2,n2,1
+ (p16) xma.lu alo[7]=ai7,bj[7],ahi[6]
+ (p16) nop.i 0 };;
+{ .mii; (p16) getf.sig a6=alo[6] // 29:
+ (p48) add t[6]=t[6],a2 // (p16) t[6]+=a2
+ (p50) add t[6]=t[6],a2,1 };;
+{ .mfi; (p16) nop.m 0 // 30:
+ (p16) xma.hu nhi[5]=ni5,mj[0],nhi[4] // np[5]*m0
+ (p40) cmp.ltu p41,p39=a2,n2 }
+{ .mfi; (p42) cmp.leu p41,p39=a2,n2
+ (p16) xma.lu nlo[5]=ni5,mj[0],nhi[4]
+ (p16) nop.i 0 };;
+{ .mfi; (p16) getf.sig n4=nlo[4] // 31:
+ (p16) nop.f 0
+ (p48) cmp.ltu p49,p47=t[6],a2 }
+{ .mfb; (p50) cmp.leu p49,p47=t[6],a2
+ (p16) nop.f 0
+ br.ctop.sptk.many .Louter_8_ctop };;
+.Louter_8_cend:
+
+// above loop has to execute one more time, without (p16), which is
+// replaced with merged move of np[8] to GPR bank
+ .pred.rel "mutex",p40,p42
+ .pred.rel "mutex",p48,p50
+{ .mmi; (p0) getf.sig n1=ni0 // 0:
+ (p40) add a3=a3,n3 // (p17) a3+=n3
+ (p42) add a3=a3,n3,1 };;
+{ .mii; (p17) getf.sig a7=alo[8] // 1:
+ (p48) add t[6]=t[6],a3 // (p17) t[6]+=a3
+ (p50) add t[6]=t[6],a3,1 };;
+{ .mfi; (p17) getf.sig a8=ahi[8] // 2:
+ (p17) xma.hu nhi[7]=ni6,mj[1],nhi[6] // np[6]*m0
+ (p40) cmp.ltu p43,p41=a3,n3 }
+{ .mfi; (p42) cmp.leu p43,p41=a3,n3
+ (p17) xma.lu nlo[7]=ni6,mj[1],nhi[6]
+ (p0) nop.i 0 };;
+{ .mii; (p17) getf.sig n5=nlo[6] // 3:
+ (p48) cmp.ltu p51,p49=t[6],a3
+ (p50) cmp.leu p51,p49=t[6],a3 };;
+ .pred.rel "mutex",p41,p43
+ .pred.rel "mutex",p49,p51
+{ .mmi; (p0) getf.sig n2=ni1 // 4:
+ (p41) add a4=a4,n4 // (p17) a4+=n4
+ (p43) add a4=a4,n4,1 };;
+{ .mfi; (p49) add t[5]=t[5],a4 // 5: (p17) t[5]+=a4
+ (p0) nop.f 0
+ (p51) add t[5]=t[5],a4,1 };;
+{ .mfi; (p0) getf.sig n3=ni2 // 6:
+ (p17) xma.hu nhi[8]=ni7,mj[1],nhi[7] // np[7]*m0
+ (p41) cmp.ltu p42,p40=a4,n4 }
+{ .mfi; (p43) cmp.leu p42,p40=a4,n4
+ (p17) xma.lu nlo[8]=ni7,mj[1],nhi[7]
+ (p0) nop.i 0 };;
+{ .mii; (p17) getf.sig n6=nlo[7] // 7:
+ (p49) cmp.ltu p50,p48=t[5],a4
+ (p51) cmp.leu p50,p48=t[5],a4 };;
+ .pred.rel "mutex",p40,p42
+ .pred.rel "mutex",p48,p50
+{ .mii; (p0) getf.sig n4=ni3 // 8:
+ (p40) add a5=a5,n5 // (p17) a5+=n5
+ (p42) add a5=a5,n5,1 };;
+{ .mii; (p0) nop.m 0 // 9:
+ (p48) add t[4]=t[4],a5 // p(17) t[4]+=a5
+ (p50) add t[4]=t[4],a5,1 };;
+{ .mii; (p0) nop.m 0 // 10:
+ (p40) cmp.ltu p43,p41=a5,n5
+ (p42) cmp.leu p43,p41=a5,n5 };;
+{ .mii; (p17) getf.sig n7=nlo[8] // 11:
+ (p48) cmp.ltu p51,p49=t[4],a5
+ (p50) cmp.leu p51,p49=t[4],a5 };;
+ .pred.rel "mutex",p41,p43
+ .pred.rel "mutex",p49,p51
+{ .mii; (p17) getf.sig n8=nhi[8] // 12:
+ (p41) add a6=a6,n6 // (p17) a6+=n6
+ (p43) add a6=a6,n6,1 };;
+{ .mii; (p0) getf.sig n5=ni4 // 13:
+ (p49) add t[3]=t[3],a6 // (p17) t[3]+=a6
+ (p51) add t[3]=t[3],a6,1 };;
+{ .mii; (p0) nop.m 0 // 14:
+ (p41) cmp.ltu p42,p40=a6,n6
+ (p43) cmp.leu p42,p40=a6,n6 };;
+{ .mii; (p0) getf.sig n6=ni5 // 15:
+ (p49) cmp.ltu p50,p48=t[3],a6
+ (p51) cmp.leu p50,p48=t[3],a6 };;
+ .pred.rel "mutex",p40,p42
+ .pred.rel "mutex",p48,p50
+{ .mii; (p0) nop.m 0 // 16:
+ (p40) add a7=a7,n7 // (p17) a7+=n7
+ (p42) add a7=a7,n7,1 };;
+{ .mii; (p0) nop.m 0 // 17:
+ (p48) add t[2]=t[2],a7 // (p17) t[2]+=a7
+ (p50) add t[2]=t[2],a7,1 };;
+{ .mii; (p0) nop.m 0 // 18:
+ (p40) cmp.ltu p43,p41=a7,n7
+ (p42) cmp.leu p43,p41=a7,n7 };;
+{ .mii; (p0) getf.sig n7=ni6 // 19:
+ (p48) cmp.ltu p51,p49=t[2],a7
+ (p50) cmp.leu p51,p49=t[2],a7 };;
+ .pred.rel "mutex",p41,p43
+ .pred.rel "mutex",p49,p51
+{ .mii; (p0) nop.m 0 // 20:
+ (p41) add a8=a8,n8 // (p17) a8+=n8
+ (p43) add a8=a8,n8,1 };;
+{ .mmi; (p0) nop.m 0 // 21:
+ (p49) add t[1]=t[1],a8 // (p17) t[1]+=a8
+ (p51) add t[1]=t[1],a8,1 }
+{ .mmi; (p17) mov t[0]=r0
+ (p41) cmp.ltu p42,p40=a8,n8
+ (p43) cmp.leu p42,p40=a8,n8 };;
+{ .mmi; (p0) getf.sig n8=ni7 // 22:
+ (p49) cmp.ltu p50,p48=t[1],a8
+ (p51) cmp.leu p50,p48=t[1],a8 }
+{ .mmi; (p42) add t[0]=t[0],r0,1
+ (p0) add r16=-7*16,prevsp
+ (p0) add r17=-6*16,prevsp };;
+
+// subtract np[8] from carrybit|tmp[8]
+// carrybit|tmp[8] layout upon exit from above loop is:
+// t[0]|t[1]|t[2]|t[3]|t[4]|t[5]|t[6]|t[7]|t0 (least significant)
+{ .mmi; (p50)add t[0]=t[0],r0,1
+ add r18=-5*16,prevsp
+ sub n1=t0,n1 };;
+{ .mmi; cmp.gtu p34,p32=n1,t0;;
+ .pred.rel "mutex",p32,p34
+ (p32)sub n2=t[7],n2
+ (p34)sub n2=t[7],n2,1 };;
+{ .mii; (p32)cmp.gtu p35,p33=n2,t[7]
+ (p34)cmp.geu p35,p33=n2,t[7];;
+ .pred.rel "mutex",p33,p35
+ (p33)sub n3=t[6],n3 }
+{ .mmi; (p35)sub n3=t[6],n3,1;;
+ (p33)cmp.gtu p34,p32=n3,t[6]
+ (p35)cmp.geu p34,p32=n3,t[6] };;
+ .pred.rel "mutex",p32,p34
+{ .mii; (p32)sub n4=t[5],n4
+ (p34)sub n4=t[5],n4,1;;
+ (p32)cmp.gtu p35,p33=n4,t[5] }
+{ .mmi; (p34)cmp.geu p35,p33=n4,t[5];;
+ .pred.rel "mutex",p33,p35
+ (p33)sub n5=t[4],n5
+ (p35)sub n5=t[4],n5,1 };;
+{ .mii; (p33)cmp.gtu p34,p32=n5,t[4]
+ (p35)cmp.geu p34,p32=n5,t[4];;
+ .pred.rel "mutex",p32,p34
+ (p32)sub n6=t[3],n6 }
+{ .mmi; (p34)sub n6=t[3],n6,1;;
+ (p32)cmp.gtu p35,p33=n6,t[3]
+ (p34)cmp.geu p35,p33=n6,t[3] };;
+ .pred.rel "mutex",p33,p35
+{ .mii; (p33)sub n7=t[2],n7
+ (p35)sub n7=t[2],n7,1;;
+ (p33)cmp.gtu p34,p32=n7,t[2] }
+{ .mmi; (p35)cmp.geu p34,p32=n7,t[2];;
+ .pred.rel "mutex",p32,p34
+ (p32)sub n8=t[1],n8
+ (p34)sub n8=t[1],n8,1 };;
+{ .mii; (p32)cmp.gtu p35,p33=n8,t[1]
+ (p34)cmp.geu p35,p33=n8,t[1];;
+ .pred.rel "mutex",p33,p35
+ (p33)sub a8=t[0],r0 }
+{ .mmi; (p35)sub a8=t[0],r0,1;;
+ (p33)cmp.gtu p34,p32=a8,t[0]
+ (p35)cmp.geu p34,p32=a8,t[0] };;
+
+// save the result, either tmp[num] or tmp[num]-np[num]
+ .pred.rel "mutex",p32,p34
+{ .mmi; (p32)st8 [rptr]=n1,8
+ (p34)st8 [rptr]=t0,8
+ add r19=-4*16,prevsp};;
+{ .mmb; (p32)st8 [rptr]=n2,8
+ (p34)st8 [rptr]=t[7],8
+ (p5)br.cond.dpnt.few .Ldone };;
+{ .mmb; (p32)st8 [rptr]=n3,8
+ (p34)st8 [rptr]=t[6],8
+ (p7)br.cond.dpnt.few .Ldone };;
+{ .mmb; (p32)st8 [rptr]=n4,8
+ (p34)st8 [rptr]=t[5],8
+ (p9)br.cond.dpnt.few .Ldone };;
+{ .mmb; (p32)st8 [rptr]=n5,8
+ (p34)st8 [rptr]=t[4],8
+ (p11)br.cond.dpnt.few .Ldone };;
+{ .mmb; (p32)st8 [rptr]=n6,8
+ (p34)st8 [rptr]=t[3],8
+ (p13)br.cond.dpnt.few .Ldone };;
+{ .mmb; (p32)st8 [rptr]=n7,8
+ (p34)st8 [rptr]=t[2],8
+ (p15)br.cond.dpnt.few .Ldone };;
+{ .mmb; (p32)st8 [rptr]=n8,8
+ (p34)st8 [rptr]=t[1],8
+ nop.b 0 };;
+.Ldone: // epilogue
+{ .mmi; ldf.fill f16=[r16],64
+ ldf.fill f17=[r17],64
+ nop.i 0 }
+{ .mmi; ldf.fill f18=[r18],64
+ ldf.fill f19=[r19],64
+ mov pr=prevpr,0x1ffff };;
+{ .mmi; ldf.fill f20=[r16]
+ ldf.fill f21=[r17]
+ mov ar.lc=prevlc }
+{ .mmi; ldf.fill f22=[r18]
+ ldf.fill f23=[r19]
+ mov ret0=1 } // signal "handled"
+{ .mib; rum 1<<5
+ .restore sp
+ mov sp=prevsp
+ br.ret.sptk.many b0 };;
+.endp bn_mul_mont_8#
+
+.type copyright#,\@object
+copyright:
+stringz "Montgomery multiplication for IA-64, CRYPTOGAMS by <appro\@openssl.org>"
+___
+
+open STDOUT,">$output" if $output;
+print $code;
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/ia64.S b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/ia64.S
new file mode 100644
index 0000000..0a26735
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/ia64.S
@@ -0,0 +1,1565 @@
+.explicit
+.text
+.ident "ia64.S, Version 2.1"
+.ident "IA-64 ISA artwork by Andy Polyakov <appro@openssl.org>"
+
+// Copyright 2001-2019 The OpenSSL Project Authors. All Rights Reserved.
+//
+// Licensed under the OpenSSL license (the "License"). You may not use
+// this file except in compliance with the License. You can obtain a copy
+// in the file LICENSE in the source distribution or at
+// https://www.openssl.org/source/license.html
+
+//
+// ====================================================================
+// Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+// project.
+//
+// Rights for redistribution and usage in source and binary forms are
+// granted according to the OpenSSL license. Warranty of any kind is
+// disclaimed.
+// ====================================================================
+//
+// Version 2.x is Itanium2 re-tune. Few words about how Itanium2 is
+// different from Itanium to this module viewpoint. Most notably, is it
+// "wider" than Itanium? Can you experience loop scalability as
+// discussed in commentary sections? Not really:-( Itanium2 has 6
+// integer ALU ports, i.e. it's 2 ports wider, but it's not enough to
+// spin twice as fast, as I need 8 IALU ports. Amount of floating point
+// ports is the same, i.e. 2, while I need 4. In other words, to this
+// module Itanium2 remains effectively as "wide" as Itanium. Yet it's
+// essentially different in respect to this module, and a re-tune was
+// required. Well, because some instruction latencies has changed. Most
+// noticeably those intensively used:
+//
+// Itanium Itanium2
+// ldf8 9 6 L2 hit
+// ld8 2 1 L1 hit
+// getf 2 5
+// xma[->getf] 7[+1] 4[+0]
+// add[->st8] 1[+1] 1[+0]
+//
+// What does it mean? You might ratiocinate that the original code
+// should run just faster... Because sum of latencies is smaller...
+// Wrong! Note that getf latency increased. This means that if a loop is
+// scheduled for lower latency (as they were), then it will suffer from
+// stall condition and the code will therefore turn anti-scalable, e.g.
+// original bn_mul_words spun at 5*n or 2.5 times slower than expected
+// on Itanium2! What to do? Reschedule loops for Itanium2? But then
+// Itanium would exhibit anti-scalability. So I've chosen to reschedule
+// for worst latency for every instruction aiming for best *all-round*
+// performance.
+
+// Q. How much faster does it get?
+// A. Here is the output from 'openssl speed rsa dsa' for vanilla
+// 0.9.6a compiled with gcc version 2.96 20000731 (Red Hat
+// Linux 7.1 2.96-81):
+//
+// sign verify sign/s verify/s
+// rsa 512 bits 0.0036s 0.0003s 275.3 2999.2
+// rsa 1024 bits 0.0203s 0.0011s 49.3 894.1
+// rsa 2048 bits 0.1331s 0.0040s 7.5 250.9
+// rsa 4096 bits 0.9270s 0.0147s 1.1 68.1
+// sign verify sign/s verify/s
+// dsa 512 bits 0.0035s 0.0043s 288.3 234.8
+// dsa 1024 bits 0.0111s 0.0135s 90.0 74.2
+//
+// And here is similar output but for this assembler
+// implementation:-)
+//
+// sign verify sign/s verify/s
+// rsa 512 bits 0.0021s 0.0001s 549.4 9638.5
+// rsa 1024 bits 0.0055s 0.0002s 183.8 4481.1
+// rsa 2048 bits 0.0244s 0.0006s 41.4 1726.3
+// rsa 4096 bits 0.1295s 0.0018s 7.7 561.5
+// sign verify sign/s verify/s
+// dsa 512 bits 0.0012s 0.0013s 891.9 756.6
+// dsa 1024 bits 0.0023s 0.0028s 440.4 376.2
+//
+// Yes, you may argue that it's not fair comparison as it's
+// possible to craft the C implementation with BN_UMULT_HIGH
+// inline assembler macro. But of course! Here is the output
+// with the macro:
+//
+// sign verify sign/s verify/s
+// rsa 512 bits 0.0020s 0.0002s 495.0 6561.0
+// rsa 1024 bits 0.0086s 0.0004s 116.2 2235.7
+// rsa 2048 bits 0.0519s 0.0015s 19.3 667.3
+// rsa 4096 bits 0.3464s 0.0053s 2.9 187.7
+// sign verify sign/s verify/s
+// dsa 512 bits 0.0016s 0.0020s 613.1 510.5
+// dsa 1024 bits 0.0045s 0.0054s 221.0 183.9
+//
+// My code is still way faster, huh:-) And I believe that even
+// higher performance can be achieved. Note that as keys get
+// longer, performance gain is larger. Why? According to the
+// profiler there is another player in the field, namely
+// BN_from_montgomery consuming larger and larger portion of CPU
+// time as keysize decreases. I therefore consider putting effort
+// to assembler implementation of the following routine:
+//
+// void bn_mul_add_mont (BN_ULONG *rp,BN_ULONG *np,int nl,BN_ULONG n0)
+// {
+// int i,j;
+// BN_ULONG v;
+//
+// for (i=0; i<nl; i++)
+// {
+// v=bn_mul_add_words(rp,np,nl,(rp[0]*n0)&BN_MASK2);
+// nrp++;
+// rp++;
+// if (((nrp[-1]+=v)&BN_MASK2) < v)
+// for (j=0; ((++nrp[j])&BN_MASK2) == 0; j++) ;
+// }
+// }
+//
+// It might as well be beneficial to implement even combaX
+// variants, as it appears as it can literally unleash the
+// performance (see comment section to bn_mul_comba8 below).
+//
+// And finally for your reference the output for 0.9.6a compiled
+// with SGIcc version 0.01.0-12 (keep in mind that for the moment
+// of this writing it's not possible to convince SGIcc to use
+// BN_UMULT_HIGH inline assembler macro, yet the code is fast,
+// i.e. for a compiler generated one:-):
+//
+// sign verify sign/s verify/s
+// rsa 512 bits 0.0022s 0.0002s 452.7 5894.3
+// rsa 1024 bits 0.0097s 0.0005s 102.7 2002.9
+// rsa 2048 bits 0.0578s 0.0017s 17.3 600.2
+// rsa 4096 bits 0.3838s 0.0061s 2.6 164.5
+// sign verify sign/s verify/s
+// dsa 512 bits 0.0018s 0.0022s 547.3 459.6
+// dsa 1024 bits 0.0051s 0.0062s 196.6 161.3
+//
+// Oh! Benchmarks were performed on 733MHz Lion-class Itanium
+// system running Redhat Linux 7.1 (very special thanks to Ray
+// McCaffity of Williams Communications for providing an account).
+//
+// Q. What's the heck with 'rum 1<<5' at the end of every function?
+// A. Well, by clearing the "upper FP registers written" bit of the
+// User Mask I want to excuse the kernel from preserving upper
+// (f32-f128) FP register bank over process context switch, thus
+// minimizing bus bandwidth consumption during the switch (i.e.
+// after PKI operation completes and the program is off doing
+// something else like bulk symmetric encryption). Having said
+// this, I also want to point out that it might be good idea
+// to compile the whole toolkit (as well as majority of the
+// programs for that matter) with -mfixed-range=f32-f127 command
+// line option. No, it doesn't prevent the compiler from writing
+// to upper bank, but at least discourages to do so. If you don't
+// like the idea you have the option to compile the module with
+// -Drum=nop.m in command line.
+//
+
+#if defined(_HPUX_SOURCE) && !defined(_LP64)
+#define ADDP addp4
+#else
+#define ADDP add
+#endif
+#ifdef __VMS
+.alias abort, "decc$abort"
+#endif
+
+#if 1
+//
+// bn_[add|sub]_words routines.
+//
+// Loops are spinning in 2*(n+5) ticks on Itanium (provided that the
+// data reside in L1 cache, i.e. 2 ticks away). It's possible to
+// compress the epilogue and get down to 2*n+6, but at the cost of
+// scalability (the neat feature of this implementation is that it
+// shall automagically spin in n+5 on "wider" IA-64 implementations:-)
+// I consider that the epilogue is short enough as it is to trade tiny
+// performance loss on Itanium for scalability.
+//
+// BN_ULONG bn_add_words(BN_ULONG *rp, BN_ULONG *ap, BN_ULONG *bp,int num)
+//
+.global bn_add_words#
+.proc bn_add_words#
+.align 64
+.skip 32 // makes the loop body aligned at 64-byte boundary
+bn_add_words:
+ .prologue
+ .save ar.pfs,r2
+{ .mii; alloc r2=ar.pfs,4,12,0,16
+ cmp4.le p6,p0=r35,r0 };;
+{ .mfb; mov r8=r0 // return value
+(p6) br.ret.spnt.many b0 };;
+
+{ .mib; sub r10=r35,r0,1
+ .save ar.lc,r3
+ mov r3=ar.lc
+ brp.loop.imp .L_bn_add_words_ctop,.L_bn_add_words_cend-16
+ }
+{ .mib; ADDP r14=0,r32 // rp
+ .save pr,r9
+ mov r9=pr };;
+ .body
+{ .mii; ADDP r15=0,r33 // ap
+ mov ar.lc=r10
+ mov ar.ec=6 }
+{ .mib; ADDP r16=0,r34 // bp
+ mov pr.rot=1<<16 };;
+
+.L_bn_add_words_ctop:
+{ .mii; (p16) ld8 r32=[r16],8 // b=*(bp++)
+ (p18) add r39=r37,r34
+ (p19) cmp.ltu.unc p56,p0=r40,r38 }
+{ .mfb; (p0) nop.m 0x0
+ (p0) nop.f 0x0
+ (p0) nop.b 0x0 }
+{ .mii; (p16) ld8 r35=[r15],8 // a=*(ap++)
+ (p58) cmp.eq.or p57,p0=-1,r41 // (p20)
+ (p58) add r41=1,r41 } // (p20)
+{ .mfb; (p21) st8 [r14]=r42,8 // *(rp++)=r
+ (p0) nop.f 0x0
+ br.ctop.sptk .L_bn_add_words_ctop };;
+.L_bn_add_words_cend:
+
+{ .mii;
+(p59) add r8=1,r8 // return value
+ mov pr=r9,0x1ffff
+ mov ar.lc=r3 }
+{ .mbb; nop.b 0x0
+ br.ret.sptk.many b0 };;
+.endp bn_add_words#
+
+//
+// BN_ULONG bn_sub_words(BN_ULONG *rp, BN_ULONG *ap, BN_ULONG *bp,int num)
+//
+.global bn_sub_words#
+.proc bn_sub_words#
+.align 64
+.skip 32 // makes the loop body aligned at 64-byte boundary
+bn_sub_words:
+ .prologue
+ .save ar.pfs,r2
+{ .mii; alloc r2=ar.pfs,4,12,0,16
+ cmp4.le p6,p0=r35,r0 };;
+{ .mfb; mov r8=r0 // return value
+(p6) br.ret.spnt.many b0 };;
+
+{ .mib; sub r10=r35,r0,1
+ .save ar.lc,r3
+ mov r3=ar.lc
+ brp.loop.imp .L_bn_sub_words_ctop,.L_bn_sub_words_cend-16
+ }
+{ .mib; ADDP r14=0,r32 // rp
+ .save pr,r9
+ mov r9=pr };;
+ .body
+{ .mii; ADDP r15=0,r33 // ap
+ mov ar.lc=r10
+ mov ar.ec=6 }
+{ .mib; ADDP r16=0,r34 // bp
+ mov pr.rot=1<<16 };;
+
+.L_bn_sub_words_ctop:
+{ .mii; (p16) ld8 r32=[r16],8 // b=*(bp++)
+ (p18) sub r39=r37,r34
+ (p19) cmp.gtu.unc p56,p0=r40,r38 }
+{ .mfb; (p0) nop.m 0x0
+ (p0) nop.f 0x0
+ (p0) nop.b 0x0 }
+{ .mii; (p16) ld8 r35=[r15],8 // a=*(ap++)
+ (p58) cmp.eq.or p57,p0=0,r41 // (p20)
+ (p58) add r41=-1,r41 } // (p20)
+{ .mbb; (p21) st8 [r14]=r42,8 // *(rp++)=r
+ (p0) nop.b 0x0
+ br.ctop.sptk .L_bn_sub_words_ctop };;
+.L_bn_sub_words_cend:
+
+{ .mii;
+(p59) add r8=1,r8 // return value
+ mov pr=r9,0x1ffff
+ mov ar.lc=r3 }
+{ .mbb; nop.b 0x0
+ br.ret.sptk.many b0 };;
+.endp bn_sub_words#
+#endif
+
+#if 0
+#define XMA_TEMPTATION
+#endif
+
+#if 1
+//
+// BN_ULONG bn_mul_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w)
+//
+.global bn_mul_words#
+.proc bn_mul_words#
+.align 64
+.skip 32 // makes the loop body aligned at 64-byte boundary
+bn_mul_words:
+ .prologue
+ .save ar.pfs,r2
+#ifdef XMA_TEMPTATION
+{ .mfi; alloc r2=ar.pfs,4,0,0,0 };;
+#else
+{ .mfi; alloc r2=ar.pfs,4,12,0,16 };;
+#endif
+{ .mib; mov r8=r0 // return value
+ cmp4.le p6,p0=r34,r0
+(p6) br.ret.spnt.many b0 };;
+
+{ .mii; sub r10=r34,r0,1
+ .save ar.lc,r3
+ mov r3=ar.lc
+ .save pr,r9
+ mov r9=pr };;
+
+ .body
+{ .mib; setf.sig f8=r35 // w
+ mov pr.rot=0x800001<<16
+ // ------^----- serves as (p50) at first (p27)
+ brp.loop.imp .L_bn_mul_words_ctop,.L_bn_mul_words_cend-16
+ }
+
+#ifndef XMA_TEMPTATION
+
+{ .mmi; ADDP r14=0,r32 // rp
+ ADDP r15=0,r33 // ap
+ mov ar.lc=r10 }
+{ .mmi; mov r40=0 // serves as r35 at first (p27)
+ mov ar.ec=13 };;
+
+// This loop spins in 2*(n+12) ticks. It's scheduled for data in Itanium
+// L2 cache (i.e. 9 ticks away) as floating point load/store instructions
+// bypass L1 cache and L2 latency is actually best-case scenario for
+// ldf8. The loop is not scalable and shall run in 2*(n+12) even on
+// "wider" IA-64 implementations. It's a trade-off here. n+24 loop
+// would give us ~5% in *overall* performance improvement on "wider"
+// IA-64, but would hurt Itanium for about same because of longer
+// epilogue. As it's a matter of few percents in either case I've
+// chosen to trade the scalability for development time (you can see
+// this very instruction sequence in bn_mul_add_words loop which in
+// turn is scalable).
+.L_bn_mul_words_ctop:
+{ .mfi; (p25) getf.sig r36=f52 // low
+ (p21) xmpy.lu f48=f37,f8
+ (p28) cmp.ltu p54,p50=r41,r39 }
+{ .mfi; (p16) ldf8 f32=[r15],8
+ (p21) xmpy.hu f40=f37,f8
+ (p0) nop.i 0x0 };;
+{ .mii; (p25) getf.sig r32=f44 // high
+ .pred.rel "mutex",p50,p54
+ (p50) add r40=r38,r35 // (p27)
+ (p54) add r40=r38,r35,1 } // (p27)
+{ .mfb; (p28) st8 [r14]=r41,8
+ (p0) nop.f 0x0
+ br.ctop.sptk .L_bn_mul_words_ctop };;
+.L_bn_mul_words_cend:
+
+{ .mii; nop.m 0x0
+.pred.rel "mutex",p51,p55
+(p51) add r8=r36,r0
+(p55) add r8=r36,r0,1 }
+{ .mfb; nop.m 0x0
+ nop.f 0x0
+ nop.b 0x0 }
+
+#else // XMA_TEMPTATION
+
+ setf.sig f37=r0 // serves as carry at (p18) tick
+ mov ar.lc=r10
+ mov ar.ec=5;;
+
+// Most of you examining this code very likely wonder why in the name
+// of Intel the following loop is commented out? Indeed, it looks so
+// neat that you find it hard to believe that it's something wrong
+// with it, right? The catch is that every iteration depends on the
+// result from previous one and the latter isn't available instantly.
+// The loop therefore spins at the latency of xma minus 1, or in other
+// words at 6*(n+4) ticks:-( Compare to the "production" loop above
+// that runs in 2*(n+11) where the low latency problem is worked around
+// by moving the dependency to one-tick latent integer ALU. Note that
+// "distance" between ldf8 and xma is not latency of ldf8, but the
+// *difference* between xma and ldf8 latencies.
+.L_bn_mul_words_ctop:
+{ .mfi; (p16) ldf8 f32=[r33],8
+ (p18) xma.hu f38=f34,f8,f39 }
+{ .mfb; (p20) stf8 [r32]=f37,8
+ (p18) xma.lu f35=f34,f8,f39
+ br.ctop.sptk .L_bn_mul_words_ctop };;
+.L_bn_mul_words_cend:
+
+ getf.sig r8=f41 // the return value
+
+#endif // XMA_TEMPTATION
+
+{ .mii; nop.m 0x0
+ mov pr=r9,0x1ffff
+ mov ar.lc=r3 }
+{ .mfb; rum 1<<5 // clear um.mfh
+ nop.f 0x0
+ br.ret.sptk.many b0 };;
+.endp bn_mul_words#
+#endif
+
+#if 1
+//
+// BN_ULONG bn_mul_add_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w)
+//
+.global bn_mul_add_words#
+.proc bn_mul_add_words#
+.align 64
+.skip 48 // makes the loop body aligned at 64-byte boundary
+bn_mul_add_words:
+ .prologue
+ .save ar.pfs,r2
+{ .mmi; alloc r2=ar.pfs,4,4,0,8
+ cmp4.le p6,p0=r34,r0
+ .save ar.lc,r3
+ mov r3=ar.lc };;
+{ .mib; mov r8=r0 // return value
+ sub r10=r34,r0,1
+(p6) br.ret.spnt.many b0 };;
+
+{ .mib; setf.sig f8=r35 // w
+ .save pr,r9
+ mov r9=pr
+ brp.loop.imp .L_bn_mul_add_words_ctop,.L_bn_mul_add_words_cend-16
+ }
+ .body
+{ .mmi; ADDP r14=0,r32 // rp
+ ADDP r15=0,r33 // ap
+ mov ar.lc=r10 }
+{ .mii; ADDP r16=0,r32 // rp copy
+ mov pr.rot=0x2001<<16
+ // ------^----- serves as (p40) at first (p27)
+ mov ar.ec=11 };;
+
+// This loop spins in 3*(n+10) ticks on Itanium and in 2*(n+10) on
+// Itanium 2. Yes, unlike previous versions it scales:-) Previous
+// version was performing *all* additions in IALU and was starving
+// for those even on Itanium 2. In this version one addition is
+// moved to FPU and is folded with multiplication. This is at cost
+// of propagating the result from previous call to this subroutine
+// to L2 cache... In other words negligible even for shorter keys.
+// *Overall* performance improvement [over previous version] varies
+// from 11 to 22 percent depending on key length.
+.L_bn_mul_add_words_ctop:
+.pred.rel "mutex",p40,p42
+{ .mfi; (p23) getf.sig r36=f45 // low
+ (p20) xma.lu f42=f36,f8,f50 // low
+ (p40) add r39=r39,r35 } // (p27)
+{ .mfi; (p16) ldf8 f32=[r15],8 // *(ap++)
+ (p20) xma.hu f36=f36,f8,f50 // high
+ (p42) add r39=r39,r35,1 };; // (p27)
+{ .mmi; (p24) getf.sig r32=f40 // high
+ (p16) ldf8 f46=[r16],8 // *(rp1++)
+ (p40) cmp.ltu p41,p39=r39,r35 } // (p27)
+{ .mib; (p26) st8 [r14]=r39,8 // *(rp2++)
+ (p42) cmp.leu p41,p39=r39,r35 // (p27)
+ br.ctop.sptk .L_bn_mul_add_words_ctop};;
+.L_bn_mul_add_words_cend:
+
+{ .mmi; .pred.rel "mutex",p40,p42
+(p40) add r8=r35,r0
+(p42) add r8=r35,r0,1
+ mov pr=r9,0x1ffff }
+{ .mib; rum 1<<5 // clear um.mfh
+ mov ar.lc=r3
+ br.ret.sptk.many b0 };;
+.endp bn_mul_add_words#
+#endif
+
+#if 1
+//
+// void bn_sqr_words(BN_ULONG *rp, BN_ULONG *ap, int num)
+//
+.global bn_sqr_words#
+.proc bn_sqr_words#
+.align 64
+.skip 32 // makes the loop body aligned at 64-byte boundary
+bn_sqr_words:
+ .prologue
+ .save ar.pfs,r2
+{ .mii; alloc r2=ar.pfs,3,0,0,0
+ sxt4 r34=r34 };;
+{ .mii; cmp.le p6,p0=r34,r0
+ mov r8=r0 } // return value
+{ .mfb; ADDP r32=0,r32
+ nop.f 0x0
+(p6) br.ret.spnt.many b0 };;
+
+{ .mii; sub r10=r34,r0,1
+ .save ar.lc,r3
+ mov r3=ar.lc
+ .save pr,r9
+ mov r9=pr };;
+
+ .body
+{ .mib; ADDP r33=0,r33
+ mov pr.rot=1<<16
+ brp.loop.imp .L_bn_sqr_words_ctop,.L_bn_sqr_words_cend-16
+ }
+{ .mii; add r34=8,r32
+ mov ar.lc=r10
+ mov ar.ec=18 };;
+
+// 2*(n+17) on Itanium, (n+17) on "wider" IA-64 implementations. It's
+// possible to compress the epilogue (I'm getting tired to write this
+// comment over and over) and get down to 2*n+16 at the cost of
+// scalability. The decision will very likely be reconsidered after the
+// benchmark program is profiled. I.e. if performance gain on Itanium
+// will appear larger than loss on "wider" IA-64, then the loop should
+// be explicitly split and the epilogue compressed.
+.L_bn_sqr_words_ctop:
+{ .mfi; (p16) ldf8 f32=[r33],8
+ (p25) xmpy.lu f42=f41,f41
+ (p0) nop.i 0x0 }
+{ .mib; (p33) stf8 [r32]=f50,16
+ (p0) nop.i 0x0
+ (p0) nop.b 0x0 }
+{ .mfi; (p0) nop.m 0x0
+ (p25) xmpy.hu f52=f41,f41
+ (p0) nop.i 0x0 }
+{ .mib; (p33) stf8 [r34]=f60,16
+ (p0) nop.i 0x0
+ br.ctop.sptk .L_bn_sqr_words_ctop };;
+.L_bn_sqr_words_cend:
+
+{ .mii; nop.m 0x0
+ mov pr=r9,0x1ffff
+ mov ar.lc=r3 }
+{ .mfb; rum 1<<5 // clear um.mfh
+ nop.f 0x0
+ br.ret.sptk.many b0 };;
+.endp bn_sqr_words#
+#endif
+
+#if 1
+// Apparently we win nothing by implementing special bn_sqr_comba8.
+// Yes, it is possible to reduce the number of multiplications by
+// almost factor of two, but then the amount of additions would
+// increase by factor of two (as we would have to perform those
+// otherwise performed by xma ourselves). Normally we would trade
+// anyway as multiplications are way more expensive, but not this
+// time... Multiplication kernel is fully pipelined and as we drain
+// one 128-bit multiplication result per clock cycle multiplications
+// are effectively as inexpensive as additions. Special implementation
+// might become of interest for "wider" IA-64 implementation as you'll
+// be able to get through the multiplication phase faster (there won't
+// be any stall issues as discussed in the commentary section below and
+// you therefore will be able to employ all 4 FP units)... But these
+// Itanium days it's simply too hard to justify the effort so I just
+// drop down to bn_mul_comba8 code:-)
+//
+// void bn_sqr_comba8(BN_ULONG *r, BN_ULONG *a)
+//
+.global bn_sqr_comba8#
+.proc bn_sqr_comba8#
+.align 64
+bn_sqr_comba8:
+ .prologue
+ .save ar.pfs,r2
+#if defined(_HPUX_SOURCE) && !defined(_LP64)
+{ .mii; alloc r2=ar.pfs,2,1,0,0
+ addp4 r33=0,r33
+ addp4 r32=0,r32 };;
+{ .mii;
+#else
+{ .mii; alloc r2=ar.pfs,2,1,0,0
+#endif
+ mov r34=r33
+ add r14=8,r33 };;
+ .body
+{ .mii; add r17=8,r34
+ add r15=16,r33
+ add r18=16,r34 }
+{ .mfb; add r16=24,r33
+ br .L_cheat_entry_point8 };;
+.endp bn_sqr_comba8#
+#endif
+
+#if 1
+// I've estimated this routine to run in ~120 ticks, but in reality
+// (i.e. according to ar.itc) it takes ~160 ticks. Are those extra
+// cycles consumed for instructions fetch? Or did I misinterpret some
+// clause in Itanium µ-architecture manual? Comments are welcomed and
+// highly appreciated.
+//
+// On Itanium 2 it takes ~190 ticks. This is because of stalls on
+// result from getf.sig. I do nothing about it at this point for
+// reasons depicted below.
+//
+// However! It should be noted that even 160 ticks is darn good result
+// as it's over 10 (yes, ten, spelled as t-e-n) times faster than the
+// C version (compiled with gcc with inline assembler). I really
+// kicked compiler's butt here, didn't I? Yeah! This brings us to the
+// following statement. It's damn shame that this routine isn't called
+// very often nowadays! According to the profiler most CPU time is
+// consumed by bn_mul_add_words called from BN_from_montgomery. In
+// order to estimate what we're missing, I've compared the performance
+// of this routine against "traditional" implementation, i.e. against
+// following routine:
+//
+// void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
+// { r[ 8]=bn_mul_words( &(r[0]),a,8,b[0]);
+// r[ 9]=bn_mul_add_words(&(r[1]),a,8,b[1]);
+// r[10]=bn_mul_add_words(&(r[2]),a,8,b[2]);
+// r[11]=bn_mul_add_words(&(r[3]),a,8,b[3]);
+// r[12]=bn_mul_add_words(&(r[4]),a,8,b[4]);
+// r[13]=bn_mul_add_words(&(r[5]),a,8,b[5]);
+// r[14]=bn_mul_add_words(&(r[6]),a,8,b[6]);
+// r[15]=bn_mul_add_words(&(r[7]),a,8,b[7]);
+// }
+//
+// The one below is over 8 times faster than the one above:-( Even
+// more reasons to "combafy" bn_mul_add_mont...
+//
+// And yes, this routine really made me wish there were an optimizing
+// assembler! It also feels like it deserves a dedication.
+//
+// To my wife for being there and to my kids...
+//
+// void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
+//
+#define carry1 r14
+#define carry2 r15
+#define carry3 r34
+.global bn_mul_comba8#
+.proc bn_mul_comba8#
+.align 64
+bn_mul_comba8:
+ .prologue
+ .save ar.pfs,r2
+#if defined(_HPUX_SOURCE) && !defined(_LP64)
+{ .mii; alloc r2=ar.pfs,3,0,0,0
+ addp4 r33=0,r33
+ addp4 r34=0,r34 };;
+{ .mii; addp4 r32=0,r32
+#else
+{ .mii; alloc r2=ar.pfs,3,0,0,0
+#endif
+ add r14=8,r33
+ add r17=8,r34 }
+ .body
+{ .mii; add r15=16,r33
+ add r18=16,r34
+ add r16=24,r33 }
+.L_cheat_entry_point8:
+{ .mmi; add r19=24,r34
+
+ ldf8 f32=[r33],32 };;
+
+{ .mmi; ldf8 f120=[r34],32
+ ldf8 f121=[r17],32 }
+{ .mmi; ldf8 f122=[r18],32
+ ldf8 f123=[r19],32 };;
+{ .mmi; ldf8 f124=[r34]
+ ldf8 f125=[r17] }
+{ .mmi; ldf8 f126=[r18]
+ ldf8 f127=[r19] }
+
+{ .mmi; ldf8 f33=[r14],32
+ ldf8 f34=[r15],32 }
+{ .mmi; ldf8 f35=[r16],32;;
+ ldf8 f36=[r33] }
+{ .mmi; ldf8 f37=[r14]
+ ldf8 f38=[r15] }
+{ .mfi; ldf8 f39=[r16]
+// -------\ Entering multiplier's heaven /-------
+// ------------\ /------------
+// -----------------\ /-----------------
+// ----------------------\/----------------------
+ xma.hu f41=f32,f120,f0 }
+{ .mfi; xma.lu f40=f32,f120,f0 };; // (*)
+{ .mfi; xma.hu f51=f32,f121,f0 }
+{ .mfi; xma.lu f50=f32,f121,f0 };;
+{ .mfi; xma.hu f61=f32,f122,f0 }
+{ .mfi; xma.lu f60=f32,f122,f0 };;
+{ .mfi; xma.hu f71=f32,f123,f0 }
+{ .mfi; xma.lu f70=f32,f123,f0 };;
+{ .mfi; xma.hu f81=f32,f124,f0 }
+{ .mfi; xma.lu f80=f32,f124,f0 };;
+{ .mfi; xma.hu f91=f32,f125,f0 }
+{ .mfi; xma.lu f90=f32,f125,f0 };;
+{ .mfi; xma.hu f101=f32,f126,f0 }
+{ .mfi; xma.lu f100=f32,f126,f0 };;
+{ .mfi; xma.hu f111=f32,f127,f0 }
+{ .mfi; xma.lu f110=f32,f127,f0 };;//
+// (*) You can argue that splitting at every second bundle would
+// prevent "wider" IA-64 implementations from achieving the peak
+// performance. Well, not really... The catch is that if you
+// intend to keep 4 FP units busy by splitting at every fourth
+// bundle and thus perform these 16 multiplications in 4 ticks,
+// the first bundle *below* would stall because the result from
+// the first xma bundle *above* won't be available for another 3
+// ticks (if not more, being an optimist, I assume that "wider"
+// implementation will have same latency:-). This stall will hold
+// you back and the performance would be as if every second bundle
+// were split *anyway*...
+{ .mfi; getf.sig r16=f40
+ xma.hu f42=f33,f120,f41
+ add r33=8,r32 }
+{ .mfi; xma.lu f41=f33,f120,f41 };;
+{ .mfi; getf.sig r24=f50
+ xma.hu f52=f33,f121,f51 }
+{ .mfi; xma.lu f51=f33,f121,f51 };;
+{ .mfi; st8 [r32]=r16,16
+ xma.hu f62=f33,f122,f61 }
+{ .mfi; xma.lu f61=f33,f122,f61 };;
+{ .mfi; xma.hu f72=f33,f123,f71 }
+{ .mfi; xma.lu f71=f33,f123,f71 };;
+{ .mfi; xma.hu f82=f33,f124,f81 }
+{ .mfi; xma.lu f81=f33,f124,f81 };;
+{ .mfi; xma.hu f92=f33,f125,f91 }
+{ .mfi; xma.lu f91=f33,f125,f91 };;
+{ .mfi; xma.hu f102=f33,f126,f101 }
+{ .mfi; xma.lu f101=f33,f126,f101 };;
+{ .mfi; xma.hu f112=f33,f127,f111 }
+{ .mfi; xma.lu f111=f33,f127,f111 };;//
+//-------------------------------------------------//
+{ .mfi; getf.sig r25=f41
+ xma.hu f43=f34,f120,f42 }
+{ .mfi; xma.lu f42=f34,f120,f42 };;
+{ .mfi; getf.sig r16=f60
+ xma.hu f53=f34,f121,f52 }
+{ .mfi; xma.lu f52=f34,f121,f52 };;
+{ .mfi; getf.sig r17=f51
+ xma.hu f63=f34,f122,f62
+ add r25=r25,r24 }
+{ .mfi; xma.lu f62=f34,f122,f62
+ mov carry1=0 };;
+{ .mfi; cmp.ltu p6,p0=r25,r24
+ xma.hu f73=f34,f123,f72 }
+{ .mfi; xma.lu f72=f34,f123,f72 };;
+{ .mfi; st8 [r33]=r25,16
+ xma.hu f83=f34,f124,f82
+(p6) add carry1=1,carry1 }
+{ .mfi; xma.lu f82=f34,f124,f82 };;
+{ .mfi; xma.hu f93=f34,f125,f92 }
+{ .mfi; xma.lu f92=f34,f125,f92 };;
+{ .mfi; xma.hu f103=f34,f126,f102 }
+{ .mfi; xma.lu f102=f34,f126,f102 };;
+{ .mfi; xma.hu f113=f34,f127,f112 }
+{ .mfi; xma.lu f112=f34,f127,f112 };;//
+//-------------------------------------------------//
+{ .mfi; getf.sig r18=f42
+ xma.hu f44=f35,f120,f43
+ add r17=r17,r16 }
+{ .mfi; xma.lu f43=f35,f120,f43 };;
+{ .mfi; getf.sig r24=f70
+ xma.hu f54=f35,f121,f53 }
+{ .mfi; mov carry2=0
+ xma.lu f53=f35,f121,f53 };;
+{ .mfi; getf.sig r25=f61
+ xma.hu f64=f35,f122,f63
+ cmp.ltu p7,p0=r17,r16 }
+{ .mfi; add r18=r18,r17
+ xma.lu f63=f35,f122,f63 };;
+{ .mfi; getf.sig r26=f52
+ xma.hu f74=f35,f123,f73
+(p7) add carry2=1,carry2 }
+{ .mfi; cmp.ltu p7,p0=r18,r17
+ xma.lu f73=f35,f123,f73
+ add r18=r18,carry1 };;
+{ .mfi;
+ xma.hu f84=f35,f124,f83
+(p7) add carry2=1,carry2 }
+{ .mfi; cmp.ltu p7,p0=r18,carry1
+ xma.lu f83=f35,f124,f83 };;
+{ .mfi; st8 [r32]=r18,16
+ xma.hu f94=f35,f125,f93
+(p7) add carry2=1,carry2 }
+{ .mfi; xma.lu f93=f35,f125,f93 };;
+{ .mfi; xma.hu f104=f35,f126,f103 }
+{ .mfi; xma.lu f103=f35,f126,f103 };;
+{ .mfi; xma.hu f114=f35,f127,f113 }
+{ .mfi; mov carry1=0
+ xma.lu f113=f35,f127,f113
+ add r25=r25,r24 };;//
+//-------------------------------------------------//
+{ .mfi; getf.sig r27=f43
+ xma.hu f45=f36,f120,f44
+ cmp.ltu p6,p0=r25,r24 }
+{ .mfi; xma.lu f44=f36,f120,f44
+ add r26=r26,r25 };;
+{ .mfi; getf.sig r16=f80
+ xma.hu f55=f36,f121,f54
+(p6) add carry1=1,carry1 }
+{ .mfi; xma.lu f54=f36,f121,f54 };;
+{ .mfi; getf.sig r17=f71
+ xma.hu f65=f36,f122,f64
+ cmp.ltu p6,p0=r26,r25 }
+{ .mfi; xma.lu f64=f36,f122,f64
+ add r27=r27,r26 };;
+{ .mfi; getf.sig r18=f62
+ xma.hu f75=f36,f123,f74
+(p6) add carry1=1,carry1 }
+{ .mfi; cmp.ltu p6,p0=r27,r26
+ xma.lu f74=f36,f123,f74
+ add r27=r27,carry2 };;
+{ .mfi; getf.sig r19=f53
+ xma.hu f85=f36,f124,f84
+(p6) add carry1=1,carry1 }
+{ .mfi; xma.lu f84=f36,f124,f84
+ cmp.ltu p6,p0=r27,carry2 };;
+{ .mfi; st8 [r33]=r27,16
+ xma.hu f95=f36,f125,f94
+(p6) add carry1=1,carry1 }
+{ .mfi; xma.lu f94=f36,f125,f94 };;
+{ .mfi; xma.hu f105=f36,f126,f104 }
+{ .mfi; mov carry2=0
+ xma.lu f104=f36,f126,f104
+ add r17=r17,r16 };;
+{ .mfi; xma.hu f115=f36,f127,f114
+ cmp.ltu p7,p0=r17,r16 }
+{ .mfi; xma.lu f114=f36,f127,f114
+ add r18=r18,r17 };;//
+//-------------------------------------------------//
+{ .mfi; getf.sig r20=f44
+ xma.hu f46=f37,f120,f45
+(p7) add carry2=1,carry2 }
+{ .mfi; cmp.ltu p7,p0=r18,r17
+ xma.lu f45=f37,f120,f45
+ add r19=r19,r18 };;
+{ .mfi; getf.sig r24=f90
+ xma.hu f56=f37,f121,f55 }
+{ .mfi; xma.lu f55=f37,f121,f55 };;
+{ .mfi; getf.sig r25=f81
+ xma.hu f66=f37,f122,f65
+(p7) add carry2=1,carry2 }
+{ .mfi; cmp.ltu p7,p0=r19,r18
+ xma.lu f65=f37,f122,f65
+ add r20=r20,r19 };;
+{ .mfi; getf.sig r26=f72
+ xma.hu f76=f37,f123,f75
+(p7) add carry2=1,carry2 }
+{ .mfi; cmp.ltu p7,p0=r20,r19
+ xma.lu f75=f37,f123,f75
+ add r20=r20,carry1 };;
+{ .mfi; getf.sig r27=f63
+ xma.hu f86=f37,f124,f85
+(p7) add carry2=1,carry2 }
+{ .mfi; xma.lu f85=f37,f124,f85
+ cmp.ltu p7,p0=r20,carry1 };;
+{ .mfi; getf.sig r28=f54
+ xma.hu f96=f37,f125,f95
+(p7) add carry2=1,carry2 }
+{ .mfi; st8 [r32]=r20,16
+ xma.lu f95=f37,f125,f95 };;
+{ .mfi; xma.hu f106=f37,f126,f105 }
+{ .mfi; mov carry1=0
+ xma.lu f105=f37,f126,f105
+ add r25=r25,r24 };;
+{ .mfi; xma.hu f116=f37,f127,f115
+ cmp.ltu p6,p0=r25,r24 }
+{ .mfi; xma.lu f115=f37,f127,f115
+ add r26=r26,r25 };;//
+//-------------------------------------------------//
+{ .mfi; getf.sig r29=f45
+ xma.hu f47=f38,f120,f46
+(p6) add carry1=1,carry1 }
+{ .mfi; cmp.ltu p6,p0=r26,r25
+ xma.lu f46=f38,f120,f46
+ add r27=r27,r26 };;
+{ .mfi; getf.sig r16=f100
+ xma.hu f57=f38,f121,f56
+(p6) add carry1=1,carry1 }
+{ .mfi; cmp.ltu p6,p0=r27,r26
+ xma.lu f56=f38,f121,f56
+ add r28=r28,r27 };;
+{ .mfi; getf.sig r17=f91
+ xma.hu f67=f38,f122,f66
+(p6) add carry1=1,carry1 }
+{ .mfi; cmp.ltu p6,p0=r28,r27
+ xma.lu f66=f38,f122,f66
+ add r29=r29,r28 };;
+{ .mfi; getf.sig r18=f82
+ xma.hu f77=f38,f123,f76
+(p6) add carry1=1,carry1 }
+{ .mfi; cmp.ltu p6,p0=r29,r28
+ xma.lu f76=f38,f123,f76
+ add r29=r29,carry2 };;
+{ .mfi; getf.sig r19=f73
+ xma.hu f87=f38,f124,f86
+(p6) add carry1=1,carry1 }
+{ .mfi; xma.lu f86=f38,f124,f86
+ cmp.ltu p6,p0=r29,carry2 };;
+{ .mfi; getf.sig r20=f64
+ xma.hu f97=f38,f125,f96
+(p6) add carry1=1,carry1 }
+{ .mfi; st8 [r33]=r29,16
+ xma.lu f96=f38,f125,f96 };;
+{ .mfi; getf.sig r21=f55
+ xma.hu f107=f38,f126,f106 }
+{ .mfi; mov carry2=0
+ xma.lu f106=f38,f126,f106
+ add r17=r17,r16 };;
+{ .mfi; xma.hu f117=f38,f127,f116
+ cmp.ltu p7,p0=r17,r16 }
+{ .mfi; xma.lu f116=f38,f127,f116
+ add r18=r18,r17 };;//
+//-------------------------------------------------//
+{ .mfi; getf.sig r22=f46
+ xma.hu f48=f39,f120,f47
+(p7) add carry2=1,carry2 }
+{ .mfi; cmp.ltu p7,p0=r18,r17
+ xma.lu f47=f39,f120,f47
+ add r19=r19,r18 };;
+{ .mfi; getf.sig r24=f110
+ xma.hu f58=f39,f121,f57
+(p7) add carry2=1,carry2 }
+{ .mfi; cmp.ltu p7,p0=r19,r18
+ xma.lu f57=f39,f121,f57
+ add r20=r20,r19 };;
+{ .mfi; getf.sig r25=f101
+ xma.hu f68=f39,f122,f67
+(p7) add carry2=1,carry2 }
+{ .mfi; cmp.ltu p7,p0=r20,r19
+ xma.lu f67=f39,f122,f67
+ add r21=r21,r20 };;
+{ .mfi; getf.sig r26=f92
+ xma.hu f78=f39,f123,f77
+(p7) add carry2=1,carry2 }
+{ .mfi; cmp.ltu p7,p0=r21,r20
+ xma.lu f77=f39,f123,f77
+ add r22=r22,r21 };;
+{ .mfi; getf.sig r27=f83
+ xma.hu f88=f39,f124,f87
+(p7) add carry2=1,carry2 }
+{ .mfi; cmp.ltu p7,p0=r22,r21
+ xma.lu f87=f39,f124,f87
+ add r22=r22,carry1 };;
+{ .mfi; getf.sig r28=f74
+ xma.hu f98=f39,f125,f97
+(p7) add carry2=1,carry2 }
+{ .mfi; xma.lu f97=f39,f125,f97
+ cmp.ltu p7,p0=r22,carry1 };;
+{ .mfi; getf.sig r29=f65
+ xma.hu f108=f39,f126,f107
+(p7) add carry2=1,carry2 }
+{ .mfi; st8 [r32]=r22,16
+ xma.lu f107=f39,f126,f107 };;
+{ .mfi; getf.sig r30=f56
+ xma.hu f118=f39,f127,f117 }
+{ .mfi; xma.lu f117=f39,f127,f117 };;//
+//-------------------------------------------------//
+// Leaving multiplier's heaven... Quite a ride, huh?
+
+{ .mii; getf.sig r31=f47
+ add r25=r25,r24
+ mov carry1=0 };;
+{ .mii; getf.sig r16=f111
+ cmp.ltu p6,p0=r25,r24
+ add r26=r26,r25 };;
+{ .mfb; getf.sig r17=f102 }
+{ .mii;
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r26,r25
+ add r27=r27,r26 };;
+{ .mfb; nop.m 0x0 }
+{ .mii;
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r27,r26
+ add r28=r28,r27 };;
+{ .mii; getf.sig r18=f93
+ add r17=r17,r16
+ mov carry3=0 }
+{ .mii;
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r28,r27
+ add r29=r29,r28 };;
+{ .mii; getf.sig r19=f84
+ cmp.ltu p7,p0=r17,r16 }
+{ .mii;
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r29,r28
+ add r30=r30,r29 };;
+{ .mii; getf.sig r20=f75
+ add r18=r18,r17 }
+{ .mii;
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r30,r29
+ add r31=r31,r30 };;
+{ .mfb; getf.sig r21=f66 }
+{ .mii; (p7) add carry3=1,carry3
+ cmp.ltu p7,p0=r18,r17
+ add r19=r19,r18 }
+{ .mfb; nop.m 0x0 }
+{ .mii;
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r31,r30
+ add r31=r31,carry2 };;
+{ .mfb; getf.sig r22=f57 }
+{ .mii; (p7) add carry3=1,carry3
+ cmp.ltu p7,p0=r19,r18
+ add r20=r20,r19 }
+{ .mfb; nop.m 0x0 }
+{ .mii;
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r31,carry2 };;
+{ .mfb; getf.sig r23=f48 }
+{ .mii; (p7) add carry3=1,carry3
+ cmp.ltu p7,p0=r20,r19
+ add r21=r21,r20 }
+{ .mii;
+(p6) add carry1=1,carry1 }
+{ .mfb; st8 [r33]=r31,16 };;
+
+{ .mfb; getf.sig r24=f112 }
+{ .mii; (p7) add carry3=1,carry3
+ cmp.ltu p7,p0=r21,r20
+ add r22=r22,r21 };;
+{ .mfb; getf.sig r25=f103 }
+{ .mii; (p7) add carry3=1,carry3
+ cmp.ltu p7,p0=r22,r21
+ add r23=r23,r22 };;
+{ .mfb; getf.sig r26=f94 }
+{ .mii; (p7) add carry3=1,carry3
+ cmp.ltu p7,p0=r23,r22
+ add r23=r23,carry1 };;
+{ .mfb; getf.sig r27=f85 }
+{ .mii; (p7) add carry3=1,carry3
+ cmp.ltu p7,p8=r23,carry1};;
+{ .mii; getf.sig r28=f76
+ add r25=r25,r24
+ mov carry1=0 }
+{ .mii; st8 [r32]=r23,16
+ (p7) add carry2=1,carry3
+ (p8) add carry2=0,carry3 };;
+
+{ .mfb; nop.m 0x0 }
+{ .mii; getf.sig r29=f67
+ cmp.ltu p6,p0=r25,r24
+ add r26=r26,r25 };;
+{ .mfb; getf.sig r30=f58 }
+{ .mii;
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r26,r25
+ add r27=r27,r26 };;
+{ .mfb; getf.sig r16=f113 }
+{ .mii;
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r27,r26
+ add r28=r28,r27 };;
+{ .mfb; getf.sig r17=f104 }
+{ .mii;
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r28,r27
+ add r29=r29,r28 };;
+{ .mfb; getf.sig r18=f95 }
+{ .mii;
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r29,r28
+ add r30=r30,r29 };;
+{ .mii; getf.sig r19=f86
+ add r17=r17,r16
+ mov carry3=0 }
+{ .mii;
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r30,r29
+ add r30=r30,carry2 };;
+{ .mii; getf.sig r20=f77
+ cmp.ltu p7,p0=r17,r16
+ add r18=r18,r17 }
+{ .mii;
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r30,carry2 };;
+{ .mfb; getf.sig r21=f68 }
+{ .mii; st8 [r33]=r30,16
+(p6) add carry1=1,carry1 };;
+
+{ .mfb; getf.sig r24=f114 }
+{ .mii; (p7) add carry3=1,carry3
+ cmp.ltu p7,p0=r18,r17
+ add r19=r19,r18 };;
+{ .mfb; getf.sig r25=f105 }
+{ .mii; (p7) add carry3=1,carry3
+ cmp.ltu p7,p0=r19,r18
+ add r20=r20,r19 };;
+{ .mfb; getf.sig r26=f96 }
+{ .mii; (p7) add carry3=1,carry3
+ cmp.ltu p7,p0=r20,r19
+ add r21=r21,r20 };;
+{ .mfb; getf.sig r27=f87 }
+{ .mii; (p7) add carry3=1,carry3
+ cmp.ltu p7,p0=r21,r20
+ add r21=r21,carry1 };;
+{ .mib; getf.sig r28=f78
+ add r25=r25,r24 }
+{ .mib; (p7) add carry3=1,carry3
+ cmp.ltu p7,p8=r21,carry1};;
+{ .mii; st8 [r32]=r21,16
+ (p7) add carry2=1,carry3
+ (p8) add carry2=0,carry3 }
+
+{ .mii; mov carry1=0
+ cmp.ltu p6,p0=r25,r24
+ add r26=r26,r25 };;
+{ .mfb; getf.sig r16=f115 }
+{ .mii;
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r26,r25
+ add r27=r27,r26 };;
+{ .mfb; getf.sig r17=f106 }
+{ .mii;
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r27,r26
+ add r28=r28,r27 };;
+{ .mfb; getf.sig r18=f97 }
+{ .mii;
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r28,r27
+ add r28=r28,carry2 };;
+{ .mib; getf.sig r19=f88
+ add r17=r17,r16 }
+{ .mib;
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r28,carry2 };;
+{ .mii; st8 [r33]=r28,16
+(p6) add carry1=1,carry1 }
+
+{ .mii; mov carry2=0
+ cmp.ltu p7,p0=r17,r16
+ add r18=r18,r17 };;
+{ .mfb; getf.sig r24=f116 }
+{ .mii; (p7) add carry2=1,carry2
+ cmp.ltu p7,p0=r18,r17
+ add r19=r19,r18 };;
+{ .mfb; getf.sig r25=f107 }
+{ .mii; (p7) add carry2=1,carry2
+ cmp.ltu p7,p0=r19,r18
+ add r19=r19,carry1 };;
+{ .mfb; getf.sig r26=f98 }
+{ .mii; (p7) add carry2=1,carry2
+ cmp.ltu p7,p0=r19,carry1};;
+{ .mii; st8 [r32]=r19,16
+ (p7) add carry2=1,carry2 }
+
+{ .mfb; add r25=r25,r24 };;
+
+{ .mfb; getf.sig r16=f117 }
+{ .mii; mov carry1=0
+ cmp.ltu p6,p0=r25,r24
+ add r26=r26,r25 };;
+{ .mfb; getf.sig r17=f108 }
+{ .mii;
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r26,r25
+ add r26=r26,carry2 };;
+{ .mfb; nop.m 0x0 }
+{ .mii;
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r26,carry2 };;
+{ .mii; st8 [r33]=r26,16
+(p6) add carry1=1,carry1 }
+
+{ .mfb; add r17=r17,r16 };;
+{ .mfb; getf.sig r24=f118 }
+{ .mii; mov carry2=0
+ cmp.ltu p7,p0=r17,r16
+ add r17=r17,carry1 };;
+{ .mii; (p7) add carry2=1,carry2
+ cmp.ltu p7,p0=r17,carry1};;
+{ .mii; st8 [r32]=r17
+ (p7) add carry2=1,carry2 };;
+{ .mfb; add r24=r24,carry2 };;
+{ .mib; st8 [r33]=r24 }
+
+{ .mib; rum 1<<5 // clear um.mfh
+ br.ret.sptk.many b0 };;
+.endp bn_mul_comba8#
+#undef carry3
+#undef carry2
+#undef carry1
+#endif
+
+#if 1
+// It's possible to make it faster (see comment to bn_sqr_comba8), but
+// I reckon it doesn't worth the effort. Basically because the routine
+// (actually both of them) practically never called... So I just play
+// same trick as with bn_sqr_comba8.
+//
+// void bn_sqr_comba4(BN_ULONG *r, BN_ULONG *a)
+//
+.global bn_sqr_comba4#
+.proc bn_sqr_comba4#
+.align 64
+bn_sqr_comba4:
+ .prologue
+ .save ar.pfs,r2
+#if defined(_HPUX_SOURCE) && !defined(_LP64)
+{ .mii; alloc r2=ar.pfs,2,1,0,0
+ addp4 r32=0,r32
+ addp4 r33=0,r33 };;
+{ .mii;
+#else
+{ .mii; alloc r2=ar.pfs,2,1,0,0
+#endif
+ mov r34=r33
+ add r14=8,r33 };;
+ .body
+{ .mii; add r17=8,r34
+ add r15=16,r33
+ add r18=16,r34 }
+{ .mfb; add r16=24,r33
+ br .L_cheat_entry_point4 };;
+.endp bn_sqr_comba4#
+#endif
+
+#if 1
+// Runs in ~115 cycles and ~4.5 times faster than C. Well, whatever...
+//
+// void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
+//
+#define carry1 r14
+#define carry2 r15
+.global bn_mul_comba4#
+.proc bn_mul_comba4#
+.align 64
+bn_mul_comba4:
+ .prologue
+ .save ar.pfs,r2
+#if defined(_HPUX_SOURCE) && !defined(_LP64)
+{ .mii; alloc r2=ar.pfs,3,0,0,0
+ addp4 r33=0,r33
+ addp4 r34=0,r34 };;
+{ .mii; addp4 r32=0,r32
+#else
+{ .mii; alloc r2=ar.pfs,3,0,0,0
+#endif
+ add r14=8,r33
+ add r17=8,r34 }
+ .body
+{ .mii; add r15=16,r33
+ add r18=16,r34
+ add r16=24,r33 };;
+.L_cheat_entry_point4:
+{ .mmi; add r19=24,r34
+
+ ldf8 f32=[r33] }
+
+{ .mmi; ldf8 f120=[r34]
+ ldf8 f121=[r17] };;
+{ .mmi; ldf8 f122=[r18]
+ ldf8 f123=[r19] }
+
+{ .mmi; ldf8 f33=[r14]
+ ldf8 f34=[r15] }
+{ .mfi; ldf8 f35=[r16]
+
+ xma.hu f41=f32,f120,f0 }
+{ .mfi; xma.lu f40=f32,f120,f0 };;
+{ .mfi; xma.hu f51=f32,f121,f0 }
+{ .mfi; xma.lu f50=f32,f121,f0 };;
+{ .mfi; xma.hu f61=f32,f122,f0 }
+{ .mfi; xma.lu f60=f32,f122,f0 };;
+{ .mfi; xma.hu f71=f32,f123,f0 }
+{ .mfi; xma.lu f70=f32,f123,f0 };;//
+// Major stall takes place here, and 3 more places below. Result from
+// first xma is not available for another 3 ticks.
+{ .mfi; getf.sig r16=f40
+ xma.hu f42=f33,f120,f41
+ add r33=8,r32 }
+{ .mfi; xma.lu f41=f33,f120,f41 };;
+{ .mfi; getf.sig r24=f50
+ xma.hu f52=f33,f121,f51 }
+{ .mfi; xma.lu f51=f33,f121,f51 };;
+{ .mfi; st8 [r32]=r16,16
+ xma.hu f62=f33,f122,f61 }
+{ .mfi; xma.lu f61=f33,f122,f61 };;
+{ .mfi; xma.hu f72=f33,f123,f71 }
+{ .mfi; xma.lu f71=f33,f123,f71 };;//
+//-------------------------------------------------//
+{ .mfi; getf.sig r25=f41
+ xma.hu f43=f34,f120,f42 }
+{ .mfi; xma.lu f42=f34,f120,f42 };;
+{ .mfi; getf.sig r16=f60
+ xma.hu f53=f34,f121,f52 }
+{ .mfi; xma.lu f52=f34,f121,f52 };;
+{ .mfi; getf.sig r17=f51
+ xma.hu f63=f34,f122,f62
+ add r25=r25,r24 }
+{ .mfi; mov carry1=0
+ xma.lu f62=f34,f122,f62 };;
+{ .mfi; st8 [r33]=r25,16
+ xma.hu f73=f34,f123,f72
+ cmp.ltu p6,p0=r25,r24 }
+{ .mfi; xma.lu f72=f34,f123,f72 };;//
+//-------------------------------------------------//
+{ .mfi; getf.sig r18=f42
+ xma.hu f44=f35,f120,f43
+(p6) add carry1=1,carry1 }
+{ .mfi; add r17=r17,r16
+ xma.lu f43=f35,f120,f43
+ mov carry2=0 };;
+{ .mfi; getf.sig r24=f70
+ xma.hu f54=f35,f121,f53
+ cmp.ltu p7,p0=r17,r16 }
+{ .mfi; xma.lu f53=f35,f121,f53 };;
+{ .mfi; getf.sig r25=f61
+ xma.hu f64=f35,f122,f63
+ add r18=r18,r17 }
+{ .mfi; xma.lu f63=f35,f122,f63
+(p7) add carry2=1,carry2 };;
+{ .mfi; getf.sig r26=f52
+ xma.hu f74=f35,f123,f73
+ cmp.ltu p7,p0=r18,r17 }
+{ .mfi; xma.lu f73=f35,f123,f73
+ add r18=r18,carry1 };;
+//-------------------------------------------------//
+{ .mii; st8 [r32]=r18,16
+(p7) add carry2=1,carry2
+ cmp.ltu p7,p0=r18,carry1 };;
+
+{ .mfi; getf.sig r27=f43 // last major stall
+(p7) add carry2=1,carry2 };;
+{ .mii; getf.sig r16=f71
+ add r25=r25,r24
+ mov carry1=0 };;
+{ .mii; getf.sig r17=f62
+ cmp.ltu p6,p0=r25,r24
+ add r26=r26,r25 };;
+{ .mii;
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r26,r25
+ add r27=r27,r26 };;
+{ .mii;
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r27,r26
+ add r27=r27,carry2 };;
+{ .mii; getf.sig r18=f53
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r27,carry2 };;
+{ .mfi; st8 [r33]=r27,16
+(p6) add carry1=1,carry1 }
+
+{ .mii; getf.sig r19=f44
+ add r17=r17,r16
+ mov carry2=0 };;
+{ .mii; getf.sig r24=f72
+ cmp.ltu p7,p0=r17,r16
+ add r18=r18,r17 };;
+{ .mii; (p7) add carry2=1,carry2
+ cmp.ltu p7,p0=r18,r17
+ add r19=r19,r18 };;
+{ .mii; (p7) add carry2=1,carry2
+ cmp.ltu p7,p0=r19,r18
+ add r19=r19,carry1 };;
+{ .mii; getf.sig r25=f63
+ (p7) add carry2=1,carry2
+ cmp.ltu p7,p0=r19,carry1};;
+{ .mii; st8 [r32]=r19,16
+ (p7) add carry2=1,carry2 }
+
+{ .mii; getf.sig r26=f54
+ add r25=r25,r24
+ mov carry1=0 };;
+{ .mii; getf.sig r16=f73
+ cmp.ltu p6,p0=r25,r24
+ add r26=r26,r25 };;
+{ .mii;
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r26,r25
+ add r26=r26,carry2 };;
+{ .mii; getf.sig r17=f64
+(p6) add carry1=1,carry1
+ cmp.ltu p6,p0=r26,carry2 };;
+{ .mii; st8 [r33]=r26,16
+(p6) add carry1=1,carry1 }
+
+{ .mii; getf.sig r24=f74
+ add r17=r17,r16
+ mov carry2=0 };;
+{ .mii; cmp.ltu p7,p0=r17,r16
+ add r17=r17,carry1 };;
+
+{ .mii; (p7) add carry2=1,carry2
+ cmp.ltu p7,p0=r17,carry1};;
+{ .mii; st8 [r32]=r17,16
+ (p7) add carry2=1,carry2 };;
+
+{ .mii; add r24=r24,carry2 };;
+{ .mii; st8 [r33]=r24 }
+
+{ .mib; rum 1<<5 // clear um.mfh
+ br.ret.sptk.many b0 };;
+.endp bn_mul_comba4#
+#undef carry2
+#undef carry1
+#endif
+
+#if 1
+//
+// BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d)
+//
+// In the nutshell it's a port of my MIPS III/IV implementation.
+//
+#define AT r14
+#define H r16
+#define HH r20
+#define L r17
+#define D r18
+#define DH r22
+#define I r21
+
+#if 0
+// Some preprocessors (most notably HP-UX) appear to be allergic to
+// macros enclosed to parenthesis [as these three were].
+#define cont p16
+#define break p0 // p20
+#define equ p24
+#else
+cont=p16
+break=p0
+equ=p24
+#endif
+
+.global abort#
+.global bn_div_words#
+.proc bn_div_words#
+.align 64
+bn_div_words:
+ .prologue
+ .save ar.pfs,r2
+{ .mii; alloc r2=ar.pfs,3,5,0,8
+ .save b0,r3
+ mov r3=b0
+ .save pr,r10
+ mov r10=pr };;
+{ .mmb; cmp.eq p6,p0=r34,r0
+ mov r8=-1
+(p6) br.ret.spnt.many b0 };;
+
+ .body
+{ .mii; mov H=r32 // save h
+ mov ar.ec=0 // don't rotate at exit
+ mov pr.rot=0 }
+{ .mii; mov L=r33 // save l
+ mov r25=r0 // needed if abort is called on VMS
+ mov r36=r0 };;
+
+.L_divw_shift: // -vv- note signed comparison
+{ .mfi; (p0) cmp.lt p16,p0=r0,r34 // d
+ (p0) shladd r33=r34,1,r0 }
+{ .mfb; (p0) add r35=1,r36
+ (p0) nop.f 0x0
+(p16) br.wtop.dpnt .L_divw_shift };;
+
+{ .mii; mov D=r34
+ shr.u DH=r34,32
+ sub r35=64,r36 };;
+{ .mii; setf.sig f7=DH
+ shr.u AT=H,r35
+ mov I=r36 };;
+{ .mib; cmp.ne p6,p0=r0,AT
+ shl H=H,r36
+(p6) br.call.spnt.clr b0=abort };; // overflow, die...
+
+{ .mfi; fcvt.xuf.s1 f7=f7
+ shr.u AT=L,r35 };;
+{ .mii; shl L=L,r36
+ or H=H,AT };;
+
+{ .mii; nop.m 0x0
+ cmp.leu p6,p0=D,H;;
+(p6) sub H=H,D }
+
+{ .mlx; setf.sig f14=D
+ movl AT=0xffffffff };;
+///////////////////////////////////////////////////////////
+{ .mii; setf.sig f6=H
+ shr.u HH=H,32;;
+ cmp.eq p6,p7=HH,DH };;
+{ .mfb;
+(p6) setf.sig f8=AT
+(p7) fcvt.xuf.s1 f6=f6
+(p7) br.call.sptk b6=.L_udiv64_32_b6 };;
+
+{ .mfi; getf.sig r33=f8 // q
+ xmpy.lu f9=f8,f14 }
+{ .mfi; xmpy.hu f10=f8,f14
+ shrp H=H,L,32 };;
+
+{ .mmi; getf.sig r35=f9 // tl
+ getf.sig r31=f10 };; // th
+
+.L_divw_1st_iter:
+{ .mii; (p0) add r32=-1,r33
+ (p0) cmp.eq equ,cont=HH,r31 };;
+{ .mii; (p0) cmp.ltu p8,p0=r35,D
+ (p0) sub r34=r35,D
+ (equ) cmp.leu break,cont=r35,H };;
+{ .mib; (cont) cmp.leu cont,break=HH,r31
+ (p8) add r31=-1,r31
+(cont) br.wtop.spnt .L_divw_1st_iter };;
+///////////////////////////////////////////////////////////
+{ .mii; sub H=H,r35
+ shl r8=r33,32
+ shl L=L,32 };;
+///////////////////////////////////////////////////////////
+{ .mii; setf.sig f6=H
+ shr.u HH=H,32;;
+ cmp.eq p6,p7=HH,DH };;
+{ .mfb;
+(p6) setf.sig f8=AT
+(p7) fcvt.xuf.s1 f6=f6
+(p7) br.call.sptk b6=.L_udiv64_32_b6 };;
+
+{ .mfi; getf.sig r33=f8 // q
+ xmpy.lu f9=f8,f14 }
+{ .mfi; xmpy.hu f10=f8,f14
+ shrp H=H,L,32 };;
+
+{ .mmi; getf.sig r35=f9 // tl
+ getf.sig r31=f10 };; // th
+
+.L_divw_2nd_iter:
+{ .mii; (p0) add r32=-1,r33
+ (p0) cmp.eq equ,cont=HH,r31 };;
+{ .mii; (p0) cmp.ltu p8,p0=r35,D
+ (p0) sub r34=r35,D
+ (equ) cmp.leu break,cont=r35,H };;
+{ .mib; (cont) cmp.leu cont,break=HH,r31
+ (p8) add r31=-1,r31
+(cont) br.wtop.spnt .L_divw_2nd_iter };;
+///////////////////////////////////////////////////////////
+{ .mii; sub H=H,r35
+ or r8=r8,r33
+ mov ar.pfs=r2 };;
+{ .mii; shr.u r9=H,I // remainder if anybody wants it
+ mov pr=r10,0x1ffff }
+{ .mfb; br.ret.sptk.many b0 };;
+
+// Unsigned 64 by 32 (well, by 64 for the moment) bit integer division
+// procedure.
+//
+// inputs: f6 = (double)a, f7 = (double)b
+// output: f8 = (int)(a/b)
+// clobbered: f8,f9,f10,f11,pred
+pred=p15
+// This snippet is based on text found in the "Divide, Square
+// Root and Remainder" section at
+// http://www.intel.com/software/products/opensource/libraries/num.htm.
+// Yes, I admit that the referred code was used as template,
+// but after I realized that there hardly is any other instruction
+// sequence which would perform this operation. I mean I figure that
+// any independent attempt to implement high-performance division
+// will result in code virtually identical to the Intel code. It
+// should be noted though that below division kernel is 1 cycle
+// faster than Intel one (note commented splits:-), not to mention
+// original prologue (rather lack of one) and epilogue.
+.align 32
+.skip 16
+.L_udiv64_32_b6:
+ frcpa.s1 f8,pred=f6,f7;; // [0] y0 = 1 / b
+
+(pred) fnma.s1 f9=f7,f8,f1 // [5] e0 = 1 - b * y0
+(pred) fmpy.s1 f10=f6,f8;; // [5] q0 = a * y0
+(pred) fmpy.s1 f11=f9,f9 // [10] e1 = e0 * e0
+(pred) fma.s1 f10=f9,f10,f10;; // [10] q1 = q0 + e0 * q0
+(pred) fma.s1 f8=f9,f8,f8 //;; // [15] y1 = y0 + e0 * y0
+(pred) fma.s1 f9=f11,f10,f10;; // [15] q2 = q1 + e1 * q1
+(pred) fma.s1 f8=f11,f8,f8 //;; // [20] y2 = y1 + e1 * y1
+(pred) fnma.s1 f10=f7,f9,f6;; // [20] r2 = a - b * q2
+(pred) fma.s1 f8=f10,f8,f9;; // [25] q3 = q2 + r2 * y2
+
+ fcvt.fxu.trunc.s1 f8=f8 // [30] q = trunc(q3)
+ br.ret.sptk.many b6;;
+.endp bn_div_words#
+#endif
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/mips-mont.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/mips-mont.pl
new file mode 100644
index 0000000..3b79a4b
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/mips-mont.pl
@@ -0,0 +1,433 @@
+#! /usr/bin/env perl
+# Copyright 2010-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+#
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+
+# This module doesn't present direct interest for OpenSSL, because it
+# doesn't provide better performance for longer keys, at least not on
+# in-order-execution cores. While 512-bit RSA sign operations can be
+# 65% faster in 64-bit mode, 1024-bit ones are only 15% faster, and
+# 4096-bit ones are up to 15% slower. In 32-bit mode it varies from
+# 16% improvement for 512-bit RSA sign to -33% for 4096-bit RSA
+# verify:-( All comparisons are against bn_mul_mont-free assembler.
+# The module might be of interest to embedded system developers, as
+# the code is smaller than 1KB, yet offers >3x improvement on MIPS64
+# and 75-30% [less for longer keys] on MIPS32 over compiler-generated
+# code.
+
+######################################################################
+# There is a number of MIPS ABI in use, O32 and N32/64 are most
+# widely used. Then there is a new contender: NUBI. It appears that if
+# one picks the latter, it's possible to arrange code in ABI neutral
+# manner. Therefore let's stick to NUBI register layout:
+#
+($zero,$at,$t0,$t1,$t2)=map("\$$_",(0..2,24,25));
+($a0,$a1,$a2,$a3,$a4,$a5,$a6,$a7)=map("\$$_",(4..11));
+($s0,$s1,$s2,$s3,$s4,$s5,$s6,$s7,$s8,$s9,$s10,$s11)=map("\$$_",(12..23));
+($gp,$tp,$sp,$fp,$ra)=map("\$$_",(3,28..31));
+#
+# The return value is placed in $a0. Following coding rules facilitate
+# interoperability:
+#
+# - never ever touch $tp, "thread pointer", former $gp;
+# - copy return value to $t0, former $v0 [or to $a0 if you're adapting
+# old code];
+# - on O32 populate $a4-$a7 with 'lw $aN,4*N($sp)' if necessary;
+#
+# For reference here is register layout for N32/64 MIPS ABIs:
+#
+# ($zero,$at,$v0,$v1)=map("\$$_",(0..3));
+# ($a0,$a1,$a2,$a3,$a4,$a5,$a6,$a7)=map("\$$_",(4..11));
+# ($t0,$t1,$t2,$t3,$t8,$t9)=map("\$$_",(12..15,24,25));
+# ($s0,$s1,$s2,$s3,$s4,$s5,$s6,$s7)=map("\$$_",(16..23));
+# ($gp,$sp,$fp,$ra)=map("\$$_",(28..31));
+#
+$flavour = shift || "o32"; # supported flavours are o32,n32,64,nubi32,nubi64
+
+if ($flavour =~ /64|n32/i) {
+ $PTR_ADD="daddu"; # incidentally works even on n32
+ $PTR_SUB="dsubu"; # incidentally works even on n32
+ $REG_S="sd";
+ $REG_L="ld";
+ $SZREG=8;
+} else {
+ $PTR_ADD="addu";
+ $PTR_SUB="subu";
+ $REG_S="sw";
+ $REG_L="lw";
+ $SZREG=4;
+}
+$SAVED_REGS_MASK = ($flavour =~ /nubi/i) ? 0x00fff000 : 0x00ff0000;
+#
+# <appro@openssl.org>
+#
+######################################################################
+
+while (($output=shift) && ($output!~/\w[\w\-]*\.\w+$/)) {}
+open STDOUT,">$output";
+
+if ($flavour =~ /64|n32/i) {
+ $LD="ld";
+ $ST="sd";
+ $MULTU="dmultu";
+ $ADDU="daddu";
+ $SUBU="dsubu";
+ $BNSZ=8;
+} else {
+ $LD="lw";
+ $ST="sw";
+ $MULTU="multu";
+ $ADDU="addu";
+ $SUBU="subu";
+ $BNSZ=4;
+}
+
+# int bn_mul_mont(
+$rp=$a0; # BN_ULONG *rp,
+$ap=$a1; # const BN_ULONG *ap,
+$bp=$a2; # const BN_ULONG *bp,
+$np=$a3; # const BN_ULONG *np,
+$n0=$a4; # const BN_ULONG *n0,
+$num=$a5; # int num);
+
+$lo0=$a6;
+$hi0=$a7;
+$lo1=$t1;
+$hi1=$t2;
+$aj=$s0;
+$bi=$s1;
+$nj=$s2;
+$tp=$s3;
+$alo=$s4;
+$ahi=$s5;
+$nlo=$s6;
+$nhi=$s7;
+$tj=$s8;
+$i=$s9;
+$j=$s10;
+$m1=$s11;
+
+$FRAMESIZE=14;
+
+$code=<<___;
+#include "mips_arch.h"
+
+.text
+
+.set noat
+.set noreorder
+
+.align 5
+.globl bn_mul_mont
+.ent bn_mul_mont
+bn_mul_mont:
+___
+$code.=<<___ if ($flavour =~ /o32/i);
+ lw $n0,16($sp)
+ lw $num,20($sp)
+___
+$code.=<<___;
+ slt $at,$num,4
+ bnez $at,1f
+ li $t0,0
+ slt $at,$num,17 # on in-order CPU
+ bnez $at,bn_mul_mont_internal
+ nop
+1: jr $ra
+ li $a0,0
+.end bn_mul_mont
+
+.align 5
+.ent bn_mul_mont_internal
+bn_mul_mont_internal:
+ .frame $fp,$FRAMESIZE*$SZREG,$ra
+ .mask 0x40000000|$SAVED_REGS_MASK,-$SZREG
+ $PTR_SUB $sp,$FRAMESIZE*$SZREG
+ $REG_S $fp,($FRAMESIZE-1)*$SZREG($sp)
+ $REG_S $s11,($FRAMESIZE-2)*$SZREG($sp)
+ $REG_S $s10,($FRAMESIZE-3)*$SZREG($sp)
+ $REG_S $s9,($FRAMESIZE-4)*$SZREG($sp)
+ $REG_S $s8,($FRAMESIZE-5)*$SZREG($sp)
+ $REG_S $s7,($FRAMESIZE-6)*$SZREG($sp)
+ $REG_S $s6,($FRAMESIZE-7)*$SZREG($sp)
+ $REG_S $s5,($FRAMESIZE-8)*$SZREG($sp)
+ $REG_S $s4,($FRAMESIZE-9)*$SZREG($sp)
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ $REG_S $s3,($FRAMESIZE-10)*$SZREG($sp)
+ $REG_S $s2,($FRAMESIZE-11)*$SZREG($sp)
+ $REG_S $s1,($FRAMESIZE-12)*$SZREG($sp)
+ $REG_S $s0,($FRAMESIZE-13)*$SZREG($sp)
+___
+$code.=<<___;
+ move $fp,$sp
+
+ .set reorder
+ $LD $n0,0($n0)
+ $LD $bi,0($bp) # bp[0]
+ $LD $aj,0($ap) # ap[0]
+ $LD $nj,0($np) # np[0]
+
+ $PTR_SUB $sp,2*$BNSZ # place for two extra words
+ sll $num,`log($BNSZ)/log(2)`
+ li $at,-4096
+ $PTR_SUB $sp,$num
+ and $sp,$at
+
+ $MULTU ($aj,$bi)
+ $LD $ahi,$BNSZ($ap)
+ $LD $nhi,$BNSZ($np)
+ mflo ($lo0,$aj,$bi)
+ mfhi ($hi0,$aj,$bi)
+ $MULTU ($lo0,$n0)
+ mflo ($m1,$lo0,$n0)
+
+ $MULTU ($ahi,$bi)
+ mflo ($alo,$ahi,$bi)
+ mfhi ($ahi,$ahi,$bi)
+
+ $MULTU ($nj,$m1)
+ mflo ($lo1,$nj,$m1)
+ mfhi ($hi1,$nj,$m1)
+ $MULTU ($nhi,$m1)
+ $ADDU $lo1,$lo0
+ sltu $at,$lo1,$lo0
+ $ADDU $hi1,$at
+ mflo ($nlo,$nhi,$m1)
+ mfhi ($nhi,$nhi,$m1)
+
+ move $tp,$sp
+ li $j,2*$BNSZ
+.align 4
+.L1st:
+ .set noreorder
+ $PTR_ADD $aj,$ap,$j
+ $PTR_ADD $nj,$np,$j
+ $LD $aj,($aj)
+ $LD $nj,($nj)
+
+ $MULTU ($aj,$bi)
+ $ADDU $lo0,$alo,$hi0
+ $ADDU $lo1,$nlo,$hi1
+ sltu $at,$lo0,$hi0
+ sltu $t0,$lo1,$hi1
+ $ADDU $hi0,$ahi,$at
+ $ADDU $hi1,$nhi,$t0
+ mflo ($alo,$aj,$bi)
+ mfhi ($ahi,$aj,$bi)
+
+ $ADDU $lo1,$lo0
+ sltu $at,$lo1,$lo0
+ $MULTU ($nj,$m1)
+ $ADDU $hi1,$at
+ addu $j,$BNSZ
+ $ST $lo1,($tp)
+ sltu $t0,$j,$num
+ mflo ($nlo,$nj,$m1)
+ mfhi ($nhi,$nj,$m1)
+
+ bnez $t0,.L1st
+ $PTR_ADD $tp,$BNSZ
+ .set reorder
+
+ $ADDU $lo0,$alo,$hi0
+ sltu $at,$lo0,$hi0
+ $ADDU $hi0,$ahi,$at
+
+ $ADDU $lo1,$nlo,$hi1
+ sltu $t0,$lo1,$hi1
+ $ADDU $hi1,$nhi,$t0
+ $ADDU $lo1,$lo0
+ sltu $at,$lo1,$lo0
+ $ADDU $hi1,$at
+
+ $ST $lo1,($tp)
+
+ $ADDU $hi1,$hi0
+ sltu $at,$hi1,$hi0
+ $ST $hi1,$BNSZ($tp)
+ $ST $at,2*$BNSZ($tp)
+
+ li $i,$BNSZ
+.align 4
+.Louter:
+ $PTR_ADD $bi,$bp,$i
+ $LD $bi,($bi)
+ $LD $aj,($ap)
+ $LD $ahi,$BNSZ($ap)
+ $LD $tj,($sp)
+
+ $MULTU ($aj,$bi)
+ $LD $nj,($np)
+ $LD $nhi,$BNSZ($np)
+ mflo ($lo0,$aj,$bi)
+ mfhi ($hi0,$aj,$bi)
+ $ADDU $lo0,$tj
+ $MULTU ($lo0,$n0)
+ sltu $at,$lo0,$tj
+ $ADDU $hi0,$at
+ mflo ($m1,$lo0,$n0)
+
+ $MULTU ($ahi,$bi)
+ mflo ($alo,$ahi,$bi)
+ mfhi ($ahi,$ahi,$bi)
+
+ $MULTU ($nj,$m1)
+ mflo ($lo1,$nj,$m1)
+ mfhi ($hi1,$nj,$m1)
+
+ $MULTU ($nhi,$m1)
+ $ADDU $lo1,$lo0
+ sltu $at,$lo1,$lo0
+ $ADDU $hi1,$at
+ mflo ($nlo,$nhi,$m1)
+ mfhi ($nhi,$nhi,$m1)
+
+ move $tp,$sp
+ li $j,2*$BNSZ
+ $LD $tj,$BNSZ($tp)
+.align 4
+.Linner:
+ .set noreorder
+ $PTR_ADD $aj,$ap,$j
+ $PTR_ADD $nj,$np,$j
+ $LD $aj,($aj)
+ $LD $nj,($nj)
+
+ $MULTU ($aj,$bi)
+ $ADDU $lo0,$alo,$hi0
+ $ADDU $lo1,$nlo,$hi1
+ sltu $at,$lo0,$hi0
+ sltu $t0,$lo1,$hi1
+ $ADDU $hi0,$ahi,$at
+ $ADDU $hi1,$nhi,$t0
+ mflo ($alo,$aj,$bi)
+ mfhi ($ahi,$aj,$bi)
+
+ $ADDU $lo0,$tj
+ addu $j,$BNSZ
+ $MULTU ($nj,$m1)
+ sltu $at,$lo0,$tj
+ $ADDU $lo1,$lo0
+ $ADDU $hi0,$at
+ sltu $t0,$lo1,$lo0
+ $LD $tj,2*$BNSZ($tp)
+ $ADDU $hi1,$t0
+ sltu $at,$j,$num
+ mflo ($nlo,$nj,$m1)
+ mfhi ($nhi,$nj,$m1)
+ $ST $lo1,($tp)
+ bnez $at,.Linner
+ $PTR_ADD $tp,$BNSZ
+ .set reorder
+
+ $ADDU $lo0,$alo,$hi0
+ sltu $at,$lo0,$hi0
+ $ADDU $hi0,$ahi,$at
+ $ADDU $lo0,$tj
+ sltu $t0,$lo0,$tj
+ $ADDU $hi0,$t0
+
+ $LD $tj,2*$BNSZ($tp)
+ $ADDU $lo1,$nlo,$hi1
+ sltu $at,$lo1,$hi1
+ $ADDU $hi1,$nhi,$at
+ $ADDU $lo1,$lo0
+ sltu $t0,$lo1,$lo0
+ $ADDU $hi1,$t0
+ $ST $lo1,($tp)
+
+ $ADDU $lo1,$hi1,$hi0
+ sltu $hi1,$lo1,$hi0
+ $ADDU $lo1,$tj
+ sltu $at,$lo1,$tj
+ $ADDU $hi1,$at
+ $ST $lo1,$BNSZ($tp)
+ $ST $hi1,2*$BNSZ($tp)
+
+ addu $i,$BNSZ
+ sltu $t0,$i,$num
+ bnez $t0,.Louter
+
+ .set noreorder
+ $PTR_ADD $tj,$sp,$num # &tp[num]
+ move $tp,$sp
+ move $ap,$sp
+ li $hi0,0 # clear borrow bit
+
+.align 4
+.Lsub: $LD $lo0,($tp)
+ $LD $lo1,($np)
+ $PTR_ADD $tp,$BNSZ
+ $PTR_ADD $np,$BNSZ
+ $SUBU $lo1,$lo0,$lo1 # tp[i]-np[i]
+ sgtu $at,$lo1,$lo0
+ $SUBU $lo0,$lo1,$hi0
+ sgtu $hi0,$lo0,$lo1
+ $ST $lo0,($rp)
+ or $hi0,$at
+ sltu $at,$tp,$tj
+ bnez $at,.Lsub
+ $PTR_ADD $rp,$BNSZ
+
+ $SUBU $hi0,$hi1,$hi0 # handle upmost overflow bit
+ move $tp,$sp
+ $PTR_SUB $rp,$num # restore rp
+ not $hi1,$hi0
+
+.Lcopy: $LD $nj,($tp) # conditional move
+ $LD $aj,($rp)
+ $ST $zero,($tp)
+ $PTR_ADD $tp,$BNSZ
+ and $nj,$hi0
+ and $aj,$hi1
+ or $aj,$nj
+ sltu $at,$tp,$tj
+ $ST $aj,($rp)
+ bnez $at,.Lcopy
+ $PTR_ADD $rp,$BNSZ
+
+ li $a0,1
+ li $t0,1
+
+ .set noreorder
+ move $sp,$fp
+ $REG_L $fp,($FRAMESIZE-1)*$SZREG($sp)
+ $REG_L $s11,($FRAMESIZE-2)*$SZREG($sp)
+ $REG_L $s10,($FRAMESIZE-3)*$SZREG($sp)
+ $REG_L $s9,($FRAMESIZE-4)*$SZREG($sp)
+ $REG_L $s8,($FRAMESIZE-5)*$SZREG($sp)
+ $REG_L $s7,($FRAMESIZE-6)*$SZREG($sp)
+ $REG_L $s6,($FRAMESIZE-7)*$SZREG($sp)
+ $REG_L $s5,($FRAMESIZE-8)*$SZREG($sp)
+ $REG_L $s4,($FRAMESIZE-9)*$SZREG($sp)
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ $REG_L $s3,($FRAMESIZE-10)*$SZREG($sp)
+ $REG_L $s2,($FRAMESIZE-11)*$SZREG($sp)
+ $REG_L $s1,($FRAMESIZE-12)*$SZREG($sp)
+ $REG_L $s0,($FRAMESIZE-13)*$SZREG($sp)
+___
+$code.=<<___;
+ jr $ra
+ $PTR_ADD $sp,$FRAMESIZE*$SZREG
+.end bn_mul_mont_internal
+.rdata
+.asciiz "Montgomery Multiplication for MIPS, CRYPTOGAMS by <appro\@openssl.org>"
+___
+
+$code =~ s/\`([^\`]*)\`/eval $1/gem;
+
+print $code;
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/mips.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/mips.pl
new file mode 100644
index 0000000..76fe823
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/mips.pl
@@ -0,0 +1,2267 @@
+#! /usr/bin/env perl
+# Copyright 2010-2021 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+#
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project.
+#
+# Rights for redistribution and usage in source and binary forms are
+# granted according to the OpenSSL license. Warranty of any kind is
+# disclaimed.
+# ====================================================================
+
+
+# July 1999
+#
+# This is drop-in MIPS III/IV ISA replacement for crypto/bn/bn_asm.c.
+#
+# The module is designed to work with either of the "new" MIPS ABI(5),
+# namely N32 or N64, offered by IRIX 6.x. It's not meant to work under
+# IRIX 5.x not only because it doesn't support new ABIs but also
+# because 5.x kernels put R4x00 CPU into 32-bit mode and all those
+# 64-bit instructions (daddu, dmultu, etc.) found below gonna only
+# cause illegal instruction exception:-(
+#
+# In addition the code depends on preprocessor flags set up by MIPSpro
+# compiler driver (either as or cc) and therefore (probably?) can't be
+# compiled by the GNU assembler. GNU C driver manages fine though...
+# I mean as long as -mmips-as is specified or is the default option,
+# because then it simply invokes /usr/bin/as which in turn takes
+# perfect care of the preprocessor definitions. Another neat feature
+# offered by the MIPSpro assembler is an optimization pass. This gave
+# me the opportunity to have the code looking more regular as all those
+# architecture dependent instruction rescheduling details were left to
+# the assembler. Cool, huh?
+#
+# Performance improvement is astonishing! 'apps/openssl speed rsa dsa'
+# goes way over 3 times faster!
+#
+# <appro@openssl.org>
+
+# October 2010
+#
+# Adapt the module even for 32-bit ABIs and other OSes. The former was
+# achieved by mechanical replacement of 64-bit arithmetic instructions
+# such as dmultu, daddu, etc. with their 32-bit counterparts and
+# adjusting offsets denoting multiples of BN_ULONG. Above mentioned
+# >3x performance improvement naturally does not apply to 32-bit code
+# [because there is no instruction 32-bit compiler can't use], one
+# has to content with 40-85% improvement depending on benchmark and
+# key length, more for longer keys.
+
+$flavour = shift || "o32";
+while (($output=shift) && ($output!~/\w[\w\-]*\.\w+$/)) {}
+open STDOUT,">$output";
+
+if ($flavour =~ /64|n32/i) {
+ $LD="ld";
+ $ST="sd";
+ $MULTU="dmultu";
+ $DIVU="ddivu";
+ $ADDU="daddu";
+ $SUBU="dsubu";
+ $SRL="dsrl";
+ $SLL="dsll";
+ $BNSZ=8;
+ $PTR_ADD="daddu";
+ $PTR_SUB="dsubu";
+ $SZREG=8;
+ $REG_S="sd";
+ $REG_L="ld";
+} else {
+ $LD="lw";
+ $ST="sw";
+ $MULTU="multu";
+ $DIVU="divu";
+ $ADDU="addu";
+ $SUBU="subu";
+ $SRL="srl";
+ $SLL="sll";
+ $BNSZ=4;
+ $PTR_ADD="addu";
+ $PTR_SUB="subu";
+ $SZREG=4;
+ $REG_S="sw";
+ $REG_L="lw";
+ $code="#if !(defined (__mips_isa_rev) && (__mips_isa_rev >= 6))\n.set mips2\n#endif\n";
+}
+
+# Below is N32/64 register layout used in the original module.
+#
+($zero,$at,$v0,$v1)=map("\$$_",(0..3));
+($a0,$a1,$a2,$a3,$a4,$a5,$a6,$a7)=map("\$$_",(4..11));
+($t0,$t1,$t2,$t3,$t8,$t9)=map("\$$_",(12..15,24,25));
+($s0,$s1,$s2,$s3,$s4,$s5,$s6,$s7)=map("\$$_",(16..23));
+($gp,$sp,$fp,$ra)=map("\$$_",(28..31));
+($ta0,$ta1,$ta2,$ta3)=($a4,$a5,$a6,$a7);
+#
+# No special adaptation is required for O32. NUBI on the other hand
+# is treated by saving/restoring ($v1,$t0..$t3).
+
+$gp=$v1 if ($flavour =~ /nubi/i);
+
+$minus4=$v1;
+
+$code.=<<___;
+#include "mips_arch.h"
+
+#if defined(_MIPS_ARCH_MIPS64R6)
+# define ddivu(rs,rt)
+# define mfqt(rd,rs,rt) ddivu rd,rs,rt
+# define mfrm(rd,rs,rt) dmodu rd,rs,rt
+#elif defined(_MIPS_ARCH_MIPS32R6)
+# define divu(rs,rt)
+# define mfqt(rd,rs,rt) divu rd,rs,rt
+# define mfrm(rd,rs,rt) modu rd,rs,rt
+#else
+# define $DIVU(rs,rt) $DIVU $zero,rs,rt
+# define mfqt(rd,rs,rt) mflo rd
+# define mfrm(rd,rs,rt) mfhi rd
+#endif
+
+.rdata
+.asciiz "mips3.s, Version 1.2"
+.asciiz "MIPS II/III/IV ISA artwork by Andy Polyakov <appro\@fy.chalmers.se>"
+
+.text
+.set noat
+
+.align 5
+.globl bn_mul_add_words
+.ent bn_mul_add_words
+bn_mul_add_words:
+ .set noreorder
+ bgtz $a2,bn_mul_add_words_internal
+ move $v0,$zero
+ jr $ra
+ move $a0,$v0
+.end bn_mul_add_words
+
+.align 5
+.ent bn_mul_add_words_internal
+bn_mul_add_words_internal:
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ .frame $sp,6*$SZREG,$ra
+ .mask 0x8000f008,-$SZREG
+ .set noreorder
+ $PTR_SUB $sp,6*$SZREG
+ $REG_S $ra,5*$SZREG($sp)
+ $REG_S $t3,4*$SZREG($sp)
+ $REG_S $t2,3*$SZREG($sp)
+ $REG_S $t1,2*$SZREG($sp)
+ $REG_S $t0,1*$SZREG($sp)
+ $REG_S $gp,0*$SZREG($sp)
+___
+$code.=<<___;
+ .set reorder
+ li $minus4,-4
+ and $ta0,$a2,$minus4
+ beqz $ta0,.L_bn_mul_add_words_tail
+
+.L_bn_mul_add_words_loop:
+ $LD $t0,0($a1)
+ $MULTU ($t0,$a3)
+ $LD $t1,0($a0)
+ $LD $t2,$BNSZ($a1)
+ $LD $t3,$BNSZ($a0)
+ $LD $ta0,2*$BNSZ($a1)
+ $LD $ta1,2*$BNSZ($a0)
+ $ADDU $t1,$v0
+ sltu $v0,$t1,$v0 # All manuals say it "compares 32-bit
+ # values", but it seems to work fine
+ # even on 64-bit registers.
+ mflo ($at,$t0,$a3)
+ mfhi ($t0,$t0,$a3)
+ $ADDU $t1,$at
+ $ADDU $v0,$t0
+ $MULTU ($t2,$a3)
+ sltu $at,$t1,$at
+ $ST $t1,0($a0)
+ $ADDU $v0,$at
+
+ $LD $ta2,3*$BNSZ($a1)
+ $LD $ta3,3*$BNSZ($a0)
+ $ADDU $t3,$v0
+ sltu $v0,$t3,$v0
+ mflo ($at,$t2,$a3)
+ mfhi ($t2,$t2,$a3)
+ $ADDU $t3,$at
+ $ADDU $v0,$t2
+ $MULTU ($ta0,$a3)
+ sltu $at,$t3,$at
+ $ST $t3,$BNSZ($a0)
+ $ADDU $v0,$at
+
+ subu $a2,4
+ $PTR_ADD $a0,4*$BNSZ
+ $PTR_ADD $a1,4*$BNSZ
+ $ADDU $ta1,$v0
+ sltu $v0,$ta1,$v0
+ mflo ($at,$ta0,$a3)
+ mfhi ($ta0,$ta0,$a3)
+ $ADDU $ta1,$at
+ $ADDU $v0,$ta0
+ $MULTU ($ta2,$a3)
+ sltu $at,$ta1,$at
+ $ST $ta1,-2*$BNSZ($a0)
+ $ADDU $v0,$at
+
+
+ and $ta0,$a2,$minus4
+ $ADDU $ta3,$v0
+ sltu $v0,$ta3,$v0
+ mflo ($at,$ta2,$a3)
+ mfhi ($ta2,$ta2,$a3)
+ $ADDU $ta3,$at
+ $ADDU $v0,$ta2
+ sltu $at,$ta3,$at
+ $ST $ta3,-$BNSZ($a0)
+ .set noreorder
+ bgtz $ta0,.L_bn_mul_add_words_loop
+ $ADDU $v0,$at
+
+ beqz $a2,.L_bn_mul_add_words_return
+ nop
+
+.L_bn_mul_add_words_tail:
+ .set reorder
+ $LD $t0,0($a1)
+ $MULTU ($t0,$a3)
+ $LD $t1,0($a0)
+ subu $a2,1
+ $ADDU $t1,$v0
+ sltu $v0,$t1,$v0
+ mflo ($at,$t0,$a3)
+ mfhi ($t0,$t0,$a3)
+ $ADDU $t1,$at
+ $ADDU $v0,$t0
+ sltu $at,$t1,$at
+ $ST $t1,0($a0)
+ $ADDU $v0,$at
+ beqz $a2,.L_bn_mul_add_words_return
+
+ $LD $t0,$BNSZ($a1)
+ $MULTU ($t0,$a3)
+ $LD $t1,$BNSZ($a0)
+ subu $a2,1
+ $ADDU $t1,$v0
+ sltu $v0,$t1,$v0
+ mflo ($at,$t0,$a3)
+ mfhi ($t0,$t0,$a3)
+ $ADDU $t1,$at
+ $ADDU $v0,$t0
+ sltu $at,$t1,$at
+ $ST $t1,$BNSZ($a0)
+ $ADDU $v0,$at
+ beqz $a2,.L_bn_mul_add_words_return
+
+ $LD $t0,2*$BNSZ($a1)
+ $MULTU ($t0,$a3)
+ $LD $t1,2*$BNSZ($a0)
+ $ADDU $t1,$v0
+ sltu $v0,$t1,$v0
+ mflo ($at,$t0,$a3)
+ mfhi ($t0,$t0,$a3)
+ $ADDU $t1,$at
+ $ADDU $v0,$t0
+ sltu $at,$t1,$at
+ $ST $t1,2*$BNSZ($a0)
+ $ADDU $v0,$at
+
+.L_bn_mul_add_words_return:
+ .set noreorder
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ $REG_L $t3,4*$SZREG($sp)
+ $REG_L $t2,3*$SZREG($sp)
+ $REG_L $t1,2*$SZREG($sp)
+ $REG_L $t0,1*$SZREG($sp)
+ $REG_L $gp,0*$SZREG($sp)
+ $PTR_ADD $sp,6*$SZREG
+___
+$code.=<<___;
+ jr $ra
+ move $a0,$v0
+.end bn_mul_add_words_internal
+
+.align 5
+.globl bn_mul_words
+.ent bn_mul_words
+bn_mul_words:
+ .set noreorder
+ bgtz $a2,bn_mul_words_internal
+ move $v0,$zero
+ jr $ra
+ move $a0,$v0
+.end bn_mul_words
+
+.align 5
+.ent bn_mul_words_internal
+bn_mul_words_internal:
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ .frame $sp,6*$SZREG,$ra
+ .mask 0x8000f008,-$SZREG
+ .set noreorder
+ $PTR_SUB $sp,6*$SZREG
+ $REG_S $ra,5*$SZREG($sp)
+ $REG_S $t3,4*$SZREG($sp)
+ $REG_S $t2,3*$SZREG($sp)
+ $REG_S $t1,2*$SZREG($sp)
+ $REG_S $t0,1*$SZREG($sp)
+ $REG_S $gp,0*$SZREG($sp)
+___
+$code.=<<___;
+ .set reorder
+ li $minus4,-4
+ and $ta0,$a2,$minus4
+ beqz $ta0,.L_bn_mul_words_tail
+
+.L_bn_mul_words_loop:
+ $LD $t0,0($a1)
+ $MULTU ($t0,$a3)
+ $LD $t2,$BNSZ($a1)
+ $LD $ta0,2*$BNSZ($a1)
+ $LD $ta2,3*$BNSZ($a1)
+ mflo ($at,$t0,$a3)
+ mfhi ($t0,$t0,$a3)
+ $ADDU $v0,$at
+ sltu $t1,$v0,$at
+ $MULTU ($t2,$a3)
+ $ST $v0,0($a0)
+ $ADDU $v0,$t1,$t0
+
+ subu $a2,4
+ $PTR_ADD $a0,4*$BNSZ
+ $PTR_ADD $a1,4*$BNSZ
+ mflo ($at,$t2,$a3)
+ mfhi ($t2,$t2,$a3)
+ $ADDU $v0,$at
+ sltu $t3,$v0,$at
+ $MULTU ($ta0,$a3)
+ $ST $v0,-3*$BNSZ($a0)
+ $ADDU $v0,$t3,$t2
+
+ mflo ($at,$ta0,$a3)
+ mfhi ($ta0,$ta0,$a3)
+ $ADDU $v0,$at
+ sltu $ta1,$v0,$at
+ $MULTU ($ta2,$a3)
+ $ST $v0,-2*$BNSZ($a0)
+ $ADDU $v0,$ta1,$ta0
+
+ and $ta0,$a2,$minus4
+ mflo ($at,$ta2,$a3)
+ mfhi ($ta2,$ta2,$a3)
+ $ADDU $v0,$at
+ sltu $ta3,$v0,$at
+ $ST $v0,-$BNSZ($a0)
+ .set noreorder
+ bgtz $ta0,.L_bn_mul_words_loop
+ $ADDU $v0,$ta3,$ta2
+
+ beqz $a2,.L_bn_mul_words_return
+ nop
+
+.L_bn_mul_words_tail:
+ .set reorder
+ $LD $t0,0($a1)
+ $MULTU ($t0,$a3)
+ subu $a2,1
+ mflo ($at,$t0,$a3)
+ mfhi ($t0,$t0,$a3)
+ $ADDU $v0,$at
+ sltu $t1,$v0,$at
+ $ST $v0,0($a0)
+ $ADDU $v0,$t1,$t0
+ beqz $a2,.L_bn_mul_words_return
+
+ $LD $t0,$BNSZ($a1)
+ $MULTU ($t0,$a3)
+ subu $a2,1
+ mflo ($at,$t0,$a3)
+ mfhi ($t0,$t0,$a3)
+ $ADDU $v0,$at
+ sltu $t1,$v0,$at
+ $ST $v0,$BNSZ($a0)
+ $ADDU $v0,$t1,$t0
+ beqz $a2,.L_bn_mul_words_return
+
+ $LD $t0,2*$BNSZ($a1)
+ $MULTU ($t0,$a3)
+ mflo ($at,$t0,$a3)
+ mfhi ($t0,$t0,$a3)
+ $ADDU $v0,$at
+ sltu $t1,$v0,$at
+ $ST $v0,2*$BNSZ($a0)
+ $ADDU $v0,$t1,$t0
+
+.L_bn_mul_words_return:
+ .set noreorder
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ $REG_L $t3,4*$SZREG($sp)
+ $REG_L $t2,3*$SZREG($sp)
+ $REG_L $t1,2*$SZREG($sp)
+ $REG_L $t0,1*$SZREG($sp)
+ $REG_L $gp,0*$SZREG($sp)
+ $PTR_ADD $sp,6*$SZREG
+___
+$code.=<<___;
+ jr $ra
+ move $a0,$v0
+.end bn_mul_words_internal
+
+.align 5
+.globl bn_sqr_words
+.ent bn_sqr_words
+bn_sqr_words:
+ .set noreorder
+ bgtz $a2,bn_sqr_words_internal
+ move $v0,$zero
+ jr $ra
+ move $a0,$v0
+.end bn_sqr_words
+
+.align 5
+.ent bn_sqr_words_internal
+bn_sqr_words_internal:
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ .frame $sp,6*$SZREG,$ra
+ .mask 0x8000f008,-$SZREG
+ .set noreorder
+ $PTR_SUB $sp,6*$SZREG
+ $REG_S $ra,5*$SZREG($sp)
+ $REG_S $t3,4*$SZREG($sp)
+ $REG_S $t2,3*$SZREG($sp)
+ $REG_S $t1,2*$SZREG($sp)
+ $REG_S $t0,1*$SZREG($sp)
+ $REG_S $gp,0*$SZREG($sp)
+___
+$code.=<<___;
+ .set reorder
+ li $minus4,-4
+ and $ta0,$a2,$minus4
+ beqz $ta0,.L_bn_sqr_words_tail
+
+.L_bn_sqr_words_loop:
+ $LD $t0,0($a1)
+ $MULTU ($t0,$t0)
+ $LD $t2,$BNSZ($a1)
+ $LD $ta0,2*$BNSZ($a1)
+ $LD $ta2,3*$BNSZ($a1)
+ mflo ($t1,$t0,$t0)
+ mfhi ($t0,$t0,$t0)
+ $ST $t1,0($a0)
+ $ST $t0,$BNSZ($a0)
+
+ $MULTU ($t2,$t2)
+ subu $a2,4
+ $PTR_ADD $a0,8*$BNSZ
+ $PTR_ADD $a1,4*$BNSZ
+ mflo ($t3,$t2,$t2)
+ mfhi ($t2,$t2,$t2)
+ $ST $t3,-6*$BNSZ($a0)
+ $ST $t2,-5*$BNSZ($a0)
+
+ $MULTU ($ta0,$ta0)
+ mflo ($ta1,$ta0,$ta0)
+ mfhi ($ta0,$ta0,$ta0)
+ $ST $ta1,-4*$BNSZ($a0)
+ $ST $ta0,-3*$BNSZ($a0)
+
+
+ $MULTU ($ta2,$ta2)
+ and $ta0,$a2,$minus4
+ mflo ($ta3,$ta2,$ta2)
+ mfhi ($ta2,$ta2,$ta2)
+ $ST $ta3,-2*$BNSZ($a0)
+
+ .set noreorder
+ bgtz $ta0,.L_bn_sqr_words_loop
+ $ST $ta2,-$BNSZ($a0)
+
+ beqz $a2,.L_bn_sqr_words_return
+ nop
+
+.L_bn_sqr_words_tail:
+ .set reorder
+ $LD $t0,0($a1)
+ $MULTU ($t0,$t0)
+ subu $a2,1
+ mflo ($t1,$t0,$t0)
+ mfhi ($t0,$t0,$t0)
+ $ST $t1,0($a0)
+ $ST $t0,$BNSZ($a0)
+ beqz $a2,.L_bn_sqr_words_return
+
+ $LD $t0,$BNSZ($a1)
+ $MULTU ($t0,$t0)
+ subu $a2,1
+ mflo ($t1,$t0,$t0)
+ mfhi ($t0,$t0,$t0)
+ $ST $t1,2*$BNSZ($a0)
+ $ST $t0,3*$BNSZ($a0)
+ beqz $a2,.L_bn_sqr_words_return
+
+ $LD $t0,2*$BNSZ($a1)
+ $MULTU ($t0,$t0)
+ mflo ($t1,$t0,$t0)
+ mfhi ($t0,$t0,$t0)
+ $ST $t1,4*$BNSZ($a0)
+ $ST $t0,5*$BNSZ($a0)
+
+.L_bn_sqr_words_return:
+ .set noreorder
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ $REG_L $t3,4*$SZREG($sp)
+ $REG_L $t2,3*$SZREG($sp)
+ $REG_L $t1,2*$SZREG($sp)
+ $REG_L $t0,1*$SZREG($sp)
+ $REG_L $gp,0*$SZREG($sp)
+ $PTR_ADD $sp,6*$SZREG
+___
+$code.=<<___;
+ jr $ra
+ move $a0,$v0
+
+.end bn_sqr_words_internal
+
+.align 5
+.globl bn_add_words
+.ent bn_add_words
+bn_add_words:
+ .set noreorder
+ bgtz $a3,bn_add_words_internal
+ move $v0,$zero
+ jr $ra
+ move $a0,$v0
+.end bn_add_words
+
+.align 5
+.ent bn_add_words_internal
+bn_add_words_internal:
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ .frame $sp,6*$SZREG,$ra
+ .mask 0x8000f008,-$SZREG
+ .set noreorder
+ $PTR_SUB $sp,6*$SZREG
+ $REG_S $ra,5*$SZREG($sp)
+ $REG_S $t3,4*$SZREG($sp)
+ $REG_S $t2,3*$SZREG($sp)
+ $REG_S $t1,2*$SZREG($sp)
+ $REG_S $t0,1*$SZREG($sp)
+ $REG_S $gp,0*$SZREG($sp)
+___
+$code.=<<___;
+ .set reorder
+ li $minus4,-4
+ and $at,$a3,$minus4
+ beqz $at,.L_bn_add_words_tail
+
+.L_bn_add_words_loop:
+ $LD $t0,0($a1)
+ $LD $ta0,0($a2)
+ subu $a3,4
+ $LD $t1,$BNSZ($a1)
+ and $at,$a3,$minus4
+ $LD $t2,2*$BNSZ($a1)
+ $PTR_ADD $a2,4*$BNSZ
+ $LD $t3,3*$BNSZ($a1)
+ $PTR_ADD $a0,4*$BNSZ
+ $LD $ta1,-3*$BNSZ($a2)
+ $PTR_ADD $a1,4*$BNSZ
+ $LD $ta2,-2*$BNSZ($a2)
+ $LD $ta3,-$BNSZ($a2)
+ $ADDU $ta0,$t0
+ sltu $t8,$ta0,$t0
+ $ADDU $t0,$ta0,$v0
+ sltu $v0,$t0,$ta0
+ $ST $t0,-4*$BNSZ($a0)
+ $ADDU $v0,$t8
+
+ $ADDU $ta1,$t1
+ sltu $t9,$ta1,$t1
+ $ADDU $t1,$ta1,$v0
+ sltu $v0,$t1,$ta1
+ $ST $t1,-3*$BNSZ($a0)
+ $ADDU $v0,$t9
+
+ $ADDU $ta2,$t2
+ sltu $t8,$ta2,$t2
+ $ADDU $t2,$ta2,$v0
+ sltu $v0,$t2,$ta2
+ $ST $t2,-2*$BNSZ($a0)
+ $ADDU $v0,$t8
+
+ $ADDU $ta3,$t3
+ sltu $t9,$ta3,$t3
+ $ADDU $t3,$ta3,$v0
+ sltu $v0,$t3,$ta3
+ $ST $t3,-$BNSZ($a0)
+
+ .set noreorder
+ bgtz $at,.L_bn_add_words_loop
+ $ADDU $v0,$t9
+
+ beqz $a3,.L_bn_add_words_return
+ nop
+
+.L_bn_add_words_tail:
+ .set reorder
+ $LD $t0,0($a1)
+ $LD $ta0,0($a2)
+ $ADDU $ta0,$t0
+ subu $a3,1
+ sltu $t8,$ta0,$t0
+ $ADDU $t0,$ta0,$v0
+ sltu $v0,$t0,$ta0
+ $ST $t0,0($a0)
+ $ADDU $v0,$t8
+ beqz $a3,.L_bn_add_words_return
+
+ $LD $t1,$BNSZ($a1)
+ $LD $ta1,$BNSZ($a2)
+ $ADDU $ta1,$t1
+ subu $a3,1
+ sltu $t9,$ta1,$t1
+ $ADDU $t1,$ta1,$v0
+ sltu $v0,$t1,$ta1
+ $ST $t1,$BNSZ($a0)
+ $ADDU $v0,$t9
+ beqz $a3,.L_bn_add_words_return
+
+ $LD $t2,2*$BNSZ($a1)
+ $LD $ta2,2*$BNSZ($a2)
+ $ADDU $ta2,$t2
+ sltu $t8,$ta2,$t2
+ $ADDU $t2,$ta2,$v0
+ sltu $v0,$t2,$ta2
+ $ST $t2,2*$BNSZ($a0)
+ $ADDU $v0,$t8
+
+.L_bn_add_words_return:
+ .set noreorder
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ $REG_L $t3,4*$SZREG($sp)
+ $REG_L $t2,3*$SZREG($sp)
+ $REG_L $t1,2*$SZREG($sp)
+ $REG_L $t0,1*$SZREG($sp)
+ $REG_L $gp,0*$SZREG($sp)
+ $PTR_ADD $sp,6*$SZREG
+___
+$code.=<<___;
+ jr $ra
+ move $a0,$v0
+
+.end bn_add_words_internal
+
+.align 5
+.globl bn_sub_words
+.ent bn_sub_words
+bn_sub_words:
+ .set noreorder
+ bgtz $a3,bn_sub_words_internal
+ move $v0,$zero
+ jr $ra
+ move $a0,$zero
+.end bn_sub_words
+
+.align 5
+.ent bn_sub_words_internal
+bn_sub_words_internal:
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ .frame $sp,6*$SZREG,$ra
+ .mask 0x8000f008,-$SZREG
+ .set noreorder
+ $PTR_SUB $sp,6*$SZREG
+ $REG_S $ra,5*$SZREG($sp)
+ $REG_S $t3,4*$SZREG($sp)
+ $REG_S $t2,3*$SZREG($sp)
+ $REG_S $t1,2*$SZREG($sp)
+ $REG_S $t0,1*$SZREG($sp)
+ $REG_S $gp,0*$SZREG($sp)
+___
+$code.=<<___;
+ .set reorder
+ li $minus4,-4
+ and $at,$a3,$minus4
+ beqz $at,.L_bn_sub_words_tail
+
+.L_bn_sub_words_loop:
+ $LD $t0,0($a1)
+ $LD $ta0,0($a2)
+ subu $a3,4
+ $LD $t1,$BNSZ($a1)
+ and $at,$a3,$minus4
+ $LD $t2,2*$BNSZ($a1)
+ $PTR_ADD $a2,4*$BNSZ
+ $LD $t3,3*$BNSZ($a1)
+ $PTR_ADD $a0,4*$BNSZ
+ $LD $ta1,-3*$BNSZ($a2)
+ $PTR_ADD $a1,4*$BNSZ
+ $LD $ta2,-2*$BNSZ($a2)
+ $LD $ta3,-$BNSZ($a2)
+ sltu $t8,$t0,$ta0
+ $SUBU $ta0,$t0,$ta0
+ $SUBU $t0,$ta0,$v0
+ sgtu $v0,$t0,$ta0
+ $ST $t0,-4*$BNSZ($a0)
+ $ADDU $v0,$t8
+
+ sltu $t9,$t1,$ta1
+ $SUBU $ta1,$t1,$ta1
+ $SUBU $t1,$ta1,$v0
+ sgtu $v0,$t1,$ta1
+ $ST $t1,-3*$BNSZ($a0)
+ $ADDU $v0,$t9
+
+
+ sltu $t8,$t2,$ta2
+ $SUBU $ta2,$t2,$ta2
+ $SUBU $t2,$ta2,$v0
+ sgtu $v0,$t2,$ta2
+ $ST $t2,-2*$BNSZ($a0)
+ $ADDU $v0,$t8
+
+ sltu $t9,$t3,$ta3
+ $SUBU $ta3,$t3,$ta3
+ $SUBU $t3,$ta3,$v0
+ sgtu $v0,$t3,$ta3
+ $ST $t3,-$BNSZ($a0)
+
+ .set noreorder
+ bgtz $at,.L_bn_sub_words_loop
+ $ADDU $v0,$t9
+
+ beqz $a3,.L_bn_sub_words_return
+ nop
+
+.L_bn_sub_words_tail:
+ .set reorder
+ $LD $t0,0($a1)
+ $LD $ta0,0($a2)
+ subu $a3,1
+ sltu $t8,$t0,$ta0
+ $SUBU $ta0,$t0,$ta0
+ $SUBU $t0,$ta0,$v0
+ sgtu $v0,$t0,$ta0
+ $ST $t0,0($a0)
+ $ADDU $v0,$t8
+ beqz $a3,.L_bn_sub_words_return
+
+ $LD $t1,$BNSZ($a1)
+ subu $a3,1
+ $LD $ta1,$BNSZ($a2)
+ sltu $t9,$t1,$ta1
+ $SUBU $ta1,$t1,$ta1
+ $SUBU $t1,$ta1,$v0
+ sgtu $v0,$t1,$ta1
+ $ST $t1,$BNSZ($a0)
+ $ADDU $v0,$t9
+ beqz $a3,.L_bn_sub_words_return
+
+ $LD $t2,2*$BNSZ($a1)
+ $LD $ta2,2*$BNSZ($a2)
+ sltu $t8,$t2,$ta2
+ $SUBU $ta2,$t2,$ta2
+ $SUBU $t2,$ta2,$v0
+ sgtu $v0,$t2,$ta2
+ $ST $t2,2*$BNSZ($a0)
+ $ADDU $v0,$t8
+
+.L_bn_sub_words_return:
+ .set noreorder
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ $REG_L $t3,4*$SZREG($sp)
+ $REG_L $t2,3*$SZREG($sp)
+ $REG_L $t1,2*$SZREG($sp)
+ $REG_L $t0,1*$SZREG($sp)
+ $REG_L $gp,0*$SZREG($sp)
+ $PTR_ADD $sp,6*$SZREG
+___
+$code.=<<___;
+ jr $ra
+ move $a0,$v0
+.end bn_sub_words_internal
+
+#if 0
+/*
+ * The bn_div_3_words entry point is re-used for constant-time interface.
+ * Implementation is retained as historical reference.
+ */
+.align 5
+.globl bn_div_3_words
+.ent bn_div_3_words
+bn_div_3_words:
+ .set noreorder
+ move $a3,$a0 # we know that bn_div_words does not
+ # touch $a3, $ta2, $ta3 and preserves $a2
+ # so that we can save two arguments
+ # and return address in registers
+ # instead of stack:-)
+
+ $LD $a0,($a3)
+ move $ta2,$a1
+ bne $a0,$a2,bn_div_3_words_internal
+ $LD $a1,-$BNSZ($a3)
+ li $v0,-1
+ jr $ra
+ move $a0,$v0
+.end bn_div_3_words
+
+.align 5
+.ent bn_div_3_words_internal
+bn_div_3_words_internal:
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ .frame $sp,6*$SZREG,$ra
+ .mask 0x8000f008,-$SZREG
+ .set noreorder
+ $PTR_SUB $sp,6*$SZREG
+ $REG_S $ra,5*$SZREG($sp)
+ $REG_S $t3,4*$SZREG($sp)
+ $REG_S $t2,3*$SZREG($sp)
+ $REG_S $t1,2*$SZREG($sp)
+ $REG_S $t0,1*$SZREG($sp)
+ $REG_S $gp,0*$SZREG($sp)
+___
+$code.=<<___;
+ .set reorder
+ move $ta3,$ra
+ bal bn_div_words_internal
+ move $ra,$ta3
+ $MULTU ($ta2,$v0)
+ $LD $t2,-2*$BNSZ($a3)
+ move $ta0,$zero
+ mfhi ($t1,$ta2,$v0)
+ mflo ($t0,$ta2,$v0)
+ sltu $t8,$t1,$a1
+.L_bn_div_3_words_inner_loop:
+ bnez $t8,.L_bn_div_3_words_inner_loop_done
+ sgeu $at,$t2,$t0
+ seq $t9,$t1,$a1
+ and $at,$t9
+ sltu $t3,$t0,$ta2
+ $ADDU $a1,$a2
+ $SUBU $t1,$t3
+ $SUBU $t0,$ta2
+ sltu $t8,$t1,$a1
+ sltu $ta0,$a1,$a2
+ or $t8,$ta0
+ .set noreorder
+ beqz $at,.L_bn_div_3_words_inner_loop
+ $SUBU $v0,1
+ $ADDU $v0,1
+ .set reorder
+.L_bn_div_3_words_inner_loop_done:
+ .set noreorder
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ $REG_L $t3,4*$SZREG($sp)
+ $REG_L $t2,3*$SZREG($sp)
+ $REG_L $t1,2*$SZREG($sp)
+ $REG_L $t0,1*$SZREG($sp)
+ $REG_L $gp,0*$SZREG($sp)
+ $PTR_ADD $sp,6*$SZREG
+___
+$code.=<<___;
+ jr $ra
+ move $a0,$v0
+.end bn_div_3_words_internal
+#endif
+
+.align 5
+.globl bn_div_words
+.ent bn_div_words
+bn_div_words:
+ .set noreorder
+ bnez $a2,bn_div_words_internal
+ li $v0,-1 # I would rather signal div-by-zero
+ # which can be done with 'break 7'
+ jr $ra
+ move $a0,$v0
+.end bn_div_words
+
+.align 5
+.ent bn_div_words_internal
+bn_div_words_internal:
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ .frame $sp,6*$SZREG,$ra
+ .mask 0x8000f008,-$SZREG
+ .set noreorder
+ $PTR_SUB $sp,6*$SZREG
+ $REG_S $ra,5*$SZREG($sp)
+ $REG_S $t3,4*$SZREG($sp)
+ $REG_S $t2,3*$SZREG($sp)
+ $REG_S $t1,2*$SZREG($sp)
+ $REG_S $t0,1*$SZREG($sp)
+ $REG_S $gp,0*$SZREG($sp)
+___
+$code.=<<___;
+ move $v1,$zero
+ bltz $a2,.L_bn_div_words_body
+ move $t9,$v1
+ $SLL $a2,1
+ bgtz $a2,.-4
+ addu $t9,1
+
+ .set reorder
+ negu $t1,$t9
+ li $t2,-1
+ $SLL $t2,$t1
+ and $t2,$a0
+ $SRL $at,$a1,$t1
+ .set noreorder
+ beqz $t2,.+12
+ nop
+ break 6 # signal overflow
+ .set reorder
+ $SLL $a0,$t9
+ $SLL $a1,$t9
+ or $a0,$at
+___
+$QT=$ta0;
+$HH=$ta1;
+$DH=$v1;
+$code.=<<___;
+.L_bn_div_words_body:
+ $SRL $DH,$a2,4*$BNSZ # bits
+ sgeu $at,$a0,$a2
+ .set noreorder
+ beqz $at,.+12
+ nop
+ $SUBU $a0,$a2
+ .set reorder
+
+ li $QT,-1
+ $SRL $HH,$a0,4*$BNSZ # bits
+ $SRL $QT,4*$BNSZ # q=0xffffffff
+ beq $DH,$HH,.L_bn_div_words_skip_div1
+ $DIVU ($a0,$DH)
+ mfqt ($QT,$a0,$DH)
+.L_bn_div_words_skip_div1:
+ $MULTU ($a2,$QT)
+ $SLL $t3,$a0,4*$BNSZ # bits
+ $SRL $at,$a1,4*$BNSZ # bits
+ or $t3,$at
+ mflo ($t0,$a2,$QT)
+ mfhi ($t1,$a2,$QT)
+.L_bn_div_words_inner_loop1:
+ sltu $t2,$t3,$t0
+ seq $t8,$HH,$t1
+ sltu $at,$HH,$t1
+ and $t2,$t8
+ sltu $v0,$t0,$a2
+ or $at,$t2
+ .set noreorder
+ beqz $at,.L_bn_div_words_inner_loop1_done
+ $SUBU $t1,$v0
+ $SUBU $t0,$a2
+ b .L_bn_div_words_inner_loop1
+ $SUBU $QT,1
+ .set reorder
+.L_bn_div_words_inner_loop1_done:
+
+ $SLL $a1,4*$BNSZ # bits
+ $SUBU $a0,$t3,$t0
+ $SLL $v0,$QT,4*$BNSZ # bits
+
+ li $QT,-1
+ $SRL $HH,$a0,4*$BNSZ # bits
+ $SRL $QT,4*$BNSZ # q=0xffffffff
+ beq $DH,$HH,.L_bn_div_words_skip_div2
+ $DIVU ($a0,$DH)
+ mfqt ($QT,$a0,$DH)
+.L_bn_div_words_skip_div2:
+ $MULTU ($a2,$QT)
+ $SLL $t3,$a0,4*$BNSZ # bits
+ $SRL $at,$a1,4*$BNSZ # bits
+ or $t3,$at
+ mflo ($t0,$a2,$QT)
+ mfhi ($t1,$a2,$QT)
+.L_bn_div_words_inner_loop2:
+ sltu $t2,$t3,$t0
+ seq $t8,$HH,$t1
+ sltu $at,$HH,$t1
+ and $t2,$t8
+ sltu $v1,$t0,$a2
+ or $at,$t2
+ .set noreorder
+ beqz $at,.L_bn_div_words_inner_loop2_done
+ $SUBU $t1,$v1
+ $SUBU $t0,$a2
+ b .L_bn_div_words_inner_loop2
+ $SUBU $QT,1
+ .set reorder
+.L_bn_div_words_inner_loop2_done:
+
+ $SUBU $a0,$t3,$t0
+ or $v0,$QT
+ $SRL $v1,$a0,$t9 # $v1 contains remainder if anybody wants it
+ $SRL $a2,$t9 # restore $a2
+
+ .set noreorder
+ move $a1,$v1
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ $REG_L $t3,4*$SZREG($sp)
+ $REG_L $t2,3*$SZREG($sp)
+ $REG_L $t1,2*$SZREG($sp)
+ $REG_L $t0,1*$SZREG($sp)
+ $REG_L $gp,0*$SZREG($sp)
+ $PTR_ADD $sp,6*$SZREG
+___
+$code.=<<___;
+ jr $ra
+ move $a0,$v0
+.end bn_div_words_internal
+___
+undef $HH; undef $QT; undef $DH;
+
+($a_0,$a_1,$a_2,$a_3)=($t0,$t1,$t2,$t3);
+($b_0,$b_1,$b_2,$b_3)=($ta0,$ta1,$ta2,$ta3);
+
+($a_4,$a_5,$a_6,$a_7)=($s0,$s2,$s4,$a1); # once we load a[7], no use for $a1
+($b_4,$b_5,$b_6,$b_7)=($s1,$s3,$s5,$a2); # once we load b[7], no use for $a2
+
+($t_1,$t_2,$c_1,$c_2,$c_3)=($t8,$t9,$v0,$v1,$a3);
+
+$code.=<<___;
+
+.align 5
+.globl bn_mul_comba8
+.ent bn_mul_comba8
+bn_mul_comba8:
+ .set noreorder
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ .frame $sp,12*$SZREG,$ra
+ .mask 0x803ff008,-$SZREG
+ $PTR_SUB $sp,12*$SZREG
+ $REG_S $ra,11*$SZREG($sp)
+ $REG_S $s5,10*$SZREG($sp)
+ $REG_S $s4,9*$SZREG($sp)
+ $REG_S $s3,8*$SZREG($sp)
+ $REG_S $s2,7*$SZREG($sp)
+ $REG_S $s1,6*$SZREG($sp)
+ $REG_S $s0,5*$SZREG($sp)
+ $REG_S $t3,4*$SZREG($sp)
+ $REG_S $t2,3*$SZREG($sp)
+ $REG_S $t1,2*$SZREG($sp)
+ $REG_S $t0,1*$SZREG($sp)
+ $REG_S $gp,0*$SZREG($sp)
+___
+$code.=<<___ if ($flavour !~ /nubi/i);
+ .frame $sp,6*$SZREG,$ra
+ .mask 0x003f0000,-$SZREG
+ $PTR_SUB $sp,6*$SZREG
+ $REG_S $s5,5*$SZREG($sp)
+ $REG_S $s4,4*$SZREG($sp)
+ $REG_S $s3,3*$SZREG($sp)
+ $REG_S $s2,2*$SZREG($sp)
+ $REG_S $s1,1*$SZREG($sp)
+ $REG_S $s0,0*$SZREG($sp)
+___
+$code.=<<___;
+
+ .set reorder
+ $LD $a_0,0($a1) # If compiled with -mips3 option on
+ # R5000 box assembler barks on this
+ # 1ine with "should not have mult/div
+ # as last instruction in bb (R10K
+ # bug)" warning. If anybody out there
+ # has a clue about how to circumvent
+ # this do send me a note.
+ # <appro\@fy.chalmers.se>
+
+ $LD $b_0,0($a2)
+ $LD $a_1,$BNSZ($a1)
+ $LD $a_2,2*$BNSZ($a1)
+ $MULTU ($a_0,$b_0) # mul_add_c(a[0],b[0],c1,c2,c3);
+ $LD $a_3,3*$BNSZ($a1)
+ $LD $b_1,$BNSZ($a2)
+ $LD $b_2,2*$BNSZ($a2)
+ $LD $b_3,3*$BNSZ($a2)
+ mflo ($c_1,$a_0,$b_0)
+ mfhi ($c_2,$a_0,$b_0)
+
+ $LD $a_4,4*$BNSZ($a1)
+ $LD $a_5,5*$BNSZ($a1)
+ $MULTU ($a_0,$b_1) # mul_add_c(a[0],b[1],c2,c3,c1);
+ $LD $a_6,6*$BNSZ($a1)
+ $LD $a_7,7*$BNSZ($a1)
+ $LD $b_4,4*$BNSZ($a2)
+ $LD $b_5,5*$BNSZ($a2)
+ mflo ($t_1,$a_0,$b_1)
+ mfhi ($t_2,$a_0,$b_1)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_1,$b_0) # mul_add_c(a[1],b[0],c2,c3,c1);
+ $ADDU $c_3,$t_2,$at
+ $LD $b_6,6*$BNSZ($a2)
+ $LD $b_7,7*$BNSZ($a2)
+ $ST $c_1,0($a0) # r[0]=c1;
+ mflo ($t_1,$a_1,$b_0)
+ mfhi ($t_2,$a_1,$b_0)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_2,$b_0) # mul_add_c(a[2],b[0],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $c_1,$c_3,$t_2
+ $ST $c_2,$BNSZ($a0) # r[1]=c2;
+
+ mflo ($t_1,$a_2,$b_0)
+ mfhi ($t_2,$a_2,$b_0)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_1,$b_1) # mul_add_c(a[1],b[1],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ mflo ($t_1,$a_1,$b_1)
+ mfhi ($t_2,$a_1,$b_1)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_0,$b_2) # mul_add_c(a[0],b[2],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $c_2,$c_1,$t_2
+ mflo ($t_1,$a_0,$b_2)
+ mfhi ($t_2,$a_0,$b_2)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_0,$b_3) # mul_add_c(a[0],b[3],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $at,$c_1,$t_2
+ $ADDU $c_2,$at
+ $ST $c_3,2*$BNSZ($a0) # r[2]=c3;
+
+ mflo ($t_1,$a_0,$b_3)
+ mfhi ($t_2,$a_0,$b_3)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_1,$b_2) # mul_add_c(a[1],b[2],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $c_3,$c_2,$t_2
+ mflo ($t_1,$a_1,$b_2)
+ mfhi ($t_2,$a_1,$b_2)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_2,$b_1) # mul_add_c(a[2],b[1],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $at,$c_2,$t_2
+ $ADDU $c_3,$at
+ mflo ($t_1,$a_2,$b_1)
+ mfhi ($t_2,$a_2,$b_1)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_3,$b_0) # mul_add_c(a[3],b[0],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $at,$c_2,$t_2
+ $ADDU $c_3,$at
+ mflo ($t_1,$a_3,$b_0)
+ mfhi ($t_2,$a_3,$b_0)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_4,$b_0) # mul_add_c(a[4],b[0],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $at,$c_2,$t_2
+ $ADDU $c_3,$at
+ $ST $c_1,3*$BNSZ($a0) # r[3]=c1;
+
+ mflo ($t_1,$a_4,$b_0)
+ mfhi ($t_2,$a_4,$b_0)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_3,$b_1) # mul_add_c(a[3],b[1],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $c_1,$c_3,$t_2
+ mflo ($t_1,$a_3,$b_1)
+ mfhi ($t_2,$a_3,$b_1)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_2,$b_2) # mul_add_c(a[2],b[2],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $at,$c_3,$t_2
+ $ADDU $c_1,$at
+ mflo ($t_1,$a_2,$b_2)
+ mfhi ($t_2,$a_2,$b_2)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_1,$b_3) # mul_add_c(a[1],b[3],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $at,$c_3,$t_2
+ $ADDU $c_1,$at
+ mflo ($t_1,$a_1,$b_3)
+ mfhi ($t_2,$a_1,$b_3)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_0,$b_4) # mul_add_c(a[0],b[4],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $at,$c_3,$t_2
+ $ADDU $c_1,$at
+ mflo ($t_1,$a_0,$b_4)
+ mfhi ($t_2,$a_0,$b_4)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_0,$b_5) # mul_add_c(a[0],b[5],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $at,$c_3,$t_2
+ $ADDU $c_1,$at
+ $ST $c_2,4*$BNSZ($a0) # r[4]=c2;
+
+ mflo ($t_1,$a_0,$b_5)
+ mfhi ($t_2,$a_0,$b_5)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_1,$b_4) # mul_add_c(a[1],b[4],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $c_2,$c_1,$t_2
+ mflo ($t_1,$a_1,$b_4)
+ mfhi ($t_2,$a_1,$b_4)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_2,$b_3) # mul_add_c(a[2],b[3],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $at,$c_1,$t_2
+ $ADDU $c_2,$at
+ mflo ($t_1,$a_2,$b_3)
+ mfhi ($t_2,$a_2,$b_3)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_3,$b_2) # mul_add_c(a[3],b[2],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $at,$c_1,$t_2
+ $ADDU $c_2,$at
+ mflo ($t_1,$a_3,$b_2)
+ mfhi ($t_2,$a_3,$b_2)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_4,$b_1) # mul_add_c(a[4],b[1],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $at,$c_1,$t_2
+ $ADDU $c_2,$at
+ mflo ($t_1,$a_4,$b_1)
+ mfhi ($t_2,$a_4,$b_1)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_5,$b_0) # mul_add_c(a[5],b[0],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $at,$c_1,$t_2
+ $ADDU $c_2,$at
+ mflo ($t_1,$a_5,$b_0)
+ mfhi ($t_2,$a_5,$b_0)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_6,$b_0) # mul_add_c(a[6],b[0],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $at,$c_1,$t_2
+ $ADDU $c_2,$at
+ $ST $c_3,5*$BNSZ($a0) # r[5]=c3;
+
+ mflo ($t_1,$a_6,$b_0)
+ mfhi ($t_2,$a_6,$b_0)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_5,$b_1) # mul_add_c(a[5],b[1],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $c_3,$c_2,$t_2
+ mflo ($t_1,$a_5,$b_1)
+ mfhi ($t_2,$a_5,$b_1)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_4,$b_2) # mul_add_c(a[4],b[2],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $at,$c_2,$t_2
+ $ADDU $c_3,$at
+ mflo ($t_1,$a_4,$b_2)
+ mfhi ($t_2,$a_4,$b_2)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_3,$b_3) # mul_add_c(a[3],b[3],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $at,$c_2,$t_2
+ $ADDU $c_3,$at
+ mflo ($t_1,$a_3,$b_3)
+ mfhi ($t_2,$a_3,$b_3)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_2,$b_4) # mul_add_c(a[2],b[4],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $at,$c_2,$t_2
+ $ADDU $c_3,$at
+ mflo ($t_1,$a_2,$b_4)
+ mfhi ($t_2,$a_2,$b_4)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_1,$b_5) # mul_add_c(a[1],b[5],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $at,$c_2,$t_2
+ $ADDU $c_3,$at
+ mflo ($t_1,$a_1,$b_5)
+ mfhi ($t_2,$a_1,$b_5)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_0,$b_6) # mul_add_c(a[0],b[6],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $at,$c_2,$t_2
+ $ADDU $c_3,$at
+ mflo ($t_1,$a_0,$b_6)
+ mfhi ($t_2,$a_0,$b_6)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_0,$b_7) # mul_add_c(a[0],b[7],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $at,$c_2,$t_2
+ $ADDU $c_3,$at
+ $ST $c_1,6*$BNSZ($a0) # r[6]=c1;
+
+ mflo ($t_1,$a_0,$b_7)
+ mfhi ($t_2,$a_0,$b_7)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_1,$b_6) # mul_add_c(a[1],b[6],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $c_1,$c_3,$t_2
+ mflo ($t_1,$a_1,$b_6)
+ mfhi ($t_2,$a_1,$b_6)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_2,$b_5) # mul_add_c(a[2],b[5],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $at,$c_3,$t_2
+ $ADDU $c_1,$at
+ mflo ($t_1,$a_2,$b_5)
+ mfhi ($t_2,$a_2,$b_5)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_3,$b_4) # mul_add_c(a[3],b[4],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $at,$c_3,$t_2
+ $ADDU $c_1,$at
+ mflo ($t_1,$a_3,$b_4)
+ mfhi ($t_2,$a_3,$b_4)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_4,$b_3) # mul_add_c(a[4],b[3],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $at,$c_3,$t_2
+ $ADDU $c_1,$at
+ mflo ($t_1,$a_4,$b_3)
+ mfhi ($t_2,$a_4,$b_3)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_5,$b_2) # mul_add_c(a[5],b[2],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $at,$c_3,$t_2
+ $ADDU $c_1,$at
+ mflo ($t_1,$a_5,$b_2)
+ mfhi ($t_2,$a_5,$b_2)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_6,$b_1) # mul_add_c(a[6],b[1],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $at,$c_3,$t_2
+ $ADDU $c_1,$at
+ mflo ($t_1,$a_6,$b_1)
+ mfhi ($t_2,$a_6,$b_1)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_7,$b_0) # mul_add_c(a[7],b[0],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $at,$c_3,$t_2
+ $ADDU $c_1,$at
+ mflo ($t_1,$a_7,$b_0)
+ mfhi ($t_2,$a_7,$b_0)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_7,$b_1) # mul_add_c(a[7],b[1],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $at,$c_3,$t_2
+ $ADDU $c_1,$at
+ $ST $c_2,7*$BNSZ($a0) # r[7]=c2;
+
+ mflo ($t_1,$a_7,$b_1)
+ mfhi ($t_2,$a_7,$b_1)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_6,$b_2) # mul_add_c(a[6],b[2],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $c_2,$c_1,$t_2
+ mflo ($t_1,$a_6,$b_2)
+ mfhi ($t_2,$a_6,$b_2)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_5,$b_3) # mul_add_c(a[5],b[3],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $at,$c_1,$t_2
+ $ADDU $c_2,$at
+ mflo ($t_1,$a_5,$b_3)
+ mfhi ($t_2,$a_5,$b_3)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_4,$b_4) # mul_add_c(a[4],b[4],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $at,$c_1,$t_2
+ $ADDU $c_2,$at
+ mflo ($t_1,$a_4,$b_4)
+ mfhi ($t_2,$a_4,$b_4)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_3,$b_5) # mul_add_c(a[3],b[5],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $at,$c_1,$t_2
+ $ADDU $c_2,$at
+ mflo ($t_1,$a_3,$b_5)
+ mfhi ($t_2,$a_3,$b_5)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_2,$b_6) # mul_add_c(a[2],b[6],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $at,$c_1,$t_2
+ $ADDU $c_2,$at
+ mflo ($t_1,$a_2,$b_6)
+ mfhi ($t_2,$a_2,$b_6)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_1,$b_7) # mul_add_c(a[1],b[7],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $at,$c_1,$t_2
+ $ADDU $c_2,$at
+ mflo ($t_1,$a_1,$b_7)
+ mfhi ($t_2,$a_1,$b_7)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_2,$b_7) # mul_add_c(a[2],b[7],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $at,$c_1,$t_2
+ $ADDU $c_2,$at
+ $ST $c_3,8*$BNSZ($a0) # r[8]=c3;
+
+ mflo ($t_1,$a_2,$b_7)
+ mfhi ($t_2,$a_2,$b_7)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_3,$b_6) # mul_add_c(a[3],b[6],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $c_3,$c_2,$t_2
+ mflo ($t_1,$a_3,$b_6)
+ mfhi ($t_2,$a_3,$b_6)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_4,$b_5) # mul_add_c(a[4],b[5],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $at,$c_2,$t_2
+ $ADDU $c_3,$at
+ mflo ($t_1,$a_4,$b_5)
+ mfhi ($t_2,$a_4,$b_5)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_5,$b_4) # mul_add_c(a[5],b[4],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $at,$c_2,$t_2
+ $ADDU $c_3,$at
+ mflo ($t_1,$a_5,$b_4)
+ mfhi ($t_2,$a_5,$b_4)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_6,$b_3) # mul_add_c(a[6],b[3],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $at,$c_2,$t_2
+ $ADDU $c_3,$at
+ mflo ($t_1,$a_6,$b_3)
+ mfhi ($t_2,$a_6,$b_3)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_7,$b_2) # mul_add_c(a[7],b[2],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $at,$c_2,$t_2
+ $ADDU $c_3,$at
+ mflo ($t_1,$a_7,$b_2)
+ mfhi ($t_2,$a_7,$b_2)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_7,$b_3) # mul_add_c(a[7],b[3],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $at,$c_2,$t_2
+ $ADDU $c_3,$at
+ $ST $c_1,9*$BNSZ($a0) # r[9]=c1;
+
+ mflo ($t_1,$a_7,$b_3)
+ mfhi ($t_2,$a_7,$b_3)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_6,$b_4) # mul_add_c(a[6],b[4],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $c_1,$c_3,$t_2
+ mflo ($t_1,$a_6,$b_4)
+ mfhi ($t_2,$a_6,$b_4)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_5,$b_5) # mul_add_c(a[5],b[5],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $at,$c_3,$t_2
+ $ADDU $c_1,$at
+ mflo ($t_1,$a_5,$b_5)
+ mfhi ($t_2,$a_5,$b_5)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_4,$b_6) # mul_add_c(a[4],b[6],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $at,$c_3,$t_2
+ $ADDU $c_1,$at
+ mflo ($t_1,$a_4,$b_6)
+ mfhi ($t_2,$a_4,$b_6)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_3,$b_7) # mul_add_c(a[3],b[7],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $at,$c_3,$t_2
+ $ADDU $c_1,$at
+ mflo ($t_1,$a_3,$b_7)
+ mfhi ($t_2,$a_3,$b_7)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_4,$b_7) # mul_add_c(a[4],b[7],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $at,$c_3,$t_2
+ $ADDU $c_1,$at
+ $ST $c_2,10*$BNSZ($a0) # r[10]=c2;
+
+ mflo ($t_1,$a_4,$b_7)
+ mfhi ($t_2,$a_4,$b_7)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_5,$b_6) # mul_add_c(a[5],b[6],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $c_2,$c_1,$t_2
+ mflo ($t_1,$a_5,$b_6)
+ mfhi ($t_2,$a_5,$b_6)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_6,$b_5) # mul_add_c(a[6],b[5],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $at,$c_1,$t_2
+ $ADDU $c_2,$at
+ mflo ($t_1,$a_6,$b_5)
+ mfhi ($t_2,$a_6,$b_5)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_7,$b_4) # mul_add_c(a[7],b[4],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $at,$c_1,$t_2
+ $ADDU $c_2,$at
+ mflo ($t_1,$a_7,$b_4)
+ mfhi ($t_2,$a_7,$b_4)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_7,$b_5) # mul_add_c(a[7],b[5],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $at,$c_1,$t_2
+ $ADDU $c_2,$at
+ $ST $c_3,11*$BNSZ($a0) # r[11]=c3;
+
+ mflo ($t_1,$a_7,$b_5)
+ mfhi ($t_2,$a_7,$b_5)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_6,$b_6) # mul_add_c(a[6],b[6],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $c_3,$c_2,$t_2
+ mflo ($t_1,$a_6,$b_6)
+ mfhi ($t_2,$a_6,$b_6)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_5,$b_7) # mul_add_c(a[5],b[7],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $at,$c_2,$t_2
+ $ADDU $c_3,$at
+ mflo ($t_1,$a_5,$b_7)
+ mfhi ($t_2,$a_5,$b_7)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_6,$b_7) # mul_add_c(a[6],b[7],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $at,$c_2,$t_2
+ $ADDU $c_3,$at
+ $ST $c_1,12*$BNSZ($a0) # r[12]=c1;
+
+ mflo ($t_1,$a_6,$b_7)
+ mfhi ($t_2,$a_6,$b_7)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_7,$b_6) # mul_add_c(a[7],b[6],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $c_1,$c_3,$t_2
+ mflo ($t_1,$a_7,$b_6)
+ mfhi ($t_2,$a_7,$b_6)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_7,$b_7) # mul_add_c(a[7],b[7],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $at,$c_3,$t_2
+ $ADDU $c_1,$at
+ $ST $c_2,13*$BNSZ($a0) # r[13]=c2;
+
+ mflo ($t_1,$a_7,$b_7)
+ mfhi ($t_2,$a_7,$b_7)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ $ST $c_3,14*$BNSZ($a0) # r[14]=c3;
+ $ST $c_1,15*$BNSZ($a0) # r[15]=c1;
+
+ .set noreorder
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ $REG_L $s5,10*$SZREG($sp)
+ $REG_L $s4,9*$SZREG($sp)
+ $REG_L $s3,8*$SZREG($sp)
+ $REG_L $s2,7*$SZREG($sp)
+ $REG_L $s1,6*$SZREG($sp)
+ $REG_L $s0,5*$SZREG($sp)
+ $REG_L $t3,4*$SZREG($sp)
+ $REG_L $t2,3*$SZREG($sp)
+ $REG_L $t1,2*$SZREG($sp)
+ $REG_L $t0,1*$SZREG($sp)
+ $REG_L $gp,0*$SZREG($sp)
+ jr $ra
+ $PTR_ADD $sp,12*$SZREG
+___
+$code.=<<___ if ($flavour !~ /nubi/i);
+ $REG_L $s5,5*$SZREG($sp)
+ $REG_L $s4,4*$SZREG($sp)
+ $REG_L $s3,3*$SZREG($sp)
+ $REG_L $s2,2*$SZREG($sp)
+ $REG_L $s1,1*$SZREG($sp)
+ $REG_L $s0,0*$SZREG($sp)
+ jr $ra
+ $PTR_ADD $sp,6*$SZREG
+___
+$code.=<<___;
+.end bn_mul_comba8
+
+.align 5
+.globl bn_mul_comba4
+.ent bn_mul_comba4
+bn_mul_comba4:
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ .frame $sp,6*$SZREG,$ra
+ .mask 0x8000f008,-$SZREG
+ .set noreorder
+ $PTR_SUB $sp,6*$SZREG
+ $REG_S $ra,5*$SZREG($sp)
+ $REG_S $t3,4*$SZREG($sp)
+ $REG_S $t2,3*$SZREG($sp)
+ $REG_S $t1,2*$SZREG($sp)
+ $REG_S $t0,1*$SZREG($sp)
+ $REG_S $gp,0*$SZREG($sp)
+___
+$code.=<<___;
+ .set reorder
+ $LD $a_0,0($a1)
+ $LD $b_0,0($a2)
+ $LD $a_1,$BNSZ($a1)
+ $LD $a_2,2*$BNSZ($a1)
+ $MULTU ($a_0,$b_0) # mul_add_c(a[0],b[0],c1,c2,c3);
+ $LD $a_3,3*$BNSZ($a1)
+ $LD $b_1,$BNSZ($a2)
+ $LD $b_2,2*$BNSZ($a2)
+ $LD $b_3,3*$BNSZ($a2)
+ mflo ($c_1,$a_0,$b_0)
+ mfhi ($c_2,$a_0,$b_0)
+ $ST $c_1,0($a0)
+
+ $MULTU ($a_0,$b_1) # mul_add_c(a[0],b[1],c2,c3,c1);
+ mflo ($t_1,$a_0,$b_1)
+ mfhi ($t_2,$a_0,$b_1)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_1,$b_0) # mul_add_c(a[1],b[0],c2,c3,c1);
+ $ADDU $c_3,$t_2,$at
+ mflo ($t_1,$a_1,$b_0)
+ mfhi ($t_2,$a_1,$b_0)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_2,$b_0) # mul_add_c(a[2],b[0],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $c_1,$c_3,$t_2
+ $ST $c_2,$BNSZ($a0)
+
+ mflo ($t_1,$a_2,$b_0)
+ mfhi ($t_2,$a_2,$b_0)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_1,$b_1) # mul_add_c(a[1],b[1],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ mflo ($t_1,$a_1,$b_1)
+ mfhi ($t_2,$a_1,$b_1)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_0,$b_2) # mul_add_c(a[0],b[2],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $c_2,$c_1,$t_2
+ mflo ($t_1,$a_0,$b_2)
+ mfhi ($t_2,$a_0,$b_2)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_0,$b_3) # mul_add_c(a[0],b[3],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $at,$c_1,$t_2
+ $ADDU $c_2,$at
+ $ST $c_3,2*$BNSZ($a0)
+
+ mflo ($t_1,$a_0,$b_3)
+ mfhi ($t_2,$a_0,$b_3)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_1,$b_2) # mul_add_c(a[1],b[2],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $c_3,$c_2,$t_2
+ mflo ($t_1,$a_1,$b_2)
+ mfhi ($t_2,$a_1,$b_2)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_2,$b_1) # mul_add_c(a[2],b[1],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $at,$c_2,$t_2
+ $ADDU $c_3,$at
+ mflo ($t_1,$a_2,$b_1)
+ mfhi ($t_2,$a_2,$b_1)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_3,$b_0) # mul_add_c(a[3],b[0],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $at,$c_2,$t_2
+ $ADDU $c_3,$at
+ mflo ($t_1,$a_3,$b_0)
+ mfhi ($t_2,$a_3,$b_0)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_3,$b_1) # mul_add_c(a[3],b[1],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $at,$c_2,$t_2
+ $ADDU $c_3,$at
+ $ST $c_1,3*$BNSZ($a0)
+
+ mflo ($t_1,$a_3,$b_1)
+ mfhi ($t_2,$a_3,$b_1)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_2,$b_2) # mul_add_c(a[2],b[2],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $c_1,$c_3,$t_2
+ mflo ($t_1,$a_2,$b_2)
+ mfhi ($t_2,$a_2,$b_2)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_1,$b_3) # mul_add_c(a[1],b[3],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $at,$c_3,$t_2
+ $ADDU $c_1,$at
+ mflo ($t_1,$a_1,$b_3)
+ mfhi ($t_2,$a_1,$b_3)
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_2,$b_3) # mul_add_c(a[2],b[3],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $at,$c_3,$t_2
+ $ADDU $c_1,$at
+ $ST $c_2,4*$BNSZ($a0)
+
+ mflo ($t_1,$a_2,$b_3)
+ mfhi ($t_2,$a_2,$b_3)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_3,$b_2) # mul_add_c(a[3],b[2],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $c_2,$c_1,$t_2
+ mflo ($t_1,$a_3,$b_2)
+ mfhi ($t_2,$a_3,$b_2)
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_3,$b_3) # mul_add_c(a[3],b[3],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $at,$c_1,$t_2
+ $ADDU $c_2,$at
+ $ST $c_3,5*$BNSZ($a0)
+
+ mflo ($t_1,$a_3,$b_3)
+ mfhi ($t_2,$a_3,$b_3)
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ $ST $c_1,6*$BNSZ($a0)
+ $ST $c_2,7*$BNSZ($a0)
+
+ .set noreorder
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ $REG_L $t3,4*$SZREG($sp)
+ $REG_L $t2,3*$SZREG($sp)
+ $REG_L $t1,2*$SZREG($sp)
+ $REG_L $t0,1*$SZREG($sp)
+ $REG_L $gp,0*$SZREG($sp)
+ $PTR_ADD $sp,6*$SZREG
+___
+$code.=<<___;
+ jr $ra
+ nop
+.end bn_mul_comba4
+___
+
+($a_4,$a_5,$a_6,$a_7)=($b_0,$b_1,$b_2,$b_3);
+
+sub add_c2 () {
+my ($hi,$lo,$c0,$c1,$c2,
+ $warm, # !$warm denotes first call with specific sequence of
+ # $c_[XYZ] when there is no Z-carry to accumulate yet;
+ $an,$bn # these two are arguments for multiplication which
+ # result is used in *next* step [which is why it's
+ # commented as "forward multiplication" below];
+ )=@_;
+$code.=<<___;
+ $ADDU $c0,$lo
+ sltu $at,$c0,$lo
+ $MULTU ($an,$bn) # forward multiplication
+ $ADDU $c0,$lo
+ $ADDU $at,$hi
+ sltu $lo,$c0,$lo
+ $ADDU $c1,$at
+ $ADDU $hi,$lo
+___
+$code.=<<___ if (!$warm);
+ sltu $c2,$c1,$at
+ $ADDU $c1,$hi
+___
+$code.=<<___ if ($warm);
+ sltu $at,$c1,$at
+ $ADDU $c1,$hi
+ $ADDU $c2,$at
+___
+$code.=<<___;
+ sltu $hi,$c1,$hi
+ $ADDU $c2,$hi
+ mflo ($lo,$an,$bn)
+ mfhi ($hi,$an,$bn)
+___
+}
+
+$code.=<<___;
+
+.align 5
+.globl bn_sqr_comba8
+.ent bn_sqr_comba8
+bn_sqr_comba8:
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ .frame $sp,6*$SZREG,$ra
+ .mask 0x8000f008,-$SZREG
+ .set noreorder
+ $PTR_SUB $sp,6*$SZREG
+ $REG_S $ra,5*$SZREG($sp)
+ $REG_S $t3,4*$SZREG($sp)
+ $REG_S $t2,3*$SZREG($sp)
+ $REG_S $t1,2*$SZREG($sp)
+ $REG_S $t0,1*$SZREG($sp)
+ $REG_S $gp,0*$SZREG($sp)
+___
+$code.=<<___;
+ .set reorder
+ $LD $a_0,0($a1)
+ $LD $a_1,$BNSZ($a1)
+ $LD $a_2,2*$BNSZ($a1)
+ $LD $a_3,3*$BNSZ($a1)
+
+ $MULTU ($a_0,$a_0) # mul_add_c(a[0],b[0],c1,c2,c3);
+ $LD $a_4,4*$BNSZ($a1)
+ $LD $a_5,5*$BNSZ($a1)
+ $LD $a_6,6*$BNSZ($a1)
+ $LD $a_7,7*$BNSZ($a1)
+ mflo ($c_1,$a_0,$a_0)
+ mfhi ($c_2,$a_0,$a_0)
+ $ST $c_1,0($a0)
+
+ $MULTU ($a_0,$a_1) # mul_add_c2(a[0],b[1],c2,c3,c1);
+ mflo ($t_1,$a_0,$a_1)
+ mfhi ($t_2,$a_0,$a_1)
+ slt $c_1,$t_2,$zero
+ $SLL $t_2,1
+ $MULTU ($a_2,$a_0) # mul_add_c2(a[2],b[0],c3,c1,c2);
+ slt $a2,$t_1,$zero
+ $ADDU $t_2,$a2
+ $SLL $t_1,1
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $ADDU $c_3,$t_2,$at
+ $ST $c_2,$BNSZ($a0)
+ sltu $at,$c_3,$t_2
+ $ADDU $c_1,$at
+ mflo ($t_1,$a_2,$a_0)
+ mfhi ($t_2,$a_2,$a_0)
+___
+ &add_c2($t_2,$t_1,$c_3,$c_1,$c_2,0,
+ $a_1,$a_1); # mul_add_c(a[1],b[1],c3,c1,c2);
+$code.=<<___;
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_0,$a_3) # mul_add_c2(a[0],b[3],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $at,$c_1,$t_2
+ $ADDU $c_2,$at
+ $ST $c_3,2*$BNSZ($a0)
+ mflo ($t_1,$a_0,$a_3)
+ mfhi ($t_2,$a_0,$a_3)
+___
+ &add_c2($t_2,$t_1,$c_1,$c_2,$c_3,0,
+ $a_1,$a_2); # mul_add_c2(a[1],b[2],c1,c2,c3);
+ &add_c2($t_2,$t_1,$c_1,$c_2,$c_3,1,
+ $a_4,$a_0); # mul_add_c2(a[4],b[0],c2,c3,c1);
+$code.=<<___;
+ $ST $c_1,3*$BNSZ($a0)
+___
+ &add_c2($t_2,$t_1,$c_2,$c_3,$c_1,0,
+ $a_3,$a_1); # mul_add_c2(a[3],b[1],c2,c3,c1);
+ &add_c2($t_2,$t_1,$c_2,$c_3,$c_1,1,
+ $a_2,$a_2); # mul_add_c(a[2],b[2],c2,c3,c1);
+$code.=<<___;
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_0,$a_5) # mul_add_c2(a[0],b[5],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $at,$c_3,$t_2
+ $ADDU $c_1,$at
+ $ST $c_2,4*$BNSZ($a0)
+ mflo ($t_1,$a_0,$a_5)
+ mfhi ($t_2,$a_0,$a_5)
+___
+ &add_c2($t_2,$t_1,$c_3,$c_1,$c_2,0,
+ $a_1,$a_4); # mul_add_c2(a[1],b[4],c3,c1,c2);
+ &add_c2($t_2,$t_1,$c_3,$c_1,$c_2,1,
+ $a_2,$a_3); # mul_add_c2(a[2],b[3],c3,c1,c2);
+ &add_c2($t_2,$t_1,$c_3,$c_1,$c_2,1,
+ $a_6,$a_0); # mul_add_c2(a[6],b[0],c1,c2,c3);
+$code.=<<___;
+ $ST $c_3,5*$BNSZ($a0)
+___
+ &add_c2($t_2,$t_1,$c_1,$c_2,$c_3,0,
+ $a_5,$a_1); # mul_add_c2(a[5],b[1],c1,c2,c3);
+ &add_c2($t_2,$t_1,$c_1,$c_2,$c_3,1,
+ $a_4,$a_2); # mul_add_c2(a[4],b[2],c1,c2,c3);
+ &add_c2($t_2,$t_1,$c_1,$c_2,$c_3,1,
+ $a_3,$a_3); # mul_add_c(a[3],b[3],c1,c2,c3);
+$code.=<<___;
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_0,$a_7) # mul_add_c2(a[0],b[7],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $at,$c_2,$t_2
+ $ADDU $c_3,$at
+ $ST $c_1,6*$BNSZ($a0)
+ mflo ($t_1,$a_0,$a_7)
+ mfhi ($t_2,$a_0,$a_7)
+___
+ &add_c2($t_2,$t_1,$c_2,$c_3,$c_1,0,
+ $a_1,$a_6); # mul_add_c2(a[1],b[6],c2,c3,c1);
+ &add_c2($t_2,$t_1,$c_2,$c_3,$c_1,1,
+ $a_2,$a_5); # mul_add_c2(a[2],b[5],c2,c3,c1);
+ &add_c2($t_2,$t_1,$c_2,$c_3,$c_1,1,
+ $a_3,$a_4); # mul_add_c2(a[3],b[4],c2,c3,c1);
+ &add_c2($t_2,$t_1,$c_2,$c_3,$c_1,1,
+ $a_7,$a_1); # mul_add_c2(a[7],b[1],c3,c1,c2);
+$code.=<<___;
+ $ST $c_2,7*$BNSZ($a0)
+___
+ &add_c2($t_2,$t_1,$c_3,$c_1,$c_2,0,
+ $a_6,$a_2); # mul_add_c2(a[6],b[2],c3,c1,c2);
+ &add_c2($t_2,$t_1,$c_3,$c_1,$c_2,1,
+ $a_5,$a_3); # mul_add_c2(a[5],b[3],c3,c1,c2);
+ &add_c2($t_2,$t_1,$c_3,$c_1,$c_2,1,
+ $a_4,$a_4); # mul_add_c(a[4],b[4],c3,c1,c2);
+$code.=<<___;
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_2,$a_7) # mul_add_c2(a[2],b[7],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $at,$c_1,$t_2
+ $ADDU $c_2,$at
+ $ST $c_3,8*$BNSZ($a0)
+ mflo ($t_1,$a_2,$a_7)
+ mfhi ($t_2,$a_2,$a_7)
+___
+ &add_c2($t_2,$t_1,$c_1,$c_2,$c_3,0,
+ $a_3,$a_6); # mul_add_c2(a[3],b[6],c1,c2,c3);
+ &add_c2($t_2,$t_1,$c_1,$c_2,$c_3,1,
+ $a_4,$a_5); # mul_add_c2(a[4],b[5],c1,c2,c3);
+ &add_c2($t_2,$t_1,$c_1,$c_2,$c_3,1,
+ $a_7,$a_3); # mul_add_c2(a[7],b[3],c2,c3,c1);
+$code.=<<___;
+ $ST $c_1,9*$BNSZ($a0)
+___
+ &add_c2($t_2,$t_1,$c_2,$c_3,$c_1,0,
+ $a_6,$a_4); # mul_add_c2(a[6],b[4],c2,c3,c1);
+ &add_c2($t_2,$t_1,$c_2,$c_3,$c_1,1,
+ $a_5,$a_5); # mul_add_c(a[5],b[5],c2,c3,c1);
+$code.=<<___;
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_4,$a_7) # mul_add_c2(a[4],b[7],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $at,$c_3,$t_2
+ $ADDU $c_1,$at
+ $ST $c_2,10*$BNSZ($a0)
+ mflo ($t_1,$a_4,$a_7)
+ mfhi ($t_2,$a_4,$a_7)
+___
+ &add_c2($t_2,$t_1,$c_3,$c_1,$c_2,0,
+ $a_5,$a_6); # mul_add_c2(a[5],b[6],c3,c1,c2);
+ &add_c2($t_2,$t_1,$c_3,$c_1,$c_2,1,
+ $a_7,$a_5); # mul_add_c2(a[7],b[5],c1,c2,c3);
+$code.=<<___;
+ $ST $c_3,11*$BNSZ($a0)
+___
+ &add_c2($t_2,$t_1,$c_1,$c_2,$c_3,0,
+ $a_6,$a_6); # mul_add_c(a[6],b[6],c1,c2,c3);
+$code.=<<___;
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $MULTU ($a_6,$a_7) # mul_add_c2(a[6],b[7],c2,c3,c1);
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ sltu $at,$c_2,$t_2
+ $ADDU $c_3,$at
+ $ST $c_1,12*$BNSZ($a0)
+ mflo ($t_1,$a_6,$a_7)
+ mfhi ($t_2,$a_6,$a_7)
+___
+ &add_c2($t_2,$t_1,$c_2,$c_3,$c_1,0,
+ $a_7,$a_7); # mul_add_c(a[7],b[7],c3,c1,c2);
+$code.=<<___;
+ $ST $c_2,13*$BNSZ($a0)
+
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ $ST $c_3,14*$BNSZ($a0)
+ $ST $c_1,15*$BNSZ($a0)
+
+ .set noreorder
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ $REG_L $t3,4*$SZREG($sp)
+ $REG_L $t2,3*$SZREG($sp)
+ $REG_L $t1,2*$SZREG($sp)
+ $REG_L $t0,1*$SZREG($sp)
+ $REG_L $gp,0*$SZREG($sp)
+ $PTR_ADD $sp,6*$SZREG
+___
+$code.=<<___;
+ jr $ra
+ nop
+.end bn_sqr_comba8
+
+.align 5
+.globl bn_sqr_comba4
+.ent bn_sqr_comba4
+bn_sqr_comba4:
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ .frame $sp,6*$SZREG,$ra
+ .mask 0x8000f008,-$SZREG
+ .set noreorder
+ $PTR_SUB $sp,6*$SZREG
+ $REG_S $ra,5*$SZREG($sp)
+ $REG_S $t3,4*$SZREG($sp)
+ $REG_S $t2,3*$SZREG($sp)
+ $REG_S $t1,2*$SZREG($sp)
+ $REG_S $t0,1*$SZREG($sp)
+ $REG_S $gp,0*$SZREG($sp)
+___
+$code.=<<___;
+ .set reorder
+ $LD $a_0,0($a1)
+ $LD $a_1,$BNSZ($a1)
+ $MULTU ($a_0,$a_0) # mul_add_c(a[0],b[0],c1,c2,c3);
+ $LD $a_2,2*$BNSZ($a1)
+ $LD $a_3,3*$BNSZ($a1)
+ mflo ($c_1,$a_0,$a_0)
+ mfhi ($c_2,$a_0,$a_0)
+ $ST $c_1,0($a0)
+
+ $MULTU ($a_0,$a_1) # mul_add_c2(a[0],b[1],c2,c3,c1);
+ mflo ($t_1,$a_0,$a_1)
+ mfhi ($t_2,$a_0,$a_1)
+ slt $c_1,$t_2,$zero
+ $SLL $t_2,1
+ $MULTU ($a_2,$a_0) # mul_add_c2(a[2],b[0],c3,c1,c2);
+ slt $a2,$t_1,$zero
+ $ADDU $t_2,$a2
+ $SLL $t_1,1
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $ADDU $c_3,$t_2,$at
+ $ST $c_2,$BNSZ($a0)
+ sltu $at,$c_3,$t_2
+ $ADDU $c_1,$at
+ mflo ($t_1,$a_2,$a_0)
+ mfhi ($t_2,$a_2,$a_0)
+___
+ &add_c2($t_2,$t_1,$c_3,$c_1,$c_2,0,
+ $a_1,$a_1); # mul_add_c(a[1],b[1],c3,c1,c2);
+$code.=<<___;
+ $ADDU $c_3,$t_1
+ sltu $at,$c_3,$t_1
+ $MULTU ($a_0,$a_3) # mul_add_c2(a[0],b[3],c1,c2,c3);
+ $ADDU $t_2,$at
+ $ADDU $c_1,$t_2
+ sltu $at,$c_1,$t_2
+ $ADDU $c_2,$at
+ $ST $c_3,2*$BNSZ($a0)
+ mflo ($t_1,$a_0,$a_3)
+ mfhi ($t_2,$a_0,$a_3)
+___
+ &add_c2($t_2,$t_1,$c_1,$c_2,$c_3,0,
+ $a_1,$a_2); # mul_add_c2(a2[1],b[2],c1,c2,c3);
+ &add_c2($t_2,$t_1,$c_1,$c_2,$c_3,1,
+ $a_3,$a_1); # mul_add_c2(a[3],b[1],c2,c3,c1);
+$code.=<<___;
+ $ST $c_1,3*$BNSZ($a0)
+___
+ &add_c2($t_2,$t_1,$c_2,$c_3,$c_1,0,
+ $a_2,$a_2); # mul_add_c(a[2],b[2],c2,c3,c1);
+$code.=<<___;
+ $ADDU $c_2,$t_1
+ sltu $at,$c_2,$t_1
+ $MULTU ($a_2,$a_3) # mul_add_c2(a[2],b[3],c3,c1,c2);
+ $ADDU $t_2,$at
+ $ADDU $c_3,$t_2
+ sltu $at,$c_3,$t_2
+ $ADDU $c_1,$at
+ $ST $c_2,4*$BNSZ($a0)
+ mflo ($t_1,$a_2,$a_3)
+ mfhi ($t_2,$a_2,$a_3)
+___
+ &add_c2($t_2,$t_1,$c_3,$c_1,$c_2,0,
+ $a_3,$a_3); # mul_add_c(a[3],b[3],c1,c2,c3);
+$code.=<<___;
+ $ST $c_3,5*$BNSZ($a0)
+
+ $ADDU $c_1,$t_1
+ sltu $at,$c_1,$t_1
+ $ADDU $t_2,$at
+ $ADDU $c_2,$t_2
+ $ST $c_1,6*$BNSZ($a0)
+ $ST $c_2,7*$BNSZ($a0)
+
+ .set noreorder
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+ $REG_L $t3,4*$SZREG($sp)
+ $REG_L $t2,3*$SZREG($sp)
+ $REG_L $t1,2*$SZREG($sp)
+ $REG_L $t0,1*$SZREG($sp)
+ $REG_L $gp,0*$SZREG($sp)
+ $PTR_ADD $sp,6*$SZREG
+___
+$code.=<<___;
+ jr $ra
+ nop
+.end bn_sqr_comba4
+___
+print $code;
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/parisc-mont.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/parisc-mont.pl
new file mode 100644
index 0000000..6a7c714
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/parisc-mont.pl
@@ -0,0 +1,1006 @@
+#! /usr/bin/env perl
+# Copyright 2009-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+
+# On PA-7100LC this module performs ~90-50% better, less for longer
+# keys, than code generated by gcc 3.2 for PA-RISC 1.1. Latter means
+# that compiler utilized xmpyu instruction to perform 32x32=64-bit
+# multiplication, which in turn means that "baseline" performance was
+# optimal in respect to instruction set capabilities. Fair comparison
+# with vendor compiler is problematic, because OpenSSL doesn't define
+# BN_LLONG [presumably] for historical reasons, which drives compiler
+# toward 4 times 16x16=32-bit multiplications [plus complementary
+# shifts and additions] instead. This means that you should observe
+# several times improvement over code generated by vendor compiler
+# for PA-RISC 1.1, but the "baseline" is far from optimal. The actual
+# improvement coefficient was never collected on PA-7100LC, or any
+# other 1.1 CPU, because I don't have access to such machine with
+# vendor compiler. But to give you a taste, PA-RISC 1.1 code path
+# reportedly outperformed code generated by cc +DA1.1 +O3 by factor
+# of ~5x on PA-8600.
+#
+# On PA-RISC 2.0 it has to compete with pa-risc2[W].s, which is
+# reportedly ~2x faster than vendor compiler generated code [according
+# to comment in pa-risc2[W].s]. Here comes a catch. Execution core of
+# this implementation is actually 32-bit one, in the sense that it
+# operates on 32-bit values. But pa-risc2[W].s operates on arrays of
+# 64-bit BN_LONGs... How do they interoperate then? No problem. This
+# module picks halves of 64-bit values in reverse order and pretends
+# they were 32-bit BN_LONGs. But can 32-bit core compete with "pure"
+# 64-bit code such as pa-risc2[W].s then? Well, the thing is that
+# 32x32=64-bit multiplication is the best even PA-RISC 2.0 can do,
+# i.e. there is no "wider" multiplication like on most other 64-bit
+# platforms. This means that even being effectively 32-bit, this
+# implementation performs "64-bit" computational task in same amount
+# of arithmetic operations, most notably multiplications. It requires
+# more memory references, most notably to tp[num], but this doesn't
+# seem to exhaust memory port capacity. And indeed, dedicated PA-RISC
+# 2.0 code path provides virtually same performance as pa-risc2[W].s:
+# it's ~10% better for shortest key length and ~10% worse for longest
+# one.
+#
+# In case it wasn't clear. The module has two distinct code paths:
+# PA-RISC 1.1 and PA-RISC 2.0 ones. Latter features carry-free 64-bit
+# additions and 64-bit integer loads, not to mention specific
+# instruction scheduling. In 64-bit build naturally only 2.0 code path
+# is assembled. In 32-bit application context both code paths are
+# assembled, PA-RISC 2.0 CPU is detected at run-time and proper path
+# is taken automatically. Also, in 32-bit build the module imposes
+# couple of limitations: vector lengths has to be even and vector
+# addresses has to be 64-bit aligned. Normally neither is a problem:
+# most common key lengths are even and vectors are commonly malloc-ed,
+# which ensures alignment.
+#
+# Special thanks to polarhome.com for providing HP-UX account on
+# PA-RISC 1.1 machine, and to correspondent who chose to remain
+# anonymous for testing the code on PA-RISC 2.0 machine.
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+
+$flavour = shift;
+$output = shift;
+
+open STDOUT,">$output";
+
+if ($flavour =~ /64/) {
+ $LEVEL ="2.0W";
+ $SIZE_T =8;
+ $FRAME_MARKER =80;
+ $SAVED_RP =16;
+ $PUSH ="std";
+ $PUSHMA ="std,ma";
+ $POP ="ldd";
+ $POPMB ="ldd,mb";
+ $BN_SZ =$SIZE_T;
+} else {
+ $LEVEL ="1.1"; #$LEVEL.="\n\t.ALLOW\t2.0";
+ $SIZE_T =4;
+ $FRAME_MARKER =48;
+ $SAVED_RP =20;
+ $PUSH ="stw";
+ $PUSHMA ="stwm";
+ $POP ="ldw";
+ $POPMB ="ldwm";
+ $BN_SZ =$SIZE_T;
+ if (open CONF,"<${dir}../../opensslconf.h") {
+ while(<CONF>) {
+ if (m/#\s*define\s+SIXTY_FOUR_BIT/) {
+ $BN_SZ=8;
+ $LEVEL="2.0";
+ last;
+ }
+ }
+ close CONF;
+ }
+}
+
+$FRAME=8*$SIZE_T+$FRAME_MARKER; # 8 saved regs + frame marker
+ # [+ argument transfer]
+$LOCALS=$FRAME-$FRAME_MARKER;
+$FRAME+=32; # local variables
+
+$tp="%r31";
+$ti1="%r29";
+$ti0="%r28";
+
+$rp="%r26";
+$ap="%r25";
+$bp="%r24";
+$np="%r23";
+$n0="%r22"; # passed through stack in 32-bit
+$num="%r21"; # passed through stack in 32-bit
+$idx="%r20";
+$arrsz="%r19";
+
+$nm1="%r7";
+$nm0="%r6";
+$ab1="%r5";
+$ab0="%r4";
+
+$fp="%r3";
+$hi1="%r2";
+$hi0="%r1";
+
+$xfer=$n0; # accommodates [-16..15] offset in fld[dw]s
+
+$fm0="%fr4"; $fti=$fm0;
+$fbi="%fr5L";
+$fn0="%fr5R";
+$fai="%fr6"; $fab0="%fr7"; $fab1="%fr8";
+$fni="%fr9"; $fnm0="%fr10"; $fnm1="%fr11";
+
+$code=<<___;
+ .LEVEL $LEVEL
+ .SPACE \$TEXT\$
+ .SUBSPA \$CODE\$,QUAD=0,ALIGN=8,ACCESS=0x2C,CODE_ONLY
+
+ .EXPORT bn_mul_mont,ENTRY,ARGW0=GR,ARGW1=GR,ARGW2=GR,ARGW3=GR
+ .ALIGN 64
+bn_mul_mont
+ .PROC
+ .CALLINFO FRAME=`$FRAME-8*$SIZE_T`,NO_CALLS,SAVE_RP,SAVE_SP,ENTRY_GR=6
+ .ENTRY
+ $PUSH %r2,-$SAVED_RP(%sp) ; standard prologue
+ $PUSHMA %r3,$FRAME(%sp)
+ $PUSH %r4,`-$FRAME+1*$SIZE_T`(%sp)
+ $PUSH %r5,`-$FRAME+2*$SIZE_T`(%sp)
+ $PUSH %r6,`-$FRAME+3*$SIZE_T`(%sp)
+ $PUSH %r7,`-$FRAME+4*$SIZE_T`(%sp)
+ $PUSH %r8,`-$FRAME+5*$SIZE_T`(%sp)
+ $PUSH %r9,`-$FRAME+6*$SIZE_T`(%sp)
+ $PUSH %r10,`-$FRAME+7*$SIZE_T`(%sp)
+ ldo -$FRAME(%sp),$fp
+___
+$code.=<<___ if ($SIZE_T==4);
+ ldw `-$FRAME_MARKER-4`($fp),$n0
+ ldw `-$FRAME_MARKER-8`($fp),$num
+ nop
+ nop ; alignment
+___
+$code.=<<___ if ($BN_SZ==4);
+ comiclr,<= 6,$num,%r0 ; are vectors long enough?
+ b L\$abort
+ ldi 0,%r28 ; signal "unhandled"
+ add,ev %r0,$num,$num ; is $num even?
+ b L\$abort
+ nop
+ or $ap,$np,$ti1
+ extru,= $ti1,31,3,%r0 ; are ap and np 64-bit aligned?
+ b L\$abort
+ nop
+ nop ; alignment
+ nop
+
+ fldws 0($n0),${fn0}
+ fldws,ma 4($bp),${fbi} ; bp[0]
+___
+$code.=<<___ if ($BN_SZ==8);
+ comib,> 3,$num,L\$abort ; are vectors long enough?
+ ldi 0,%r28 ; signal "unhandled"
+ addl $num,$num,$num ; I operate on 32-bit values
+
+ fldws 4($n0),${fn0} ; only low part of n0
+ fldws 4($bp),${fbi} ; bp[0] in flipped word order
+___
+$code.=<<___;
+ fldds 0($ap),${fai} ; ap[0,1]
+ fldds 0($np),${fni} ; np[0,1]
+
+ sh2addl $num,%r0,$arrsz
+ ldi 31,$hi0
+ ldo 36($arrsz),$hi1 ; space for tp[num+1]
+ andcm $hi1,$hi0,$hi1 ; align
+ addl $hi1,%sp,%sp
+ $PUSH $fp,-$SIZE_T(%sp)
+
+ ldo `$LOCALS+16`($fp),$xfer
+ ldo `$LOCALS+32+4`($fp),$tp
+
+ xmpyu ${fai}L,${fbi},${fab0} ; ap[0]*bp[0]
+ xmpyu ${fai}R,${fbi},${fab1} ; ap[1]*bp[0]
+ xmpyu ${fn0},${fab0}R,${fm0}
+
+ addl $arrsz,$ap,$ap ; point at the end
+ addl $arrsz,$np,$np
+ subi 0,$arrsz,$idx ; j=0
+ ldo 8($idx),$idx ; j++++
+
+ xmpyu ${fni}L,${fm0}R,${fnm0} ; np[0]*m
+ xmpyu ${fni}R,${fm0}R,${fnm1} ; np[1]*m
+ fstds ${fab0},-16($xfer)
+ fstds ${fnm0},-8($xfer)
+ fstds ${fab1},0($xfer)
+ fstds ${fnm1},8($xfer)
+ flddx $idx($ap),${fai} ; ap[2,3]
+ flddx $idx($np),${fni} ; np[2,3]
+___
+$code.=<<___ if ($BN_SZ==4);
+ mtctl $hi0,%cr11 ; $hi0 still holds 31
+ extrd,u,*= $hi0,%sar,1,$hi0 ; executes on PA-RISC 1.0
+ b L\$parisc11
+ nop
+___
+$code.=<<___; # PA-RISC 2.0 code-path
+ xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[0]
+ xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m
+ ldd -16($xfer),$ab0
+ fstds ${fab0},-16($xfer)
+
+ extrd,u $ab0,31,32,$hi0
+ extrd,u $ab0,63,32,$ab0
+ ldd -8($xfer),$nm0
+ fstds ${fnm0},-8($xfer)
+ ldo 8($idx),$idx ; j++++
+ addl $ab0,$nm0,$nm0 ; low part is discarded
+ extrd,u $nm0,31,32,$hi1
+
+L\$1st
+ xmpyu ${fai}R,${fbi},${fab1} ; ap[j+1]*bp[0]
+ xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j+1]*m
+ ldd 0($xfer),$ab1
+ fstds ${fab1},0($xfer)
+ addl $hi0,$ab1,$ab1
+ extrd,u $ab1,31,32,$hi0
+ ldd 8($xfer),$nm1
+ fstds ${fnm1},8($xfer)
+ extrd,u $ab1,63,32,$ab1
+ addl $hi1,$nm1,$nm1
+ flddx $idx($ap),${fai} ; ap[j,j+1]
+ flddx $idx($np),${fni} ; np[j,j+1]
+ addl $ab1,$nm1,$nm1
+ extrd,u $nm1,31,32,$hi1
+
+ xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[0]
+ xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m
+ ldd -16($xfer),$ab0
+ fstds ${fab0},-16($xfer)
+ addl $hi0,$ab0,$ab0
+ extrd,u $ab0,31,32,$hi0
+ ldd -8($xfer),$nm0
+ fstds ${fnm0},-8($xfer)
+ extrd,u $ab0,63,32,$ab0
+ addl $hi1,$nm0,$nm0
+ stw $nm1,-4($tp) ; tp[j-1]
+ addl $ab0,$nm0,$nm0
+ stw,ma $nm0,8($tp) ; tp[j-1]
+ addib,<> 8,$idx,L\$1st ; j++++
+ extrd,u $nm0,31,32,$hi1
+
+ xmpyu ${fai}R,${fbi},${fab1} ; ap[j]*bp[0]
+ xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j]*m
+ ldd 0($xfer),$ab1
+ fstds ${fab1},0($xfer)
+ addl $hi0,$ab1,$ab1
+ extrd,u $ab1,31,32,$hi0
+ ldd 8($xfer),$nm1
+ fstds ${fnm1},8($xfer)
+ extrd,u $ab1,63,32,$ab1
+ addl $hi1,$nm1,$nm1
+ ldd -16($xfer),$ab0
+ addl $ab1,$nm1,$nm1
+ ldd -8($xfer),$nm0
+ extrd,u $nm1,31,32,$hi1
+
+ addl $hi0,$ab0,$ab0
+ extrd,u $ab0,31,32,$hi0
+ stw $nm1,-4($tp) ; tp[j-1]
+ extrd,u $ab0,63,32,$ab0
+ addl $hi1,$nm0,$nm0
+ ldd 0($xfer),$ab1
+ addl $ab0,$nm0,$nm0
+ ldd,mb 8($xfer),$nm1
+ extrd,u $nm0,31,32,$hi1
+ stw,ma $nm0,8($tp) ; tp[j-1]
+
+ ldo -1($num),$num ; i--
+ subi 0,$arrsz,$idx ; j=0
+___
+$code.=<<___ if ($BN_SZ==4);
+ fldws,ma 4($bp),${fbi} ; bp[1]
+___
+$code.=<<___ if ($BN_SZ==8);
+ fldws 0($bp),${fbi} ; bp[1] in flipped word order
+___
+$code.=<<___;
+ flddx $idx($ap),${fai} ; ap[0,1]
+ flddx $idx($np),${fni} ; np[0,1]
+ fldws 8($xfer),${fti}R ; tp[0]
+ addl $hi0,$ab1,$ab1
+ extrd,u $ab1,31,32,$hi0
+ extrd,u $ab1,63,32,$ab1
+ ldo 8($idx),$idx ; j++++
+ xmpyu ${fai}L,${fbi},${fab0} ; ap[0]*bp[1]
+ xmpyu ${fai}R,${fbi},${fab1} ; ap[1]*bp[1]
+ addl $hi1,$nm1,$nm1
+ addl $ab1,$nm1,$nm1
+ extrd,u $nm1,31,32,$hi1
+ fstws,mb ${fab0}L,-8($xfer) ; save high part
+ stw $nm1,-4($tp) ; tp[j-1]
+
+ fcpy,sgl %fr0,${fti}L ; zero high part
+ fcpy,sgl %fr0,${fab0}L
+ addl $hi1,$hi0,$hi0
+ extrd,u $hi0,31,32,$hi1
+ fcnvxf,dbl,dbl ${fti},${fti} ; 32-bit unsigned int -> double
+ fcnvxf,dbl,dbl ${fab0},${fab0}
+ stw $hi0,0($tp)
+ stw $hi1,4($tp)
+
+ fadd,dbl ${fti},${fab0},${fab0} ; add tp[0]
+ fcnvfx,dbl,dbl ${fab0},${fab0} ; double -> 33-bit unsigned int
+ xmpyu ${fn0},${fab0}R,${fm0}
+ ldo `$LOCALS+32+4`($fp),$tp
+L\$outer
+ xmpyu ${fni}L,${fm0}R,${fnm0} ; np[0]*m
+ xmpyu ${fni}R,${fm0}R,${fnm1} ; np[1]*m
+ fstds ${fab0},-16($xfer) ; 33-bit value
+ fstds ${fnm0},-8($xfer)
+ flddx $idx($ap),${fai} ; ap[2]
+ flddx $idx($np),${fni} ; np[2]
+ ldo 8($idx),$idx ; j++++
+ ldd -16($xfer),$ab0 ; 33-bit value
+ ldd -8($xfer),$nm0
+ ldw 0($xfer),$hi0 ; high part
+
+ xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[i]
+ xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m
+ extrd,u $ab0,31,32,$ti0 ; carry bit
+ extrd,u $ab0,63,32,$ab0
+ fstds ${fab1},0($xfer)
+ addl $ti0,$hi0,$hi0 ; account carry bit
+ fstds ${fnm1},8($xfer)
+ addl $ab0,$nm0,$nm0 ; low part is discarded
+ ldw 0($tp),$ti1 ; tp[1]
+ extrd,u $nm0,31,32,$hi1
+ fstds ${fab0},-16($xfer)
+ fstds ${fnm0},-8($xfer)
+
+L\$inner
+ xmpyu ${fai}R,${fbi},${fab1} ; ap[j+1]*bp[i]
+ xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j+1]*m
+ ldd 0($xfer),$ab1
+ fstds ${fab1},0($xfer)
+ addl $hi0,$ti1,$ti1
+ addl $ti1,$ab1,$ab1
+ ldd 8($xfer),$nm1
+ fstds ${fnm1},8($xfer)
+ extrd,u $ab1,31,32,$hi0
+ extrd,u $ab1,63,32,$ab1
+ flddx $idx($ap),${fai} ; ap[j,j+1]
+ flddx $idx($np),${fni} ; np[j,j+1]
+ addl $hi1,$nm1,$nm1
+ addl $ab1,$nm1,$nm1
+ ldw 4($tp),$ti0 ; tp[j]
+ stw $nm1,-4($tp) ; tp[j-1]
+
+ xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[i]
+ xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m
+ ldd -16($xfer),$ab0
+ fstds ${fab0},-16($xfer)
+ addl $hi0,$ti0,$ti0
+ addl $ti0,$ab0,$ab0
+ ldd -8($xfer),$nm0
+ fstds ${fnm0},-8($xfer)
+ extrd,u $ab0,31,32,$hi0
+ extrd,u $nm1,31,32,$hi1
+ ldw 8($tp),$ti1 ; tp[j]
+ extrd,u $ab0,63,32,$ab0
+ addl $hi1,$nm0,$nm0
+ addl $ab0,$nm0,$nm0
+ stw,ma $nm0,8($tp) ; tp[j-1]
+ addib,<> 8,$idx,L\$inner ; j++++
+ extrd,u $nm0,31,32,$hi1
+
+ xmpyu ${fai}R,${fbi},${fab1} ; ap[j]*bp[i]
+ xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j]*m
+ ldd 0($xfer),$ab1
+ fstds ${fab1},0($xfer)
+ addl $hi0,$ti1,$ti1
+ addl $ti1,$ab1,$ab1
+ ldd 8($xfer),$nm1
+ fstds ${fnm1},8($xfer)
+ extrd,u $ab1,31,32,$hi0
+ extrd,u $ab1,63,32,$ab1
+ ldw 4($tp),$ti0 ; tp[j]
+ addl $hi1,$nm1,$nm1
+ addl $ab1,$nm1,$nm1
+ ldd -16($xfer),$ab0
+ ldd -8($xfer),$nm0
+ extrd,u $nm1,31,32,$hi1
+
+ addl $hi0,$ab0,$ab0
+ addl $ti0,$ab0,$ab0
+ stw $nm1,-4($tp) ; tp[j-1]
+ extrd,u $ab0,31,32,$hi0
+ ldw 8($tp),$ti1 ; tp[j]
+ extrd,u $ab0,63,32,$ab0
+ addl $hi1,$nm0,$nm0
+ ldd 0($xfer),$ab1
+ addl $ab0,$nm0,$nm0
+ ldd,mb 8($xfer),$nm1
+ extrd,u $nm0,31,32,$hi1
+ stw,ma $nm0,8($tp) ; tp[j-1]
+
+ addib,= -1,$num,L\$outerdone ; i--
+ subi 0,$arrsz,$idx ; j=0
+___
+$code.=<<___ if ($BN_SZ==4);
+ fldws,ma 4($bp),${fbi} ; bp[i]
+___
+$code.=<<___ if ($BN_SZ==8);
+ ldi 12,$ti0 ; bp[i] in flipped word order
+ addl,ev %r0,$num,$num
+ ldi -4,$ti0
+ addl $ti0,$bp,$bp
+ fldws 0($bp),${fbi}
+___
+$code.=<<___;
+ flddx $idx($ap),${fai} ; ap[0]
+ addl $hi0,$ab1,$ab1
+ flddx $idx($np),${fni} ; np[0]
+ fldws 8($xfer),${fti}R ; tp[0]
+ addl $ti1,$ab1,$ab1
+ extrd,u $ab1,31,32,$hi0
+ extrd,u $ab1,63,32,$ab1
+
+ ldo 8($idx),$idx ; j++++
+ xmpyu ${fai}L,${fbi},${fab0} ; ap[0]*bp[i]
+ xmpyu ${fai}R,${fbi},${fab1} ; ap[1]*bp[i]
+ ldw 4($tp),$ti0 ; tp[j]
+
+ addl $hi1,$nm1,$nm1
+ fstws,mb ${fab0}L,-8($xfer) ; save high part
+ addl $ab1,$nm1,$nm1
+ extrd,u $nm1,31,32,$hi1
+ fcpy,sgl %fr0,${fti}L ; zero high part
+ fcpy,sgl %fr0,${fab0}L
+ stw $nm1,-4($tp) ; tp[j-1]
+
+ fcnvxf,dbl,dbl ${fti},${fti} ; 32-bit unsigned int -> double
+ fcnvxf,dbl,dbl ${fab0},${fab0}
+ addl $hi1,$hi0,$hi0
+ fadd,dbl ${fti},${fab0},${fab0} ; add tp[0]
+ addl $ti0,$hi0,$hi0
+ extrd,u $hi0,31,32,$hi1
+ fcnvfx,dbl,dbl ${fab0},${fab0} ; double -> 33-bit unsigned int
+ stw $hi0,0($tp)
+ stw $hi1,4($tp)
+ xmpyu ${fn0},${fab0}R,${fm0}
+
+ b L\$outer
+ ldo `$LOCALS+32+4`($fp),$tp
+
+L\$outerdone
+ addl $hi0,$ab1,$ab1
+ addl $ti1,$ab1,$ab1
+ extrd,u $ab1,31,32,$hi0
+ extrd,u $ab1,63,32,$ab1
+
+ ldw 4($tp),$ti0 ; tp[j]
+
+ addl $hi1,$nm1,$nm1
+ addl $ab1,$nm1,$nm1
+ extrd,u $nm1,31,32,$hi1
+ stw $nm1,-4($tp) ; tp[j-1]
+
+ addl $hi1,$hi0,$hi0
+ addl $ti0,$hi0,$hi0
+ extrd,u $hi0,31,32,$hi1
+ stw $hi0,0($tp)
+ stw $hi1,4($tp)
+
+ ldo `$LOCALS+32`($fp),$tp
+ sub %r0,%r0,%r0 ; clear borrow
+___
+$code.=<<___ if ($BN_SZ==4);
+ ldws,ma 4($tp),$ti0
+ extru,= $rp,31,3,%r0 ; is rp 64-bit aligned?
+ b L\$sub_pa11
+ addl $tp,$arrsz,$tp
+L\$sub
+ ldwx $idx($np),$hi0
+ subb $ti0,$hi0,$hi1
+ ldwx $idx($tp),$ti0
+ addib,<> 4,$idx,L\$sub
+ stws,ma $hi1,4($rp)
+
+ subb $ti0,%r0,$hi1
+___
+$code.=<<___ if ($BN_SZ==8);
+ ldd,ma 8($tp),$ti0
+L\$sub
+ ldd $idx($np),$hi0
+ shrpd $ti0,$ti0,32,$ti0 ; flip word order
+ std $ti0,-8($tp) ; save flipped value
+ sub,db $ti0,$hi0,$hi1
+ ldd,ma 8($tp),$ti0
+ addib,<> 8,$idx,L\$sub
+ std,ma $hi1,8($rp)
+
+ extrd,u $ti0,31,32,$ti0 ; carry in flipped word order
+ sub,db $ti0,%r0,$hi1
+___
+$code.=<<___;
+ ldo `$LOCALS+32`($fp),$tp
+ sub $rp,$arrsz,$rp ; rewind rp
+ subi 0,$arrsz,$idx
+L\$copy
+ ldd 0($tp),$ti0
+ ldd 0($rp),$hi0
+ std,ma %r0,8($tp)
+ comiclr,= 0,$hi1,%r0
+ copy $ti0,$hi0
+ addib,<> 8,$idx,L\$copy
+ std,ma $hi0,8($rp)
+___
+
+if ($BN_SZ==4) { # PA-RISC 1.1 code-path
+$ablo=$ab0;
+$abhi=$ab1;
+$nmlo0=$nm0;
+$nmhi0=$nm1;
+$nmlo1="%r9";
+$nmhi1="%r8";
+
+$code.=<<___;
+ b L\$done
+ nop
+
+ .ALIGN 8
+L\$parisc11
+ xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[0]
+ xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m
+ ldw -12($xfer),$ablo
+ ldw -16($xfer),$hi0
+ ldw -4($xfer),$nmlo0
+ ldw -8($xfer),$nmhi0
+ fstds ${fab0},-16($xfer)
+ fstds ${fnm0},-8($xfer)
+
+ ldo 8($idx),$idx ; j++++
+ add $ablo,$nmlo0,$nmlo0 ; discarded
+ addc %r0,$nmhi0,$hi1
+ ldw 4($xfer),$ablo
+ ldw 0($xfer),$abhi
+ nop
+
+L\$1st_pa11
+ xmpyu ${fai}R,${fbi},${fab1} ; ap[j+1]*bp[0]
+ flddx $idx($ap),${fai} ; ap[j,j+1]
+ xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j+1]*m
+ flddx $idx($np),${fni} ; np[j,j+1]
+ add $hi0,$ablo,$ablo
+ ldw 12($xfer),$nmlo1
+ addc %r0,$abhi,$hi0
+ ldw 8($xfer),$nmhi1
+ add $ablo,$nmlo1,$nmlo1
+ fstds ${fab1},0($xfer)
+ addc %r0,$nmhi1,$nmhi1
+ fstds ${fnm1},8($xfer)
+ add $hi1,$nmlo1,$nmlo1
+ ldw -12($xfer),$ablo
+ addc %r0,$nmhi1,$hi1
+ ldw -16($xfer),$abhi
+
+ xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[0]
+ ldw -4($xfer),$nmlo0
+ xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m
+ ldw -8($xfer),$nmhi0
+ add $hi0,$ablo,$ablo
+ stw $nmlo1,-4($tp) ; tp[j-1]
+ addc %r0,$abhi,$hi0
+ fstds ${fab0},-16($xfer)
+ add $ablo,$nmlo0,$nmlo0
+ fstds ${fnm0},-8($xfer)
+ addc %r0,$nmhi0,$nmhi0
+ ldw 0($xfer),$abhi
+ add $hi1,$nmlo0,$nmlo0
+ ldw 4($xfer),$ablo
+ stws,ma $nmlo0,8($tp) ; tp[j-1]
+ addib,<> 8,$idx,L\$1st_pa11 ; j++++
+ addc %r0,$nmhi0,$hi1
+
+ ldw 8($xfer),$nmhi1
+ ldw 12($xfer),$nmlo1
+ xmpyu ${fai}R,${fbi},${fab1} ; ap[j]*bp[0]
+ xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j]*m
+ add $hi0,$ablo,$ablo
+ fstds ${fab1},0($xfer)
+ addc %r0,$abhi,$hi0
+ fstds ${fnm1},8($xfer)
+ add $ablo,$nmlo1,$nmlo1
+ ldw -16($xfer),$abhi
+ addc %r0,$nmhi1,$nmhi1
+ ldw -12($xfer),$ablo
+ add $hi1,$nmlo1,$nmlo1
+ ldw -8($xfer),$nmhi0
+ addc %r0,$nmhi1,$hi1
+ ldw -4($xfer),$nmlo0
+
+ add $hi0,$ablo,$ablo
+ stw $nmlo1,-4($tp) ; tp[j-1]
+ addc %r0,$abhi,$hi0
+ ldw 0($xfer),$abhi
+ add $ablo,$nmlo0,$nmlo0
+ ldw 4($xfer),$ablo
+ addc %r0,$nmhi0,$nmhi0
+ ldws,mb 8($xfer),$nmhi1
+ add $hi1,$nmlo0,$nmlo0
+ ldw 4($xfer),$nmlo1
+ addc %r0,$nmhi0,$hi1
+ stws,ma $nmlo0,8($tp) ; tp[j-1]
+
+ ldo -1($num),$num ; i--
+ subi 0,$arrsz,$idx ; j=0
+
+ fldws,ma 4($bp),${fbi} ; bp[1]
+ flddx $idx($ap),${fai} ; ap[0,1]
+ flddx $idx($np),${fni} ; np[0,1]
+ fldws 8($xfer),${fti}R ; tp[0]
+ add $hi0,$ablo,$ablo
+ addc %r0,$abhi,$hi0
+ ldo 8($idx),$idx ; j++++
+ xmpyu ${fai}L,${fbi},${fab0} ; ap[0]*bp[1]
+ xmpyu ${fai}R,${fbi},${fab1} ; ap[1]*bp[1]
+ add $hi1,$nmlo1,$nmlo1
+ addc %r0,$nmhi1,$nmhi1
+ add $ablo,$nmlo1,$nmlo1
+ addc %r0,$nmhi1,$hi1
+ fstws,mb ${fab0}L,-8($xfer) ; save high part
+ stw $nmlo1,-4($tp) ; tp[j-1]
+
+ fcpy,sgl %fr0,${fti}L ; zero high part
+ fcpy,sgl %fr0,${fab0}L
+ add $hi1,$hi0,$hi0
+ addc %r0,%r0,$hi1
+ fcnvxf,dbl,dbl ${fti},${fti} ; 32-bit unsigned int -> double
+ fcnvxf,dbl,dbl ${fab0},${fab0}
+ stw $hi0,0($tp)
+ stw $hi1,4($tp)
+
+ fadd,dbl ${fti},${fab0},${fab0} ; add tp[0]
+ fcnvfx,dbl,dbl ${fab0},${fab0} ; double -> 33-bit unsigned int
+ xmpyu ${fn0},${fab0}R,${fm0}
+ ldo `$LOCALS+32+4`($fp),$tp
+L\$outer_pa11
+ xmpyu ${fni}L,${fm0}R,${fnm0} ; np[0]*m
+ xmpyu ${fni}R,${fm0}R,${fnm1} ; np[1]*m
+ fstds ${fab0},-16($xfer) ; 33-bit value
+ fstds ${fnm0},-8($xfer)
+ flddx $idx($ap),${fai} ; ap[2,3]
+ flddx $idx($np),${fni} ; np[2,3]
+ ldw -16($xfer),$abhi ; carry bit actually
+ ldo 8($idx),$idx ; j++++
+ ldw -12($xfer),$ablo
+ ldw -8($xfer),$nmhi0
+ ldw -4($xfer),$nmlo0
+ ldw 0($xfer),$hi0 ; high part
+
+ xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[i]
+ xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m
+ fstds ${fab1},0($xfer)
+ addl $abhi,$hi0,$hi0 ; account carry bit
+ fstds ${fnm1},8($xfer)
+ add $ablo,$nmlo0,$nmlo0 ; discarded
+ ldw 0($tp),$ti1 ; tp[1]
+ addc %r0,$nmhi0,$hi1
+ fstds ${fab0},-16($xfer)
+ fstds ${fnm0},-8($xfer)
+ ldw 4($xfer),$ablo
+ ldw 0($xfer),$abhi
+
+L\$inner_pa11
+ xmpyu ${fai}R,${fbi},${fab1} ; ap[j+1]*bp[i]
+ flddx $idx($ap),${fai} ; ap[j,j+1]
+ xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j+1]*m
+ flddx $idx($np),${fni} ; np[j,j+1]
+ add $hi0,$ablo,$ablo
+ ldw 4($tp),$ti0 ; tp[j]
+ addc %r0,$abhi,$abhi
+ ldw 12($xfer),$nmlo1
+ add $ti1,$ablo,$ablo
+ ldw 8($xfer),$nmhi1
+ addc %r0,$abhi,$hi0
+ fstds ${fab1},0($xfer)
+ add $ablo,$nmlo1,$nmlo1
+ fstds ${fnm1},8($xfer)
+ addc %r0,$nmhi1,$nmhi1
+ ldw -12($xfer),$ablo
+ add $hi1,$nmlo1,$nmlo1
+ ldw -16($xfer),$abhi
+ addc %r0,$nmhi1,$hi1
+
+ xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[i]
+ ldw 8($tp),$ti1 ; tp[j]
+ xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m
+ ldw -4($xfer),$nmlo0
+ add $hi0,$ablo,$ablo
+ ldw -8($xfer),$nmhi0
+ addc %r0,$abhi,$abhi
+ stw $nmlo1,-4($tp) ; tp[j-1]
+ add $ti0,$ablo,$ablo
+ fstds ${fab0},-16($xfer)
+ addc %r0,$abhi,$hi0
+ fstds ${fnm0},-8($xfer)
+ add $ablo,$nmlo0,$nmlo0
+ ldw 4($xfer),$ablo
+ addc %r0,$nmhi0,$nmhi0
+ ldw 0($xfer),$abhi
+ add $hi1,$nmlo0,$nmlo0
+ stws,ma $nmlo0,8($tp) ; tp[j-1]
+ addib,<> 8,$idx,L\$inner_pa11 ; j++++
+ addc %r0,$nmhi0,$hi1
+
+ xmpyu ${fai}R,${fbi},${fab1} ; ap[j]*bp[i]
+ ldw 12($xfer),$nmlo1
+ xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j]*m
+ ldw 8($xfer),$nmhi1
+ add $hi0,$ablo,$ablo
+ ldw 4($tp),$ti0 ; tp[j]
+ addc %r0,$abhi,$abhi
+ fstds ${fab1},0($xfer)
+ add $ti1,$ablo,$ablo
+ fstds ${fnm1},8($xfer)
+ addc %r0,$abhi,$hi0
+ ldw -16($xfer),$abhi
+ add $ablo,$nmlo1,$nmlo1
+ ldw -12($xfer),$ablo
+ addc %r0,$nmhi1,$nmhi1
+ ldw -8($xfer),$nmhi0
+ add $hi1,$nmlo1,$nmlo1
+ ldw -4($xfer),$nmlo0
+ addc %r0,$nmhi1,$hi1
+
+ add $hi0,$ablo,$ablo
+ stw $nmlo1,-4($tp) ; tp[j-1]
+ addc %r0,$abhi,$abhi
+ add $ti0,$ablo,$ablo
+ ldw 8($tp),$ti1 ; tp[j]
+ addc %r0,$abhi,$hi0
+ ldw 0($xfer),$abhi
+ add $ablo,$nmlo0,$nmlo0
+ ldw 4($xfer),$ablo
+ addc %r0,$nmhi0,$nmhi0
+ ldws,mb 8($xfer),$nmhi1
+ add $hi1,$nmlo0,$nmlo0
+ ldw 4($xfer),$nmlo1
+ addc %r0,$nmhi0,$hi1
+ stws,ma $nmlo0,8($tp) ; tp[j-1]
+
+ addib,= -1,$num,L\$outerdone_pa11; i--
+ subi 0,$arrsz,$idx ; j=0
+
+ fldws,ma 4($bp),${fbi} ; bp[i]
+ flddx $idx($ap),${fai} ; ap[0]
+ add $hi0,$ablo,$ablo
+ addc %r0,$abhi,$abhi
+ flddx $idx($np),${fni} ; np[0]
+ fldws 8($xfer),${fti}R ; tp[0]
+ add $ti1,$ablo,$ablo
+ addc %r0,$abhi,$hi0
+
+ ldo 8($idx),$idx ; j++++
+ xmpyu ${fai}L,${fbi},${fab0} ; ap[0]*bp[i]
+ xmpyu ${fai}R,${fbi},${fab1} ; ap[1]*bp[i]
+ ldw 4($tp),$ti0 ; tp[j]
+
+ add $hi1,$nmlo1,$nmlo1
+ addc %r0,$nmhi1,$nmhi1
+ fstws,mb ${fab0}L,-8($xfer) ; save high part
+ add $ablo,$nmlo1,$nmlo1
+ addc %r0,$nmhi1,$hi1
+ fcpy,sgl %fr0,${fti}L ; zero high part
+ fcpy,sgl %fr0,${fab0}L
+ stw $nmlo1,-4($tp) ; tp[j-1]
+
+ fcnvxf,dbl,dbl ${fti},${fti} ; 32-bit unsigned int -> double
+ fcnvxf,dbl,dbl ${fab0},${fab0}
+ add $hi1,$hi0,$hi0
+ addc %r0,%r0,$hi1
+ fadd,dbl ${fti},${fab0},${fab0} ; add tp[0]
+ add $ti0,$hi0,$hi0
+ addc %r0,$hi1,$hi1
+ fcnvfx,dbl,dbl ${fab0},${fab0} ; double -> 33-bit unsigned int
+ stw $hi0,0($tp)
+ stw $hi1,4($tp)
+ xmpyu ${fn0},${fab0}R,${fm0}
+
+ b L\$outer_pa11
+ ldo `$LOCALS+32+4`($fp),$tp
+
+L\$outerdone_pa11
+ add $hi0,$ablo,$ablo
+ addc %r0,$abhi,$abhi
+ add $ti1,$ablo,$ablo
+ addc %r0,$abhi,$hi0
+
+ ldw 4($tp),$ti0 ; tp[j]
+
+ add $hi1,$nmlo1,$nmlo1
+ addc %r0,$nmhi1,$nmhi1
+ add $ablo,$nmlo1,$nmlo1
+ addc %r0,$nmhi1,$hi1
+ stw $nmlo1,-4($tp) ; tp[j-1]
+
+ add $hi1,$hi0,$hi0
+ addc %r0,%r0,$hi1
+ add $ti0,$hi0,$hi0
+ addc %r0,$hi1,$hi1
+ stw $hi0,0($tp)
+ stw $hi1,4($tp)
+
+ ldo `$LOCALS+32+4`($fp),$tp
+ sub %r0,%r0,%r0 ; clear borrow
+ ldw -4($tp),$ti0
+ addl $tp,$arrsz,$tp
+L\$sub_pa11
+ ldwx $idx($np),$hi0
+ subb $ti0,$hi0,$hi1
+ ldwx $idx($tp),$ti0
+ addib,<> 4,$idx,L\$sub_pa11
+ stws,ma $hi1,4($rp)
+
+ subb $ti0,%r0,$hi1
+
+ ldo `$LOCALS+32`($fp),$tp
+ sub $rp,$arrsz,$rp ; rewind rp
+ subi 0,$arrsz,$idx
+L\$copy_pa11
+ ldw 0($tp),$ti0
+ ldw 0($rp),$hi0
+ stws,ma %r0,4($tp)
+ comiclr,= 0,$hi1,%r0
+ copy $ti0,$hi0
+ addib,<> 4,$idx,L\$copy_pa11
+ stws,ma $hi0,4($rp)
+
+ nop ; alignment
+L\$done
+___
+}
+
+$code.=<<___;
+ ldi 1,%r28 ; signal "handled"
+ ldo $FRAME($fp),%sp ; destroy tp[num+1]
+
+ $POP `-$FRAME-$SAVED_RP`(%sp),%r2 ; standard epilogue
+ $POP `-$FRAME+1*$SIZE_T`(%sp),%r4
+ $POP `-$FRAME+2*$SIZE_T`(%sp),%r5
+ $POP `-$FRAME+3*$SIZE_T`(%sp),%r6
+ $POP `-$FRAME+4*$SIZE_T`(%sp),%r7
+ $POP `-$FRAME+5*$SIZE_T`(%sp),%r8
+ $POP `-$FRAME+6*$SIZE_T`(%sp),%r9
+ $POP `-$FRAME+7*$SIZE_T`(%sp),%r10
+L\$abort
+ bv (%r2)
+ .EXIT
+ $POPMB -$FRAME(%sp),%r3
+ .PROCEND
+ .STRINGZ "Montgomery Multiplication for PA-RISC, CRYPTOGAMS by <appro\@openssl.org>"
+___
+
+# Explicitly encode PA-RISC 2.0 instructions used in this module, so
+# that it can be compiled with .LEVEL 1.0. It should be noted that I
+# wouldn't have to do this, if GNU assembler understood .ALLOW 2.0
+# directive...
+
+my $ldd = sub {
+ my ($mod,$args) = @_;
+ my $orig = "ldd$mod\t$args";
+
+ if ($args =~ /%r([0-9]+)\(%r([0-9]+)\),%r([0-9]+)/) # format 4
+ { my $opcode=(0x03<<26)|($2<<21)|($1<<16)|(3<<6)|$3;
+ sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig;
+ }
+ elsif ($args =~ /(\-?[0-9]+)\(%r([0-9]+)\),%r([0-9]+)/) # format 5
+ { my $opcode=(0x03<<26)|($2<<21)|(1<<12)|(3<<6)|$3;
+ $opcode|=(($1&0xF)<<17)|(($1&0x10)<<12); # encode offset
+ $opcode|=(1<<5) if ($mod =~ /^,m/);
+ $opcode|=(1<<13) if ($mod =~ /^,mb/);
+ sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig;
+ }
+ else { "\t".$orig; }
+};
+
+my $std = sub {
+ my ($mod,$args) = @_;
+ my $orig = "std$mod\t$args";
+
+ if ($args =~ /%r([0-9]+),(\-?[0-9]+)\(%r([0-9]+)\)/) # format 6
+ { my $opcode=(0x03<<26)|($3<<21)|($1<<16)|(1<<12)|(0xB<<6);
+ $opcode|=(($2&0xF)<<1)|(($2&0x10)>>4); # encode offset
+ $opcode|=(1<<5) if ($mod =~ /^,m/);
+ $opcode|=(1<<13) if ($mod =~ /^,mb/);
+ sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig;
+ }
+ else { "\t".$orig; }
+};
+
+my $extrd = sub {
+ my ($mod,$args) = @_;
+ my $orig = "extrd$mod\t$args";
+
+ # I only have ",u" completer, it's implicitly encoded...
+ if ($args =~ /%r([0-9]+),([0-9]+),([0-9]+),%r([0-9]+)/) # format 15
+ { my $opcode=(0x36<<26)|($1<<21)|($4<<16);
+ my $len=32-$3;
+ $opcode |= (($2&0x20)<<6)|(($2&0x1f)<<5); # encode pos
+ $opcode |= (($len&0x20)<<7)|($len&0x1f); # encode len
+ sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig;
+ }
+ elsif ($args =~ /%r([0-9]+),%sar,([0-9]+),%r([0-9]+)/) # format 12
+ { my $opcode=(0x34<<26)|($1<<21)|($3<<16)|(2<<11)|(1<<9);
+ my $len=32-$2;
+ $opcode |= (($len&0x20)<<3)|($len&0x1f); # encode len
+ $opcode |= (1<<13) if ($mod =~ /,\**=/);
+ sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig;
+ }
+ else { "\t".$orig; }
+};
+
+my $shrpd = sub {
+ my ($mod,$args) = @_;
+ my $orig = "shrpd$mod\t$args";
+
+ if ($args =~ /%r([0-9]+),%r([0-9]+),([0-9]+),%r([0-9]+)/) # format 14
+ { my $opcode=(0x34<<26)|($2<<21)|($1<<16)|(1<<10)|$4;
+ my $cpos=63-$3;
+ $opcode |= (($cpos&0x20)<<6)|(($cpos&0x1f)<<5); # encode sa
+ sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig;
+ }
+ else { "\t".$orig; }
+};
+
+my $sub = sub {
+ my ($mod,$args) = @_;
+ my $orig = "sub$mod\t$args";
+
+ if ($mod eq ",db" && $args =~ /%r([0-9]+),%r([0-9]+),%r([0-9]+)/) {
+ my $opcode=(0x02<<26)|($2<<21)|($1<<16)|$3;
+ $opcode|=(1<<10); # e1
+ $opcode|=(1<<8); # e2
+ $opcode|=(1<<5); # d
+ sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig
+ }
+ else { "\t".$orig; }
+};
+
+sub assemble {
+ my ($mnemonic,$mod,$args)=@_;
+ my $opcode = eval("\$$mnemonic");
+
+ ref($opcode) eq 'CODE' ? &$opcode($mod,$args) : "\t$mnemonic$mod\t$args";
+}
+
+if (`$ENV{CC} -Wa,-v -c -o /dev/null -x assembler /dev/null 2>&1`
+ =~ /GNU assembler/) {
+ $gnuas = 1;
+}
+
+foreach (split("\n",$code)) {
+ s/\`([^\`]*)\`/eval $1/ge;
+ # flip word order in 64-bit mode...
+ s/(xmpyu\s+)($fai|$fni)([LR])/$1.$2.($3 eq "L"?"R":"L")/e if ($BN_SZ==8);
+ # assemble 2.0 instructions in 32-bit mode...
+ s/^\s+([a-z]+)([\S]*)\s+([\S]*)/&assemble($1,$2,$3)/e if ($BN_SZ==4);
+
+ s/(\.LEVEL\s+2\.0)W/$1w/ if ($gnuas && $SIZE_T==8);
+ s/\.SPACE\s+\$TEXT\$/.text/ if ($gnuas && $SIZE_T==8);
+ s/\.SUBSPA.*// if ($gnuas && $SIZE_T==8);
+ s/\bbv\b/bve/ if ($SIZE_T==8);
+
+ print $_,"\n";
+}
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/ppc-mont.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/ppc-mont.pl
new file mode 100644
index 0000000..278314c
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/ppc-mont.pl
@@ -0,0 +1,1990 @@
+#! /usr/bin/env perl
+# Copyright 2006-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+
+# April 2006
+
+# "Teaser" Montgomery multiplication module for PowerPC. It's possible
+# to gain a bit more by modulo-scheduling outer loop, then dedicated
+# squaring procedure should give further 20% and code can be adapted
+# for 32-bit application running on 64-bit CPU. As for the latter.
+# It won't be able to achieve "native" 64-bit performance, because in
+# 32-bit application context every addc instruction will have to be
+# expanded as addc, twice right shift by 32 and finally adde, etc.
+# So far RSA *sign* performance improvement over pre-bn_mul_mont asm
+# for 64-bit application running on PPC970/G5 is:
+#
+# 512-bit +65%
+# 1024-bit +35%
+# 2048-bit +18%
+# 4096-bit +4%
+
+# September 2016
+#
+# Add multiplication procedure operating on lengths divisible by 4
+# and squaring procedure operating on lengths divisible by 8. Length
+# is expressed in number of limbs. RSA private key operations are
+# ~35-50% faster (more for longer keys) on contemporary high-end POWER
+# processors in 64-bit builds, [mysteriously enough] more in 32-bit
+# builds. On low-end 32-bit processors performance improvement turned
+# to be marginal...
+
+$flavour = shift;
+
+if ($flavour =~ /32/) {
+ $BITS= 32;
+ $BNSZ= $BITS/8;
+ $SIZE_T=4;
+ $RZONE= 224;
+
+ $LD= "lwz"; # load
+ $LDU= "lwzu"; # load and update
+ $LDX= "lwzx"; # load indexed
+ $ST= "stw"; # store
+ $STU= "stwu"; # store and update
+ $STX= "stwx"; # store indexed
+ $STUX= "stwux"; # store indexed and update
+ $UMULL= "mullw"; # unsigned multiply low
+ $UMULH= "mulhwu"; # unsigned multiply high
+ $UCMP= "cmplw"; # unsigned compare
+ $SHRI= "srwi"; # unsigned shift right by immediate
+ $SHLI= "slwi"; # unsigned shift left by immediate
+ $PUSH= $ST;
+ $POP= $LD;
+} elsif ($flavour =~ /64/) {
+ $BITS= 64;
+ $BNSZ= $BITS/8;
+ $SIZE_T=8;
+ $RZONE= 288;
+
+ # same as above, but 64-bit mnemonics...
+ $LD= "ld"; # load
+ $LDU= "ldu"; # load and update
+ $LDX= "ldx"; # load indexed
+ $ST= "std"; # store
+ $STU= "stdu"; # store and update
+ $STX= "stdx"; # store indexed
+ $STUX= "stdux"; # store indexed and update
+ $UMULL= "mulld"; # unsigned multiply low
+ $UMULH= "mulhdu"; # unsigned multiply high
+ $UCMP= "cmpld"; # unsigned compare
+ $SHRI= "srdi"; # unsigned shift right by immediate
+ $SHLI= "sldi"; # unsigned shift left by immediate
+ $PUSH= $ST;
+ $POP= $LD;
+} else { die "nonsense $flavour"; }
+
+$FRAME=8*$SIZE_T+$RZONE;
+$LOCALS=8*$SIZE_T;
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or
+( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or
+die "can't locate ppc-xlate.pl";
+
+open STDOUT,"| $^X $xlate $flavour ".shift || die "can't call $xlate: $!";
+
+$sp="r1";
+$toc="r2";
+$rp="r3";
+$ap="r4";
+$bp="r5";
+$np="r6";
+$n0="r7";
+$num="r8";
+
+{
+my $ovf=$rp;
+my $rp="r9"; # $rp is reassigned
+my $aj="r10";
+my $nj="r11";
+my $tj="r12";
+# non-volatile registers
+my $i="r20";
+my $j="r21";
+my $tp="r22";
+my $m0="r23";
+my $m1="r24";
+my $lo0="r25";
+my $hi0="r26";
+my $lo1="r27";
+my $hi1="r28";
+my $alo="r29";
+my $ahi="r30";
+my $nlo="r31";
+#
+my $nhi="r0";
+
+$code=<<___;
+.machine "any"
+.text
+
+.globl .bn_mul_mont_int
+.align 5
+.bn_mul_mont_int:
+ mr $rp,r3 ; $rp is reassigned
+ li r3,0
+___
+$code.=<<___ if ($BNSZ==4);
+ cmpwi $num,32 ; longer key performance is not better
+ bgelr
+___
+$code.=<<___;
+ slwi $num,$num,`log($BNSZ)/log(2)`
+ li $tj,-4096
+ addi $ovf,$num,$FRAME
+ subf $ovf,$ovf,$sp ; $sp-$ovf
+ and $ovf,$ovf,$tj ; minimize TLB usage
+ subf $ovf,$sp,$ovf ; $ovf-$sp
+ mr $tj,$sp
+ srwi $num,$num,`log($BNSZ)/log(2)`
+ $STUX $sp,$sp,$ovf
+
+ $PUSH r20,`-12*$SIZE_T`($tj)
+ $PUSH r21,`-11*$SIZE_T`($tj)
+ $PUSH r22,`-10*$SIZE_T`($tj)
+ $PUSH r23,`-9*$SIZE_T`($tj)
+ $PUSH r24,`-8*$SIZE_T`($tj)
+ $PUSH r25,`-7*$SIZE_T`($tj)
+ $PUSH r26,`-6*$SIZE_T`($tj)
+ $PUSH r27,`-5*$SIZE_T`($tj)
+ $PUSH r28,`-4*$SIZE_T`($tj)
+ $PUSH r29,`-3*$SIZE_T`($tj)
+ $PUSH r30,`-2*$SIZE_T`($tj)
+ $PUSH r31,`-1*$SIZE_T`($tj)
+
+ $LD $n0,0($n0) ; pull n0[0] value
+ addi $num,$num,-2 ; adjust $num for counter register
+
+ $LD $m0,0($bp) ; m0=bp[0]
+ $LD $aj,0($ap) ; ap[0]
+ addi $tp,$sp,$LOCALS
+ $UMULL $lo0,$aj,$m0 ; ap[0]*bp[0]
+ $UMULH $hi0,$aj,$m0
+
+ $LD $aj,$BNSZ($ap) ; ap[1]
+ $LD $nj,0($np) ; np[0]
+
+ $UMULL $m1,$lo0,$n0 ; "tp[0]"*n0
+
+ $UMULL $alo,$aj,$m0 ; ap[1]*bp[0]
+ $UMULH $ahi,$aj,$m0
+
+ $UMULL $lo1,$nj,$m1 ; np[0]*m1
+ $UMULH $hi1,$nj,$m1
+ $LD $nj,$BNSZ($np) ; np[1]
+ addc $lo1,$lo1,$lo0
+ addze $hi1,$hi1
+
+ $UMULL $nlo,$nj,$m1 ; np[1]*m1
+ $UMULH $nhi,$nj,$m1
+
+ mtctr $num
+ li $j,`2*$BNSZ`
+.align 4
+L1st:
+ $LDX $aj,$ap,$j ; ap[j]
+ addc $lo0,$alo,$hi0
+ $LDX $nj,$np,$j ; np[j]
+ addze $hi0,$ahi
+ $UMULL $alo,$aj,$m0 ; ap[j]*bp[0]
+ addc $lo1,$nlo,$hi1
+ $UMULH $ahi,$aj,$m0
+ addze $hi1,$nhi
+ $UMULL $nlo,$nj,$m1 ; np[j]*m1
+ addc $lo1,$lo1,$lo0 ; np[j]*m1+ap[j]*bp[0]
+ $UMULH $nhi,$nj,$m1
+ addze $hi1,$hi1
+ $ST $lo1,0($tp) ; tp[j-1]
+
+ addi $j,$j,$BNSZ ; j++
+ addi $tp,$tp,$BNSZ ; tp++
+ bdnz L1st
+;L1st
+ addc $lo0,$alo,$hi0
+ addze $hi0,$ahi
+
+ addc $lo1,$nlo,$hi1
+ addze $hi1,$nhi
+ addc $lo1,$lo1,$lo0 ; np[j]*m1+ap[j]*bp[0]
+ addze $hi1,$hi1
+ $ST $lo1,0($tp) ; tp[j-1]
+
+ li $ovf,0
+ addc $hi1,$hi1,$hi0
+ addze $ovf,$ovf ; upmost overflow bit
+ $ST $hi1,$BNSZ($tp)
+
+ li $i,$BNSZ
+.align 4
+Louter:
+ $LDX $m0,$bp,$i ; m0=bp[i]
+ $LD $aj,0($ap) ; ap[0]
+ addi $tp,$sp,$LOCALS
+ $LD $tj,$LOCALS($sp); tp[0]
+ $UMULL $lo0,$aj,$m0 ; ap[0]*bp[i]
+ $UMULH $hi0,$aj,$m0
+ $LD $aj,$BNSZ($ap) ; ap[1]
+ $LD $nj,0($np) ; np[0]
+ addc $lo0,$lo0,$tj ; ap[0]*bp[i]+tp[0]
+ $UMULL $alo,$aj,$m0 ; ap[j]*bp[i]
+ addze $hi0,$hi0
+ $UMULL $m1,$lo0,$n0 ; tp[0]*n0
+ $UMULH $ahi,$aj,$m0
+ $UMULL $lo1,$nj,$m1 ; np[0]*m1
+ $UMULH $hi1,$nj,$m1
+ $LD $nj,$BNSZ($np) ; np[1]
+ addc $lo1,$lo1,$lo0
+ $UMULL $nlo,$nj,$m1 ; np[1]*m1
+ addze $hi1,$hi1
+ $UMULH $nhi,$nj,$m1
+
+ mtctr $num
+ li $j,`2*$BNSZ`
+.align 4
+Linner:
+ $LDX $aj,$ap,$j ; ap[j]
+ addc $lo0,$alo,$hi0
+ $LD $tj,$BNSZ($tp) ; tp[j]
+ addze $hi0,$ahi
+ $LDX $nj,$np,$j ; np[j]
+ addc $lo1,$nlo,$hi1
+ $UMULL $alo,$aj,$m0 ; ap[j]*bp[i]
+ addze $hi1,$nhi
+ $UMULH $ahi,$aj,$m0
+ addc $lo0,$lo0,$tj ; ap[j]*bp[i]+tp[j]
+ $UMULL $nlo,$nj,$m1 ; np[j]*m1
+ addze $hi0,$hi0
+ $UMULH $nhi,$nj,$m1
+ addc $lo1,$lo1,$lo0 ; np[j]*m1+ap[j]*bp[i]+tp[j]
+ addi $j,$j,$BNSZ ; j++
+ addze $hi1,$hi1
+ $ST $lo1,0($tp) ; tp[j-1]
+ addi $tp,$tp,$BNSZ ; tp++
+ bdnz Linner
+;Linner
+ $LD $tj,$BNSZ($tp) ; tp[j]
+ addc $lo0,$alo,$hi0
+ addze $hi0,$ahi
+ addc $lo0,$lo0,$tj ; ap[j]*bp[i]+tp[j]
+ addze $hi0,$hi0
+
+ addc $lo1,$nlo,$hi1
+ addze $hi1,$nhi
+ addc $lo1,$lo1,$lo0 ; np[j]*m1+ap[j]*bp[i]+tp[j]
+ addze $hi1,$hi1
+ $ST $lo1,0($tp) ; tp[j-1]
+
+ addic $ovf,$ovf,-1 ; move upmost overflow to XER[CA]
+ li $ovf,0
+ adde $hi1,$hi1,$hi0
+ addze $ovf,$ovf
+ $ST $hi1,$BNSZ($tp)
+;
+ slwi $tj,$num,`log($BNSZ)/log(2)`
+ $UCMP $i,$tj
+ addi $i,$i,$BNSZ
+ ble Louter
+
+ addi $num,$num,2 ; restore $num
+ subfc $j,$j,$j ; j=0 and "clear" XER[CA]
+ addi $tp,$sp,$LOCALS
+ mtctr $num
+
+.align 4
+Lsub: $LDX $tj,$tp,$j
+ $LDX $nj,$np,$j
+ subfe $aj,$nj,$tj ; tp[j]-np[j]
+ $STX $aj,$rp,$j
+ addi $j,$j,$BNSZ
+ bdnz Lsub
+
+ li $j,0
+ mtctr $num
+ subfe $ovf,$j,$ovf ; handle upmost overflow bit
+
+.align 4
+Lcopy: ; conditional copy
+ $LDX $tj,$tp,$j
+ $LDX $aj,$rp,$j
+ and $tj,$tj,$ovf
+ andc $aj,$aj,$ovf
+ $STX $j,$tp,$j ; zap at once
+ or $aj,$aj,$tj
+ $STX $aj,$rp,$j
+ addi $j,$j,$BNSZ
+ bdnz Lcopy
+
+ $POP $tj,0($sp)
+ li r3,1
+ $POP r20,`-12*$SIZE_T`($tj)
+ $POP r21,`-11*$SIZE_T`($tj)
+ $POP r22,`-10*$SIZE_T`($tj)
+ $POP r23,`-9*$SIZE_T`($tj)
+ $POP r24,`-8*$SIZE_T`($tj)
+ $POP r25,`-7*$SIZE_T`($tj)
+ $POP r26,`-6*$SIZE_T`($tj)
+ $POP r27,`-5*$SIZE_T`($tj)
+ $POP r28,`-4*$SIZE_T`($tj)
+ $POP r29,`-3*$SIZE_T`($tj)
+ $POP r30,`-2*$SIZE_T`($tj)
+ $POP r31,`-1*$SIZE_T`($tj)
+ mr $sp,$tj
+ blr
+ .long 0
+ .byte 0,12,4,0,0x80,12,6,0
+ .long 0
+.size .bn_mul_mont_int,.-.bn_mul_mont_int
+___
+}
+if (1) {
+my ($a0,$a1,$a2,$a3,
+ $t0,$t1,$t2,$t3,
+ $m0,$m1,$m2,$m3,
+ $acc0,$acc1,$acc2,$acc3,$acc4,
+ $bi,$mi,$tp,$ap_end,$cnt) = map("r$_",(9..12,14..31));
+my ($carry,$zero) = ($rp,"r0");
+
+# sp----------->+-------------------------------+
+# | saved sp |
+# +-------------------------------+
+# . .
+# +8*size_t +-------------------------------+
+# | 4 "n0*t0" |
+# . .
+# . .
+# +12*size_t +-------------------------------+
+# | size_t tmp[num] |
+# . .
+# . .
+# . .
+# +-------------------------------+
+# | topmost carry |
+# . .
+# -18*size_t +-------------------------------+
+# | 18 saved gpr, r14-r31 |
+# . .
+# . .
+# +-------------------------------+
+$code.=<<___;
+.globl .bn_mul4x_mont_int
+.align 5
+.bn_mul4x_mont_int:
+ andi. r0,$num,7
+ bne .Lmul4x_do
+ $UCMP $ap,$bp
+ bne .Lmul4x_do
+ b .Lsqr8x_do
+.Lmul4x_do:
+ slwi $num,$num,`log($SIZE_T)/log(2)`
+ mr $a0,$sp
+ li $a1,-32*$SIZE_T
+ sub $a1,$a1,$num
+ $STUX $sp,$sp,$a1 # alloca
+
+ $PUSH r14,-$SIZE_T*18($a0)
+ $PUSH r15,-$SIZE_T*17($a0)
+ $PUSH r16,-$SIZE_T*16($a0)
+ $PUSH r17,-$SIZE_T*15($a0)
+ $PUSH r18,-$SIZE_T*14($a0)
+ $PUSH r19,-$SIZE_T*13($a0)
+ $PUSH r20,-$SIZE_T*12($a0)
+ $PUSH r21,-$SIZE_T*11($a0)
+ $PUSH r22,-$SIZE_T*10($a0)
+ $PUSH r23,-$SIZE_T*9($a0)
+ $PUSH r24,-$SIZE_T*8($a0)
+ $PUSH r25,-$SIZE_T*7($a0)
+ $PUSH r26,-$SIZE_T*6($a0)
+ $PUSH r27,-$SIZE_T*5($a0)
+ $PUSH r28,-$SIZE_T*4($a0)
+ $PUSH r29,-$SIZE_T*3($a0)
+ $PUSH r30,-$SIZE_T*2($a0)
+ $PUSH r31,-$SIZE_T*1($a0)
+
+ subi $ap,$ap,$SIZE_T # bias by -1
+ subi $np,$np,$SIZE_T # bias by -1
+ subi $rp,$rp,$SIZE_T # bias by -1
+ $LD $n0,0($n0) # *n0
+
+ add $t0,$bp,$num
+ add $ap_end,$ap,$num
+ subi $t0,$t0,$SIZE_T*4 # &b[num-4]
+
+ $LD $bi,$SIZE_T*0($bp) # b[0]
+ li $acc0,0
+ $LD $a0,$SIZE_T*1($ap) # a[0..3]
+ li $acc1,0
+ $LD $a1,$SIZE_T*2($ap)
+ li $acc2,0
+ $LD $a2,$SIZE_T*3($ap)
+ li $acc3,0
+ $LDU $a3,$SIZE_T*4($ap)
+ $LD $m0,$SIZE_T*1($np) # n[0..3]
+ $LD $m1,$SIZE_T*2($np)
+ $LD $m2,$SIZE_T*3($np)
+ $LDU $m3,$SIZE_T*4($np)
+
+ $PUSH $rp,$SIZE_T*6($sp) # offload rp and &b[num-4]
+ $PUSH $t0,$SIZE_T*7($sp)
+ li $carry,0
+ addic $tp,$sp,$SIZE_T*7 # &t[-1], clear carry bit
+ li $cnt,0
+ li $zero,0
+ b .Loop_mul4x_1st_reduction
+
+.align 5
+.Loop_mul4x_1st_reduction:
+ $UMULL $t0,$a0,$bi # lo(a[0..3]*b[0])
+ addze $carry,$carry # modulo-scheduled
+ $UMULL $t1,$a1,$bi
+ addi $cnt,$cnt,$SIZE_T
+ $UMULL $t2,$a2,$bi
+ andi. $cnt,$cnt,$SIZE_T*4-1
+ $UMULL $t3,$a3,$bi
+ addc $acc0,$acc0,$t0
+ $UMULH $t0,$a0,$bi # hi(a[0..3]*b[0])
+ adde $acc1,$acc1,$t1
+ $UMULH $t1,$a1,$bi
+ adde $acc2,$acc2,$t2
+ $UMULL $mi,$acc0,$n0 # t[0]*n0
+ adde $acc3,$acc3,$t3
+ $UMULH $t2,$a2,$bi
+ addze $acc4,$zero
+ $UMULH $t3,$a3,$bi
+ $LDX $bi,$bp,$cnt # next b[i] (or b[0])
+ addc $acc1,$acc1,$t0
+ # (*) mul $t0,$m0,$mi # lo(n[0..3]*t[0]*n0)
+ $STU $mi,$SIZE_T($tp) # put aside t[0]*n0 for tail processing
+ adde $acc2,$acc2,$t1
+ $UMULL $t1,$m1,$mi
+ adde $acc3,$acc3,$t2
+ $UMULL $t2,$m2,$mi
+ adde $acc4,$acc4,$t3 # can't overflow
+ $UMULL $t3,$m3,$mi
+ # (*) addc $acc0,$acc0,$t0
+ # (*) As for removal of first multiplication and addition
+ # instructions. The outcome of first addition is
+ # guaranteed to be zero, which leaves two computationally
+ # significant outcomes: it either carries or not. Then
+ # question is when does it carry? Is there alternative
+ # way to deduce it? If you follow operations, you can
+ # observe that condition for carry is quite simple:
+ # $acc0 being non-zero. So that carry can be calculated
+ # by adding -1 to $acc0. That's what next instruction does.
+ addic $acc0,$acc0,-1 # (*), discarded
+ $UMULH $t0,$m0,$mi # hi(n[0..3]*t[0]*n0)
+ adde $acc0,$acc1,$t1
+ $UMULH $t1,$m1,$mi
+ adde $acc1,$acc2,$t2
+ $UMULH $t2,$m2,$mi
+ adde $acc2,$acc3,$t3
+ $UMULH $t3,$m3,$mi
+ adde $acc3,$acc4,$carry
+ addze $carry,$zero
+ addc $acc0,$acc0,$t0
+ adde $acc1,$acc1,$t1
+ adde $acc2,$acc2,$t2
+ adde $acc3,$acc3,$t3
+ #addze $carry,$carry
+ bne .Loop_mul4x_1st_reduction
+
+ $UCMP $ap_end,$ap
+ beq .Lmul4x4_post_condition
+
+ $LD $a0,$SIZE_T*1($ap) # a[4..7]
+ $LD $a1,$SIZE_T*2($ap)
+ $LD $a2,$SIZE_T*3($ap)
+ $LDU $a3,$SIZE_T*4($ap)
+ $LD $mi,$SIZE_T*8($sp) # a[0]*n0
+ $LD $m0,$SIZE_T*1($np) # n[4..7]
+ $LD $m1,$SIZE_T*2($np)
+ $LD $m2,$SIZE_T*3($np)
+ $LDU $m3,$SIZE_T*4($np)
+ b .Loop_mul4x_1st_tail
+
+.align 5
+.Loop_mul4x_1st_tail:
+ $UMULL $t0,$a0,$bi # lo(a[4..7]*b[i])
+ addze $carry,$carry # modulo-scheduled
+ $UMULL $t1,$a1,$bi
+ addi $cnt,$cnt,$SIZE_T
+ $UMULL $t2,$a2,$bi
+ andi. $cnt,$cnt,$SIZE_T*4-1
+ $UMULL $t3,$a3,$bi
+ addc $acc0,$acc0,$t0
+ $UMULH $t0,$a0,$bi # hi(a[4..7]*b[i])
+ adde $acc1,$acc1,$t1
+ $UMULH $t1,$a1,$bi
+ adde $acc2,$acc2,$t2
+ $UMULH $t2,$a2,$bi
+ adde $acc3,$acc3,$t3
+ $UMULH $t3,$a3,$bi
+ addze $acc4,$zero
+ $LDX $bi,$bp,$cnt # next b[i] (or b[0])
+ addc $acc1,$acc1,$t0
+ $UMULL $t0,$m0,$mi # lo(n[4..7]*a[0]*n0)
+ adde $acc2,$acc2,$t1
+ $UMULL $t1,$m1,$mi
+ adde $acc3,$acc3,$t2
+ $UMULL $t2,$m2,$mi
+ adde $acc4,$acc4,$t3 # can't overflow
+ $UMULL $t3,$m3,$mi
+ addc $acc0,$acc0,$t0
+ $UMULH $t0,$m0,$mi # hi(n[4..7]*a[0]*n0)
+ adde $acc1,$acc1,$t1
+ $UMULH $t1,$m1,$mi
+ adde $acc2,$acc2,$t2
+ $UMULH $t2,$m2,$mi
+ adde $acc3,$acc3,$t3
+ adde $acc4,$acc4,$carry
+ $UMULH $t3,$m3,$mi
+ addze $carry,$zero
+ addi $mi,$sp,$SIZE_T*8
+ $LDX $mi,$mi,$cnt # next t[0]*n0
+ $STU $acc0,$SIZE_T($tp) # word of result
+ addc $acc0,$acc1,$t0
+ adde $acc1,$acc2,$t1
+ adde $acc2,$acc3,$t2
+ adde $acc3,$acc4,$t3
+ #addze $carry,$carry
+ bne .Loop_mul4x_1st_tail
+
+ sub $t1,$ap_end,$num # rewinded $ap
+ $UCMP $ap_end,$ap # done yet?
+ beq .Lmul4x_proceed
+
+ $LD $a0,$SIZE_T*1($ap)
+ $LD $a1,$SIZE_T*2($ap)
+ $LD $a2,$SIZE_T*3($ap)
+ $LDU $a3,$SIZE_T*4($ap)
+ $LD $m0,$SIZE_T*1($np)
+ $LD $m1,$SIZE_T*2($np)
+ $LD $m2,$SIZE_T*3($np)
+ $LDU $m3,$SIZE_T*4($np)
+ b .Loop_mul4x_1st_tail
+
+.align 5
+.Lmul4x_proceed:
+ $LDU $bi,$SIZE_T*4($bp) # *++b
+ addze $carry,$carry # topmost carry
+ $LD $a0,$SIZE_T*1($t1)
+ $LD $a1,$SIZE_T*2($t1)
+ $LD $a2,$SIZE_T*3($t1)
+ $LD $a3,$SIZE_T*4($t1)
+ addi $ap,$t1,$SIZE_T*4
+ sub $np,$np,$num # rewind np
+
+ $ST $acc0,$SIZE_T*1($tp) # result
+ $ST $acc1,$SIZE_T*2($tp)
+ $ST $acc2,$SIZE_T*3($tp)
+ $ST $acc3,$SIZE_T*4($tp)
+ $ST $carry,$SIZE_T*5($tp) # save topmost carry
+ $LD $acc0,$SIZE_T*12($sp) # t[0..3]
+ $LD $acc1,$SIZE_T*13($sp)
+ $LD $acc2,$SIZE_T*14($sp)
+ $LD $acc3,$SIZE_T*15($sp)
+
+ $LD $m0,$SIZE_T*1($np) # n[0..3]
+ $LD $m1,$SIZE_T*2($np)
+ $LD $m2,$SIZE_T*3($np)
+ $LDU $m3,$SIZE_T*4($np)
+ addic $tp,$sp,$SIZE_T*7 # &t[-1], clear carry bit
+ li $carry,0
+ b .Loop_mul4x_reduction
+
+.align 5
+.Loop_mul4x_reduction:
+ $UMULL $t0,$a0,$bi # lo(a[0..3]*b[4])
+ addze $carry,$carry # modulo-scheduled
+ $UMULL $t1,$a1,$bi
+ addi $cnt,$cnt,$SIZE_T
+ $UMULL $t2,$a2,$bi
+ andi. $cnt,$cnt,$SIZE_T*4-1
+ $UMULL $t3,$a3,$bi
+ addc $acc0,$acc0,$t0
+ $UMULH $t0,$a0,$bi # hi(a[0..3]*b[4])
+ adde $acc1,$acc1,$t1
+ $UMULH $t1,$a1,$bi
+ adde $acc2,$acc2,$t2
+ $UMULL $mi,$acc0,$n0 # t[0]*n0
+ adde $acc3,$acc3,$t3
+ $UMULH $t2,$a2,$bi
+ addze $acc4,$zero
+ $UMULH $t3,$a3,$bi
+ $LDX $bi,$bp,$cnt # next b[i]
+ addc $acc1,$acc1,$t0
+ # (*) mul $t0,$m0,$mi
+ $STU $mi,$SIZE_T($tp) # put aside t[0]*n0 for tail processing
+ adde $acc2,$acc2,$t1
+ $UMULL $t1,$m1,$mi # lo(n[0..3]*t[0]*n0
+ adde $acc3,$acc3,$t2
+ $UMULL $t2,$m2,$mi
+ adde $acc4,$acc4,$t3 # can't overflow
+ $UMULL $t3,$m3,$mi
+ # (*) addc $acc0,$acc0,$t0
+ addic $acc0,$acc0,-1 # (*), discarded
+ $UMULH $t0,$m0,$mi # hi(n[0..3]*t[0]*n0
+ adde $acc0,$acc1,$t1
+ $UMULH $t1,$m1,$mi
+ adde $acc1,$acc2,$t2
+ $UMULH $t2,$m2,$mi
+ adde $acc2,$acc3,$t3
+ $UMULH $t3,$m3,$mi
+ adde $acc3,$acc4,$carry
+ addze $carry,$zero
+ addc $acc0,$acc0,$t0
+ adde $acc1,$acc1,$t1
+ adde $acc2,$acc2,$t2
+ adde $acc3,$acc3,$t3
+ #addze $carry,$carry
+ bne .Loop_mul4x_reduction
+
+ $LD $t0,$SIZE_T*5($tp) # t[4..7]
+ addze $carry,$carry
+ $LD $t1,$SIZE_T*6($tp)
+ $LD $t2,$SIZE_T*7($tp)
+ $LD $t3,$SIZE_T*8($tp)
+ $LD $a0,$SIZE_T*1($ap) # a[4..7]
+ $LD $a1,$SIZE_T*2($ap)
+ $LD $a2,$SIZE_T*3($ap)
+ $LDU $a3,$SIZE_T*4($ap)
+ addc $acc0,$acc0,$t0
+ adde $acc1,$acc1,$t1
+ adde $acc2,$acc2,$t2
+ adde $acc3,$acc3,$t3
+ #addze $carry,$carry
+
+ $LD $mi,$SIZE_T*8($sp) # t[0]*n0
+ $LD $m0,$SIZE_T*1($np) # n[4..7]
+ $LD $m1,$SIZE_T*2($np)
+ $LD $m2,$SIZE_T*3($np)
+ $LDU $m3,$SIZE_T*4($np)
+ b .Loop_mul4x_tail
+
+.align 5
+.Loop_mul4x_tail:
+ $UMULL $t0,$a0,$bi # lo(a[4..7]*b[4])
+ addze $carry,$carry # modulo-scheduled
+ $UMULL $t1,$a1,$bi
+ addi $cnt,$cnt,$SIZE_T
+ $UMULL $t2,$a2,$bi
+ andi. $cnt,$cnt,$SIZE_T*4-1
+ $UMULL $t3,$a3,$bi
+ addc $acc0,$acc0,$t0
+ $UMULH $t0,$a0,$bi # hi(a[4..7]*b[4])
+ adde $acc1,$acc1,$t1
+ $UMULH $t1,$a1,$bi
+ adde $acc2,$acc2,$t2
+ $UMULH $t2,$a2,$bi
+ adde $acc3,$acc3,$t3
+ $UMULH $t3,$a3,$bi
+ addze $acc4,$zero
+ $LDX $bi,$bp,$cnt # next b[i]
+ addc $acc1,$acc1,$t0
+ $UMULL $t0,$m0,$mi # lo(n[4..7]*t[0]*n0)
+ adde $acc2,$acc2,$t1
+ $UMULL $t1,$m1,$mi
+ adde $acc3,$acc3,$t2
+ $UMULL $t2,$m2,$mi
+ adde $acc4,$acc4,$t3 # can't overflow
+ $UMULL $t3,$m3,$mi
+ addc $acc0,$acc0,$t0
+ $UMULH $t0,$m0,$mi # hi(n[4..7]*t[0]*n0)
+ adde $acc1,$acc1,$t1
+ $UMULH $t1,$m1,$mi
+ adde $acc2,$acc2,$t2
+ $UMULH $t2,$m2,$mi
+ adde $acc3,$acc3,$t3
+ $UMULH $t3,$m3,$mi
+ adde $acc4,$acc4,$carry
+ addi $mi,$sp,$SIZE_T*8
+ $LDX $mi,$mi,$cnt # next a[0]*n0
+ addze $carry,$zero
+ $STU $acc0,$SIZE_T($tp) # word of result
+ addc $acc0,$acc1,$t0
+ adde $acc1,$acc2,$t1
+ adde $acc2,$acc3,$t2
+ adde $acc3,$acc4,$t3
+ #addze $carry,$carry
+ bne .Loop_mul4x_tail
+
+ $LD $t0,$SIZE_T*5($tp) # next t[i] or topmost carry
+ sub $t1,$np,$num # rewinded np?
+ addze $carry,$carry
+ $UCMP $ap_end,$ap # done yet?
+ beq .Loop_mul4x_break
+
+ $LD $t1,$SIZE_T*6($tp)
+ $LD $t2,$SIZE_T*7($tp)
+ $LD $t3,$SIZE_T*8($tp)
+ $LD $a0,$SIZE_T*1($ap)
+ $LD $a1,$SIZE_T*2($ap)
+ $LD $a2,$SIZE_T*3($ap)
+ $LDU $a3,$SIZE_T*4($ap)
+ addc $acc0,$acc0,$t0
+ adde $acc1,$acc1,$t1
+ adde $acc2,$acc2,$t2
+ adde $acc3,$acc3,$t3
+ #addze $carry,$carry
+
+ $LD $m0,$SIZE_T*1($np) # n[4..7]
+ $LD $m1,$SIZE_T*2($np)
+ $LD $m2,$SIZE_T*3($np)
+ $LDU $m3,$SIZE_T*4($np)
+ b .Loop_mul4x_tail
+
+.align 5
+.Loop_mul4x_break:
+ $POP $t2,$SIZE_T*6($sp) # pull rp and &b[num-4]
+ $POP $t3,$SIZE_T*7($sp)
+ addc $a0,$acc0,$t0 # accumulate topmost carry
+ $LD $acc0,$SIZE_T*12($sp) # t[0..3]
+ addze $a1,$acc1
+ $LD $acc1,$SIZE_T*13($sp)
+ addze $a2,$acc2
+ $LD $acc2,$SIZE_T*14($sp)
+ addze $a3,$acc3
+ $LD $acc3,$SIZE_T*15($sp)
+ addze $carry,$carry # topmost carry
+ $ST $a0,$SIZE_T*1($tp) # result
+ sub $ap,$ap_end,$num # rewind ap
+ $ST $a1,$SIZE_T*2($tp)
+ $ST $a2,$SIZE_T*3($tp)
+ $ST $a3,$SIZE_T*4($tp)
+ $ST $carry,$SIZE_T*5($tp) # store topmost carry
+
+ $LD $m0,$SIZE_T*1($t1) # n[0..3]
+ $LD $m1,$SIZE_T*2($t1)
+ $LD $m2,$SIZE_T*3($t1)
+ $LD $m3,$SIZE_T*4($t1)
+ addi $np,$t1,$SIZE_T*4
+ $UCMP $bp,$t3 # done yet?
+ beq .Lmul4x_post
+
+ $LDU $bi,$SIZE_T*4($bp)
+ $LD $a0,$SIZE_T*1($ap) # a[0..3]
+ $LD $a1,$SIZE_T*2($ap)
+ $LD $a2,$SIZE_T*3($ap)
+ $LDU $a3,$SIZE_T*4($ap)
+ li $carry,0
+ addic $tp,$sp,$SIZE_T*7 # &t[-1], clear carry bit
+ b .Loop_mul4x_reduction
+
+.align 5
+.Lmul4x_post:
+ # Final step. We see if result is larger than modulus, and
+ # if it is, subtract the modulus. But comparison implies
+ # subtraction. So we subtract modulus, see if it borrowed,
+ # and conditionally copy original value.
+ srwi $cnt,$num,`log($SIZE_T)/log(2)+2`
+ mr $bp,$t2 # &rp[-1]
+ subi $cnt,$cnt,1
+ mr $ap_end,$t2 # &rp[-1] copy
+ subfc $t0,$m0,$acc0
+ addi $tp,$sp,$SIZE_T*15
+ subfe $t1,$m1,$acc1
+
+ mtctr $cnt
+.Lmul4x_sub:
+ $LD $m0,$SIZE_T*1($np)
+ $LD $acc0,$SIZE_T*1($tp)
+ subfe $t2,$m2,$acc2
+ $LD $m1,$SIZE_T*2($np)
+ $LD $acc1,$SIZE_T*2($tp)
+ subfe $t3,$m3,$acc3
+ $LD $m2,$SIZE_T*3($np)
+ $LD $acc2,$SIZE_T*3($tp)
+ $LDU $m3,$SIZE_T*4($np)
+ $LDU $acc3,$SIZE_T*4($tp)
+ $ST $t0,$SIZE_T*1($bp)
+ $ST $t1,$SIZE_T*2($bp)
+ subfe $t0,$m0,$acc0
+ $ST $t2,$SIZE_T*3($bp)
+ $STU $t3,$SIZE_T*4($bp)
+ subfe $t1,$m1,$acc1
+ bdnz .Lmul4x_sub
+
+ $LD $a0,$SIZE_T*1($ap_end)
+ $ST $t0,$SIZE_T*1($bp)
+ $LD $t0,$SIZE_T*12($sp)
+ subfe $t2,$m2,$acc2
+ $LD $a1,$SIZE_T*2($ap_end)
+ $ST $t1,$SIZE_T*2($bp)
+ $LD $t1,$SIZE_T*13($sp)
+ subfe $t3,$m3,$acc3
+ subfe $carry,$zero,$carry # did it borrow?
+ addi $tp,$sp,$SIZE_T*12
+ $LD $a2,$SIZE_T*3($ap_end)
+ $ST $t2,$SIZE_T*3($bp)
+ $LD $t2,$SIZE_T*14($sp)
+ $LD $a3,$SIZE_T*4($ap_end)
+ $ST $t3,$SIZE_T*4($bp)
+ $LD $t3,$SIZE_T*15($sp)
+
+ mtctr $cnt
+.Lmul4x_cond_copy:
+ and $t0,$t0,$carry
+ andc $a0,$a0,$carry
+ $ST $zero,$SIZE_T*0($tp) # wipe stack clean
+ and $t1,$t1,$carry
+ andc $a1,$a1,$carry
+ $ST $zero,$SIZE_T*1($tp)
+ and $t2,$t2,$carry
+ andc $a2,$a2,$carry
+ $ST $zero,$SIZE_T*2($tp)
+ and $t3,$t3,$carry
+ andc $a3,$a3,$carry
+ $ST $zero,$SIZE_T*3($tp)
+ or $acc0,$t0,$a0
+ $LD $a0,$SIZE_T*5($ap_end)
+ $LD $t0,$SIZE_T*4($tp)
+ or $acc1,$t1,$a1
+ $LD $a1,$SIZE_T*6($ap_end)
+ $LD $t1,$SIZE_T*5($tp)
+ or $acc2,$t2,$a2
+ $LD $a2,$SIZE_T*7($ap_end)
+ $LD $t2,$SIZE_T*6($tp)
+ or $acc3,$t3,$a3
+ $LD $a3,$SIZE_T*8($ap_end)
+ $LD $t3,$SIZE_T*7($tp)
+ addi $tp,$tp,$SIZE_T*4
+ $ST $acc0,$SIZE_T*1($ap_end)
+ $ST $acc1,$SIZE_T*2($ap_end)
+ $ST $acc2,$SIZE_T*3($ap_end)
+ $STU $acc3,$SIZE_T*4($ap_end)
+ bdnz .Lmul4x_cond_copy
+
+ $POP $bp,0($sp) # pull saved sp
+ and $t0,$t0,$carry
+ andc $a0,$a0,$carry
+ $ST $zero,$SIZE_T*0($tp)
+ and $t1,$t1,$carry
+ andc $a1,$a1,$carry
+ $ST $zero,$SIZE_T*1($tp)
+ and $t2,$t2,$carry
+ andc $a2,$a2,$carry
+ $ST $zero,$SIZE_T*2($tp)
+ and $t3,$t3,$carry
+ andc $a3,$a3,$carry
+ $ST $zero,$SIZE_T*3($tp)
+ or $acc0,$t0,$a0
+ or $acc1,$t1,$a1
+ $ST $zero,$SIZE_T*4($tp)
+ or $acc2,$t2,$a2
+ or $acc3,$t3,$a3
+ $ST $acc0,$SIZE_T*1($ap_end)
+ $ST $acc1,$SIZE_T*2($ap_end)
+ $ST $acc2,$SIZE_T*3($ap_end)
+ $ST $acc3,$SIZE_T*4($ap_end)
+
+ b .Lmul4x_done
+
+.align 4
+.Lmul4x4_post_condition:
+ $POP $ap,$SIZE_T*6($sp) # pull &rp[-1]
+ $POP $bp,0($sp) # pull saved sp
+ addze $carry,$carry # modulo-scheduled
+ # $acc0-3,$carry hold result, $m0-3 hold modulus
+ subfc $a0,$m0,$acc0
+ subfe $a1,$m1,$acc1
+ subfe $a2,$m2,$acc2
+ subfe $a3,$m3,$acc3
+ subfe $carry,$zero,$carry # did it borrow?
+
+ and $m0,$m0,$carry
+ and $m1,$m1,$carry
+ addc $a0,$a0,$m0
+ and $m2,$m2,$carry
+ adde $a1,$a1,$m1
+ and $m3,$m3,$carry
+ adde $a2,$a2,$m2
+ adde $a3,$a3,$m3
+
+ $ST $a0,$SIZE_T*1($ap) # write result
+ $ST $a1,$SIZE_T*2($ap)
+ $ST $a2,$SIZE_T*3($ap)
+ $ST $a3,$SIZE_T*4($ap)
+
+.Lmul4x_done:
+ $ST $zero,$SIZE_T*8($sp) # wipe stack clean
+ $ST $zero,$SIZE_T*9($sp)
+ $ST $zero,$SIZE_T*10($sp)
+ $ST $zero,$SIZE_T*11($sp)
+ li r3,1 # signal "done"
+ $POP r14,-$SIZE_T*18($bp)
+ $POP r15,-$SIZE_T*17($bp)
+ $POP r16,-$SIZE_T*16($bp)
+ $POP r17,-$SIZE_T*15($bp)
+ $POP r18,-$SIZE_T*14($bp)
+ $POP r19,-$SIZE_T*13($bp)
+ $POP r20,-$SIZE_T*12($bp)
+ $POP r21,-$SIZE_T*11($bp)
+ $POP r22,-$SIZE_T*10($bp)
+ $POP r23,-$SIZE_T*9($bp)
+ $POP r24,-$SIZE_T*8($bp)
+ $POP r25,-$SIZE_T*7($bp)
+ $POP r26,-$SIZE_T*6($bp)
+ $POP r27,-$SIZE_T*5($bp)
+ $POP r28,-$SIZE_T*4($bp)
+ $POP r29,-$SIZE_T*3($bp)
+ $POP r30,-$SIZE_T*2($bp)
+ $POP r31,-$SIZE_T*1($bp)
+ mr $sp,$bp
+ blr
+ .long 0
+ .byte 0,12,4,0x20,0x80,18,6,0
+ .long 0
+.size .bn_mul4x_mont_int,.-.bn_mul4x_mont_int
+___
+}
+
+if (1) {
+########################################################################
+# Following is PPC adaptation of sqrx8x_mont from x86_64-mont5 module.
+
+my ($a0,$a1,$a2,$a3,$a4,$a5,$a6,$a7)=map("r$_",(9..12,14..17));
+my ($t0,$t1,$t2,$t3)=map("r$_",(18..21));
+my ($acc0,$acc1,$acc2,$acc3,$acc4,$acc5,$acc6,$acc7)=map("r$_",(22..29));
+my ($cnt,$carry,$zero)=("r30","r31","r0");
+my ($tp,$ap_end,$na0)=($bp,$np,$carry);
+
+# sp----------->+-------------------------------+
+# | saved sp |
+# +-------------------------------+
+# . .
+# +12*size_t +-------------------------------+
+# | size_t tmp[2*num] |
+# . .
+# . .
+# . .
+# +-------------------------------+
+# . .
+# -18*size_t +-------------------------------+
+# | 18 saved gpr, r14-r31 |
+# . .
+# . .
+# +-------------------------------+
+$code.=<<___;
+.align 5
+__bn_sqr8x_mont:
+.Lsqr8x_do:
+ mr $a0,$sp
+ slwi $a1,$num,`log($SIZE_T)/log(2)+1`
+ li $a2,-32*$SIZE_T
+ sub $a1,$a2,$a1
+ slwi $num,$num,`log($SIZE_T)/log(2)`
+ $STUX $sp,$sp,$a1 # alloca
+
+ $PUSH r14,-$SIZE_T*18($a0)
+ $PUSH r15,-$SIZE_T*17($a0)
+ $PUSH r16,-$SIZE_T*16($a0)
+ $PUSH r17,-$SIZE_T*15($a0)
+ $PUSH r18,-$SIZE_T*14($a0)
+ $PUSH r19,-$SIZE_T*13($a0)
+ $PUSH r20,-$SIZE_T*12($a0)
+ $PUSH r21,-$SIZE_T*11($a0)
+ $PUSH r22,-$SIZE_T*10($a0)
+ $PUSH r23,-$SIZE_T*9($a0)
+ $PUSH r24,-$SIZE_T*8($a0)
+ $PUSH r25,-$SIZE_T*7($a0)
+ $PUSH r26,-$SIZE_T*6($a0)
+ $PUSH r27,-$SIZE_T*5($a0)
+ $PUSH r28,-$SIZE_T*4($a0)
+ $PUSH r29,-$SIZE_T*3($a0)
+ $PUSH r30,-$SIZE_T*2($a0)
+ $PUSH r31,-$SIZE_T*1($a0)
+
+ subi $ap,$ap,$SIZE_T # bias by -1
+ subi $t0,$np,$SIZE_T # bias by -1
+ subi $rp,$rp,$SIZE_T # bias by -1
+ $LD $n0,0($n0) # *n0
+ li $zero,0
+
+ add $ap_end,$ap,$num
+ $LD $a0,$SIZE_T*1($ap)
+ #li $acc0,0
+ $LD $a1,$SIZE_T*2($ap)
+ li $acc1,0
+ $LD $a2,$SIZE_T*3($ap)
+ li $acc2,0
+ $LD $a3,$SIZE_T*4($ap)
+ li $acc3,0
+ $LD $a4,$SIZE_T*5($ap)
+ li $acc4,0
+ $LD $a5,$SIZE_T*6($ap)
+ li $acc5,0
+ $LD $a6,$SIZE_T*7($ap)
+ li $acc6,0
+ $LDU $a7,$SIZE_T*8($ap)
+ li $acc7,0
+
+ addi $tp,$sp,$SIZE_T*11 # &tp[-1]
+ subic. $cnt,$num,$SIZE_T*8
+ b .Lsqr8x_zero_start
+
+.align 5
+.Lsqr8x_zero:
+ subic. $cnt,$cnt,$SIZE_T*8
+ $ST $zero,$SIZE_T*1($tp)
+ $ST $zero,$SIZE_T*2($tp)
+ $ST $zero,$SIZE_T*3($tp)
+ $ST $zero,$SIZE_T*4($tp)
+ $ST $zero,$SIZE_T*5($tp)
+ $ST $zero,$SIZE_T*6($tp)
+ $ST $zero,$SIZE_T*7($tp)
+ $ST $zero,$SIZE_T*8($tp)
+.Lsqr8x_zero_start:
+ $ST $zero,$SIZE_T*9($tp)
+ $ST $zero,$SIZE_T*10($tp)
+ $ST $zero,$SIZE_T*11($tp)
+ $ST $zero,$SIZE_T*12($tp)
+ $ST $zero,$SIZE_T*13($tp)
+ $ST $zero,$SIZE_T*14($tp)
+ $ST $zero,$SIZE_T*15($tp)
+ $STU $zero,$SIZE_T*16($tp)
+ bne .Lsqr8x_zero
+
+ $PUSH $rp,$SIZE_T*6($sp) # offload &rp[-1]
+ $PUSH $t0,$SIZE_T*7($sp) # offload &np[-1]
+ $PUSH $n0,$SIZE_T*8($sp) # offload n0
+ $PUSH $tp,$SIZE_T*9($sp) # &tp[2*num-1]
+ $PUSH $zero,$SIZE_T*10($sp) # initial top-most carry
+ addi $tp,$sp,$SIZE_T*11 # &tp[-1]
+
+ # Multiply everything but a[i]*a[i]
+.align 5
+.Lsqr8x_outer_loop:
+ # a[1]a[0] (i)
+ # a[2]a[0]
+ # a[3]a[0]
+ # a[4]a[0]
+ # a[5]a[0]
+ # a[6]a[0]
+ # a[7]a[0]
+ # a[2]a[1] (ii)
+ # a[3]a[1]
+ # a[4]a[1]
+ # a[5]a[1]
+ # a[6]a[1]
+ # a[7]a[1]
+ # a[3]a[2] (iii)
+ # a[4]a[2]
+ # a[5]a[2]
+ # a[6]a[2]
+ # a[7]a[2]
+ # a[4]a[3] (iv)
+ # a[5]a[3]
+ # a[6]a[3]
+ # a[7]a[3]
+ # a[5]a[4] (v)
+ # a[6]a[4]
+ # a[7]a[4]
+ # a[6]a[5] (vi)
+ # a[7]a[5]
+ # a[7]a[6] (vii)
+
+ $UMULL $t0,$a1,$a0 # lo(a[1..7]*a[0]) (i)
+ $UMULL $t1,$a2,$a0
+ $UMULL $t2,$a3,$a0
+ $UMULL $t3,$a4,$a0
+ addc $acc1,$acc1,$t0 # t[1]+lo(a[1]*a[0])
+ $UMULL $t0,$a5,$a0
+ adde $acc2,$acc2,$t1
+ $UMULL $t1,$a6,$a0
+ adde $acc3,$acc3,$t2
+ $UMULL $t2,$a7,$a0
+ adde $acc4,$acc4,$t3
+ $UMULH $t3,$a1,$a0 # hi(a[1..7]*a[0])
+ adde $acc5,$acc5,$t0
+ $UMULH $t0,$a2,$a0
+ adde $acc6,$acc6,$t1
+ $UMULH $t1,$a3,$a0
+ adde $acc7,$acc7,$t2
+ $UMULH $t2,$a4,$a0
+ $ST $acc0,$SIZE_T*1($tp) # t[0]
+ addze $acc0,$zero # t[8]
+ $ST $acc1,$SIZE_T*2($tp) # t[1]
+ addc $acc2,$acc2,$t3 # t[2]+lo(a[1]*a[0])
+ $UMULH $t3,$a5,$a0
+ adde $acc3,$acc3,$t0
+ $UMULH $t0,$a6,$a0
+ adde $acc4,$acc4,$t1
+ $UMULH $t1,$a7,$a0
+ adde $acc5,$acc5,$t2
+ $UMULL $t2,$a2,$a1 # lo(a[2..7]*a[1]) (ii)
+ adde $acc6,$acc6,$t3
+ $UMULL $t3,$a3,$a1
+ adde $acc7,$acc7,$t0
+ $UMULL $t0,$a4,$a1
+ adde $acc0,$acc0,$t1
+
+ $UMULL $t1,$a5,$a1
+ addc $acc3,$acc3,$t2
+ $UMULL $t2,$a6,$a1
+ adde $acc4,$acc4,$t3
+ $UMULL $t3,$a7,$a1
+ adde $acc5,$acc5,$t0
+ $UMULH $t0,$a2,$a1 # hi(a[2..7]*a[1])
+ adde $acc6,$acc6,$t1
+ $UMULH $t1,$a3,$a1
+ adde $acc7,$acc7,$t2
+ $UMULH $t2,$a4,$a1
+ adde $acc0,$acc0,$t3
+ $UMULH $t3,$a5,$a1
+ $ST $acc2,$SIZE_T*3($tp) # t[2]
+ addze $acc1,$zero # t[9]
+ $ST $acc3,$SIZE_T*4($tp) # t[3]
+ addc $acc4,$acc4,$t0
+ $UMULH $t0,$a6,$a1
+ adde $acc5,$acc5,$t1
+ $UMULH $t1,$a7,$a1
+ adde $acc6,$acc6,$t2
+ $UMULL $t2,$a3,$a2 # lo(a[3..7]*a[2]) (iii)
+ adde $acc7,$acc7,$t3
+ $UMULL $t3,$a4,$a2
+ adde $acc0,$acc0,$t0
+ $UMULL $t0,$a5,$a2
+ adde $acc1,$acc1,$t1
+
+ $UMULL $t1,$a6,$a2
+ addc $acc5,$acc5,$t2
+ $UMULL $t2,$a7,$a2
+ adde $acc6,$acc6,$t3
+ $UMULH $t3,$a3,$a2 # hi(a[3..7]*a[2])
+ adde $acc7,$acc7,$t0
+ $UMULH $t0,$a4,$a2
+ adde $acc0,$acc0,$t1
+ $UMULH $t1,$a5,$a2
+ adde $acc1,$acc1,$t2
+ $UMULH $t2,$a6,$a2
+ $ST $acc4,$SIZE_T*5($tp) # t[4]
+ addze $acc2,$zero # t[10]
+ $ST $acc5,$SIZE_T*6($tp) # t[5]
+ addc $acc6,$acc6,$t3
+ $UMULH $t3,$a7,$a2
+ adde $acc7,$acc7,$t0
+ $UMULL $t0,$a4,$a3 # lo(a[4..7]*a[3]) (iv)
+ adde $acc0,$acc0,$t1
+ $UMULL $t1,$a5,$a3
+ adde $acc1,$acc1,$t2
+ $UMULL $t2,$a6,$a3
+ adde $acc2,$acc2,$t3
+
+ $UMULL $t3,$a7,$a3
+ addc $acc7,$acc7,$t0
+ $UMULH $t0,$a4,$a3 # hi(a[4..7]*a[3])
+ adde $acc0,$acc0,$t1
+ $UMULH $t1,$a5,$a3
+ adde $acc1,$acc1,$t2
+ $UMULH $t2,$a6,$a3
+ adde $acc2,$acc2,$t3
+ $UMULH $t3,$a7,$a3
+ $ST $acc6,$SIZE_T*7($tp) # t[6]
+ addze $acc3,$zero # t[11]
+ $STU $acc7,$SIZE_T*8($tp) # t[7]
+ addc $acc0,$acc0,$t0
+ $UMULL $t0,$a5,$a4 # lo(a[5..7]*a[4]) (v)
+ adde $acc1,$acc1,$t1
+ $UMULL $t1,$a6,$a4
+ adde $acc2,$acc2,$t2
+ $UMULL $t2,$a7,$a4
+ adde $acc3,$acc3,$t3
+
+ $UMULH $t3,$a5,$a4 # hi(a[5..7]*a[4])
+ addc $acc1,$acc1,$t0
+ $UMULH $t0,$a6,$a4
+ adde $acc2,$acc2,$t1
+ $UMULH $t1,$a7,$a4
+ adde $acc3,$acc3,$t2
+ $UMULL $t2,$a6,$a5 # lo(a[6..7]*a[5]) (vi)
+ addze $acc4,$zero # t[12]
+ addc $acc2,$acc2,$t3
+ $UMULL $t3,$a7,$a5
+ adde $acc3,$acc3,$t0
+ $UMULH $t0,$a6,$a5 # hi(a[6..7]*a[5])
+ adde $acc4,$acc4,$t1
+
+ $UMULH $t1,$a7,$a5
+ addc $acc3,$acc3,$t2
+ $UMULL $t2,$a7,$a6 # lo(a[7]*a[6]) (vii)
+ adde $acc4,$acc4,$t3
+ $UMULH $t3,$a7,$a6 # hi(a[7]*a[6])
+ addze $acc5,$zero # t[13]
+ addc $acc4,$acc4,$t0
+ $UCMP $ap_end,$ap # done yet?
+ adde $acc5,$acc5,$t1
+
+ addc $acc5,$acc5,$t2
+ sub $t0,$ap_end,$num # rewinded ap
+ addze $acc6,$zero # t[14]
+ add $acc6,$acc6,$t3
+
+ beq .Lsqr8x_outer_break
+
+ mr $n0,$a0
+ $LD $a0,$SIZE_T*1($tp)
+ $LD $a1,$SIZE_T*2($tp)
+ $LD $a2,$SIZE_T*3($tp)
+ $LD $a3,$SIZE_T*4($tp)
+ $LD $a4,$SIZE_T*5($tp)
+ $LD $a5,$SIZE_T*6($tp)
+ $LD $a6,$SIZE_T*7($tp)
+ $LD $a7,$SIZE_T*8($tp)
+ addc $acc0,$acc0,$a0
+ $LD $a0,$SIZE_T*1($ap)
+ adde $acc1,$acc1,$a1
+ $LD $a1,$SIZE_T*2($ap)
+ adde $acc2,$acc2,$a2
+ $LD $a2,$SIZE_T*3($ap)
+ adde $acc3,$acc3,$a3
+ $LD $a3,$SIZE_T*4($ap)
+ adde $acc4,$acc4,$a4
+ $LD $a4,$SIZE_T*5($ap)
+ adde $acc5,$acc5,$a5
+ $LD $a5,$SIZE_T*6($ap)
+ adde $acc6,$acc6,$a6
+ $LD $a6,$SIZE_T*7($ap)
+ subi $rp,$ap,$SIZE_T*7
+ addze $acc7,$a7
+ $LDU $a7,$SIZE_T*8($ap)
+ #addze $carry,$zero # moved below
+ li $cnt,0
+ b .Lsqr8x_mul
+
+ # a[8]a[0]
+ # a[9]a[0]
+ # a[a]a[0]
+ # a[b]a[0]
+ # a[c]a[0]
+ # a[d]a[0]
+ # a[e]a[0]
+ # a[f]a[0]
+ # a[8]a[1]
+ # a[f]a[1]........................
+ # a[8]a[2]
+ # a[f]a[2]........................
+ # a[8]a[3]
+ # a[f]a[3]........................
+ # a[8]a[4]
+ # a[f]a[4]........................
+ # a[8]a[5]
+ # a[f]a[5]........................
+ # a[8]a[6]
+ # a[f]a[6]........................
+ # a[8]a[7]
+ # a[f]a[7]........................
+.align 5
+.Lsqr8x_mul:
+ $UMULL $t0,$a0,$n0
+ addze $carry,$zero # carry bit, modulo-scheduled
+ $UMULL $t1,$a1,$n0
+ addi $cnt,$cnt,$SIZE_T
+ $UMULL $t2,$a2,$n0
+ andi. $cnt,$cnt,$SIZE_T*8-1
+ $UMULL $t3,$a3,$n0
+ addc $acc0,$acc0,$t0
+ $UMULL $t0,$a4,$n0
+ adde $acc1,$acc1,$t1
+ $UMULL $t1,$a5,$n0
+ adde $acc2,$acc2,$t2
+ $UMULL $t2,$a6,$n0
+ adde $acc3,$acc3,$t3
+ $UMULL $t3,$a7,$n0
+ adde $acc4,$acc4,$t0
+ $UMULH $t0,$a0,$n0
+ adde $acc5,$acc5,$t1
+ $UMULH $t1,$a1,$n0
+ adde $acc6,$acc6,$t2
+ $UMULH $t2,$a2,$n0
+ adde $acc7,$acc7,$t3
+ $UMULH $t3,$a3,$n0
+ addze $carry,$carry
+ $STU $acc0,$SIZE_T($tp)
+ addc $acc0,$acc1,$t0
+ $UMULH $t0,$a4,$n0
+ adde $acc1,$acc2,$t1
+ $UMULH $t1,$a5,$n0
+ adde $acc2,$acc3,$t2
+ $UMULH $t2,$a6,$n0
+ adde $acc3,$acc4,$t3
+ $UMULH $t3,$a7,$n0
+ $LDX $n0,$rp,$cnt
+ adde $acc4,$acc5,$t0
+ adde $acc5,$acc6,$t1
+ adde $acc6,$acc7,$t2
+ adde $acc7,$carry,$t3
+ #addze $carry,$zero # moved above
+ bne .Lsqr8x_mul
+ # note that carry flag is guaranteed
+ # to be zero at this point
+ $UCMP $ap,$ap_end # done yet?
+ beq .Lsqr8x_break
+
+ $LD $a0,$SIZE_T*1($tp)
+ $LD $a1,$SIZE_T*2($tp)
+ $LD $a2,$SIZE_T*3($tp)
+ $LD $a3,$SIZE_T*4($tp)
+ $LD $a4,$SIZE_T*5($tp)
+ $LD $a5,$SIZE_T*6($tp)
+ $LD $a6,$SIZE_T*7($tp)
+ $LD $a7,$SIZE_T*8($tp)
+ addc $acc0,$acc0,$a0
+ $LD $a0,$SIZE_T*1($ap)
+ adde $acc1,$acc1,$a1
+ $LD $a1,$SIZE_T*2($ap)
+ adde $acc2,$acc2,$a2
+ $LD $a2,$SIZE_T*3($ap)
+ adde $acc3,$acc3,$a3
+ $LD $a3,$SIZE_T*4($ap)
+ adde $acc4,$acc4,$a4
+ $LD $a4,$SIZE_T*5($ap)
+ adde $acc5,$acc5,$a5
+ $LD $a5,$SIZE_T*6($ap)
+ adde $acc6,$acc6,$a6
+ $LD $a6,$SIZE_T*7($ap)
+ adde $acc7,$acc7,$a7
+ $LDU $a7,$SIZE_T*8($ap)
+ #addze $carry,$zero # moved above
+ b .Lsqr8x_mul
+
+.align 5
+.Lsqr8x_break:
+ $LD $a0,$SIZE_T*8($rp)
+ addi $ap,$rp,$SIZE_T*15
+ $LD $a1,$SIZE_T*9($rp)
+ sub. $t0,$ap_end,$ap # is it last iteration?
+ $LD $a2,$SIZE_T*10($rp)
+ sub $t1,$tp,$t0
+ $LD $a3,$SIZE_T*11($rp)
+ $LD $a4,$SIZE_T*12($rp)
+ $LD $a5,$SIZE_T*13($rp)
+ $LD $a6,$SIZE_T*14($rp)
+ $LD $a7,$SIZE_T*15($rp)
+ beq .Lsqr8x_outer_loop
+
+ $ST $acc0,$SIZE_T*1($tp)
+ $LD $acc0,$SIZE_T*1($t1)
+ $ST $acc1,$SIZE_T*2($tp)
+ $LD $acc1,$SIZE_T*2($t1)
+ $ST $acc2,$SIZE_T*3($tp)
+ $LD $acc2,$SIZE_T*3($t1)
+ $ST $acc3,$SIZE_T*4($tp)
+ $LD $acc3,$SIZE_T*4($t1)
+ $ST $acc4,$SIZE_T*5($tp)
+ $LD $acc4,$SIZE_T*5($t1)
+ $ST $acc5,$SIZE_T*6($tp)
+ $LD $acc5,$SIZE_T*6($t1)
+ $ST $acc6,$SIZE_T*7($tp)
+ $LD $acc6,$SIZE_T*7($t1)
+ $ST $acc7,$SIZE_T*8($tp)
+ $LD $acc7,$SIZE_T*8($t1)
+ mr $tp,$t1
+ b .Lsqr8x_outer_loop
+
+.align 5
+.Lsqr8x_outer_break:
+ ####################################################################
+ # Now multiply above result by 2 and add a[n-1]*a[n-1]|...|a[0]*a[0]
+ $LD $a1,$SIZE_T*1($t0) # recall that $t0 is &a[-1]
+ $LD $a3,$SIZE_T*2($t0)
+ $LD $a5,$SIZE_T*3($t0)
+ $LD $a7,$SIZE_T*4($t0)
+ addi $ap,$t0,$SIZE_T*4
+ # "tp[x]" comments are for num==8 case
+ $LD $t1,$SIZE_T*13($sp) # =tp[1], t[0] is not interesting
+ $LD $t2,$SIZE_T*14($sp)
+ $LD $t3,$SIZE_T*15($sp)
+ $LD $t0,$SIZE_T*16($sp)
+
+ $ST $acc0,$SIZE_T*1($tp) # tp[8]=
+ srwi $cnt,$num,`log($SIZE_T)/log(2)+2`
+ $ST $acc1,$SIZE_T*2($tp)
+ subi $cnt,$cnt,1
+ $ST $acc2,$SIZE_T*3($tp)
+ $ST $acc3,$SIZE_T*4($tp)
+ $ST $acc4,$SIZE_T*5($tp)
+ $ST $acc5,$SIZE_T*6($tp)
+ $ST $acc6,$SIZE_T*7($tp)
+ #$ST $acc7,$SIZE_T*8($tp) # tp[15] is not interesting
+ addi $tp,$sp,$SIZE_T*11 # &tp[-1]
+ $UMULL $acc0,$a1,$a1
+ $UMULH $a1,$a1,$a1
+ add $acc1,$t1,$t1 # <<1
+ $SHRI $t1,$t1,$BITS-1
+ $UMULL $a2,$a3,$a3
+ $UMULH $a3,$a3,$a3
+ addc $acc1,$acc1,$a1
+ add $acc2,$t2,$t2
+ $SHRI $t2,$t2,$BITS-1
+ add $acc3,$t3,$t3
+ $SHRI $t3,$t3,$BITS-1
+ or $acc2,$acc2,$t1
+
+ mtctr $cnt
+.Lsqr4x_shift_n_add:
+ $UMULL $a4,$a5,$a5
+ $UMULH $a5,$a5,$a5
+ $LD $t1,$SIZE_T*6($tp) # =tp[5]
+ $LD $a1,$SIZE_T*1($ap)
+ adde $acc2,$acc2,$a2
+ add $acc4,$t0,$t0
+ $SHRI $t0,$t0,$BITS-1
+ or $acc3,$acc3,$t2
+ $LD $t2,$SIZE_T*7($tp) # =tp[6]
+ adde $acc3,$acc3,$a3
+ $LD $a3,$SIZE_T*2($ap)
+ add $acc5,$t1,$t1
+ $SHRI $t1,$t1,$BITS-1
+ or $acc4,$acc4,$t3
+ $LD $t3,$SIZE_T*8($tp) # =tp[7]
+ $UMULL $a6,$a7,$a7
+ $UMULH $a7,$a7,$a7
+ adde $acc4,$acc4,$a4
+ add $acc6,$t2,$t2
+ $SHRI $t2,$t2,$BITS-1
+ or $acc5,$acc5,$t0
+ $LD $t0,$SIZE_T*9($tp) # =tp[8]
+ adde $acc5,$acc5,$a5
+ $LD $a5,$SIZE_T*3($ap)
+ add $acc7,$t3,$t3
+ $SHRI $t3,$t3,$BITS-1
+ or $acc6,$acc6,$t1
+ $LD $t1,$SIZE_T*10($tp) # =tp[9]
+ $UMULL $a0,$a1,$a1
+ $UMULH $a1,$a1,$a1
+ adde $acc6,$acc6,$a6
+ $ST $acc0,$SIZE_T*1($tp) # tp[0]=
+ add $acc0,$t0,$t0
+ $SHRI $t0,$t0,$BITS-1
+ or $acc7,$acc7,$t2
+ $LD $t2,$SIZE_T*11($tp) # =tp[10]
+ adde $acc7,$acc7,$a7
+ $LDU $a7,$SIZE_T*4($ap)
+ $ST $acc1,$SIZE_T*2($tp) # tp[1]=
+ add $acc1,$t1,$t1
+ $SHRI $t1,$t1,$BITS-1
+ or $acc0,$acc0,$t3
+ $LD $t3,$SIZE_T*12($tp) # =tp[11]
+ $UMULL $a2,$a3,$a3
+ $UMULH $a3,$a3,$a3
+ adde $acc0,$acc0,$a0
+ $ST $acc2,$SIZE_T*3($tp) # tp[2]=
+ add $acc2,$t2,$t2
+ $SHRI $t2,$t2,$BITS-1
+ or $acc1,$acc1,$t0
+ $LD $t0,$SIZE_T*13($tp) # =tp[12]
+ adde $acc1,$acc1,$a1
+ $ST $acc3,$SIZE_T*4($tp) # tp[3]=
+ $ST $acc4,$SIZE_T*5($tp) # tp[4]=
+ $ST $acc5,$SIZE_T*6($tp) # tp[5]=
+ $ST $acc6,$SIZE_T*7($tp) # tp[6]=
+ $STU $acc7,$SIZE_T*8($tp) # tp[7]=
+ add $acc3,$t3,$t3
+ $SHRI $t3,$t3,$BITS-1
+ or $acc2,$acc2,$t1
+ bdnz .Lsqr4x_shift_n_add
+___
+my ($np,$np_end)=($ap,$ap_end);
+$code.=<<___;
+ $POP $np,$SIZE_T*7($sp) # pull &np[-1] and n0
+ $POP $n0,$SIZE_T*8($sp)
+
+ $UMULL $a4,$a5,$a5
+ $UMULH $a5,$a5,$a5
+ $ST $acc0,$SIZE_T*1($tp) # tp[8]=
+ $LD $acc0,$SIZE_T*12($sp) # =tp[0]
+ $LD $t1,$SIZE_T*6($tp) # =tp[13]
+ adde $acc2,$acc2,$a2
+ add $acc4,$t0,$t0
+ $SHRI $t0,$t0,$BITS-1
+ or $acc3,$acc3,$t2
+ $LD $t2,$SIZE_T*7($tp) # =tp[14]
+ adde $acc3,$acc3,$a3
+ add $acc5,$t1,$t1
+ $SHRI $t1,$t1,$BITS-1
+ or $acc4,$acc4,$t3
+ $UMULL $a6,$a7,$a7
+ $UMULH $a7,$a7,$a7
+ adde $acc4,$acc4,$a4
+ add $acc6,$t2,$t2
+ $SHRI $t2,$t2,$BITS-1
+ or $acc5,$acc5,$t0
+ $ST $acc1,$SIZE_T*2($tp) # tp[9]=
+ $LD $acc1,$SIZE_T*13($sp) # =tp[1]
+ adde $acc5,$acc5,$a5
+ or $acc6,$acc6,$t1
+ $LD $a0,$SIZE_T*1($np)
+ $LD $a1,$SIZE_T*2($np)
+ adde $acc6,$acc6,$a6
+ $LD $a2,$SIZE_T*3($np)
+ $LD $a3,$SIZE_T*4($np)
+ adde $acc7,$a7,$t2
+ $LD $a4,$SIZE_T*5($np)
+ $LD $a5,$SIZE_T*6($np)
+
+ ################################################################
+ # Reduce by 8 limbs per iteration
+ $UMULL $na0,$n0,$acc0 # t[0]*n0
+ li $cnt,8
+ $LD $a6,$SIZE_T*7($np)
+ add $np_end,$np,$num
+ $LDU $a7,$SIZE_T*8($np)
+ $ST $acc2,$SIZE_T*3($tp) # tp[10]=
+ $LD $acc2,$SIZE_T*14($sp)
+ $ST $acc3,$SIZE_T*4($tp) # tp[11]=
+ $LD $acc3,$SIZE_T*15($sp)
+ $ST $acc4,$SIZE_T*5($tp) # tp[12]=
+ $LD $acc4,$SIZE_T*16($sp)
+ $ST $acc5,$SIZE_T*6($tp) # tp[13]=
+ $LD $acc5,$SIZE_T*17($sp)
+ $ST $acc6,$SIZE_T*7($tp) # tp[14]=
+ $LD $acc6,$SIZE_T*18($sp)
+ $ST $acc7,$SIZE_T*8($tp) # tp[15]=
+ $LD $acc7,$SIZE_T*19($sp)
+ addi $tp,$sp,$SIZE_T*11 # &tp[-1]
+ mtctr $cnt
+ b .Lsqr8x_reduction
+
+.align 5
+.Lsqr8x_reduction:
+ # (*) $UMULL $t0,$a0,$na0 # lo(n[0-7])*lo(t[0]*n0)
+ $UMULL $t1,$a1,$na0
+ $UMULL $t2,$a2,$na0
+ $STU $na0,$SIZE_T($tp) # put aside t[0]*n0 for tail processing
+ $UMULL $t3,$a3,$na0
+ # (*) addc $acc0,$acc0,$t0
+ addic $acc0,$acc0,-1 # (*)
+ $UMULL $t0,$a4,$na0
+ adde $acc0,$acc1,$t1
+ $UMULL $t1,$a5,$na0
+ adde $acc1,$acc2,$t2
+ $UMULL $t2,$a6,$na0
+ adde $acc2,$acc3,$t3
+ $UMULL $t3,$a7,$na0
+ adde $acc3,$acc4,$t0
+ $UMULH $t0,$a0,$na0 # hi(n[0-7])*lo(t[0]*n0)
+ adde $acc4,$acc5,$t1
+ $UMULH $t1,$a1,$na0
+ adde $acc5,$acc6,$t2
+ $UMULH $t2,$a2,$na0
+ adde $acc6,$acc7,$t3
+ $UMULH $t3,$a3,$na0
+ addze $acc7,$zero
+ addc $acc0,$acc0,$t0
+ $UMULH $t0,$a4,$na0
+ adde $acc1,$acc1,$t1
+ $UMULH $t1,$a5,$na0
+ adde $acc2,$acc2,$t2
+ $UMULH $t2,$a6,$na0
+ adde $acc3,$acc3,$t3
+ $UMULH $t3,$a7,$na0
+ $UMULL $na0,$n0,$acc0 # next t[0]*n0
+ adde $acc4,$acc4,$t0
+ adde $acc5,$acc5,$t1
+ adde $acc6,$acc6,$t2
+ adde $acc7,$acc7,$t3
+ bdnz .Lsqr8x_reduction
+
+ $LD $t0,$SIZE_T*1($tp)
+ $LD $t1,$SIZE_T*2($tp)
+ $LD $t2,$SIZE_T*3($tp)
+ $LD $t3,$SIZE_T*4($tp)
+ subi $rp,$tp,$SIZE_T*7
+ $UCMP $np_end,$np # done yet?
+ addc $acc0,$acc0,$t0
+ $LD $t0,$SIZE_T*5($tp)
+ adde $acc1,$acc1,$t1
+ $LD $t1,$SIZE_T*6($tp)
+ adde $acc2,$acc2,$t2
+ $LD $t2,$SIZE_T*7($tp)
+ adde $acc3,$acc3,$t3
+ $LD $t3,$SIZE_T*8($tp)
+ adde $acc4,$acc4,$t0
+ adde $acc5,$acc5,$t1
+ adde $acc6,$acc6,$t2
+ adde $acc7,$acc7,$t3
+ #addze $carry,$zero # moved below
+ beq .Lsqr8x8_post_condition
+
+ $LD $n0,$SIZE_T*0($rp)
+ $LD $a0,$SIZE_T*1($np)
+ $LD $a1,$SIZE_T*2($np)
+ $LD $a2,$SIZE_T*3($np)
+ $LD $a3,$SIZE_T*4($np)
+ $LD $a4,$SIZE_T*5($np)
+ $LD $a5,$SIZE_T*6($np)
+ $LD $a6,$SIZE_T*7($np)
+ $LDU $a7,$SIZE_T*8($np)
+ li $cnt,0
+
+.align 5
+.Lsqr8x_tail:
+ $UMULL $t0,$a0,$n0
+ addze $carry,$zero # carry bit, modulo-scheduled
+ $UMULL $t1,$a1,$n0
+ addi $cnt,$cnt,$SIZE_T
+ $UMULL $t2,$a2,$n0
+ andi. $cnt,$cnt,$SIZE_T*8-1
+ $UMULL $t3,$a3,$n0
+ addc $acc0,$acc0,$t0
+ $UMULL $t0,$a4,$n0
+ adde $acc1,$acc1,$t1
+ $UMULL $t1,$a5,$n0
+ adde $acc2,$acc2,$t2
+ $UMULL $t2,$a6,$n0
+ adde $acc3,$acc3,$t3
+ $UMULL $t3,$a7,$n0
+ adde $acc4,$acc4,$t0
+ $UMULH $t0,$a0,$n0
+ adde $acc5,$acc5,$t1
+ $UMULH $t1,$a1,$n0
+ adde $acc6,$acc6,$t2
+ $UMULH $t2,$a2,$n0
+ adde $acc7,$acc7,$t3
+ $UMULH $t3,$a3,$n0
+ addze $carry,$carry
+ $STU $acc0,$SIZE_T($tp)
+ addc $acc0,$acc1,$t0
+ $UMULH $t0,$a4,$n0
+ adde $acc1,$acc2,$t1
+ $UMULH $t1,$a5,$n0
+ adde $acc2,$acc3,$t2
+ $UMULH $t2,$a6,$n0
+ adde $acc3,$acc4,$t3
+ $UMULH $t3,$a7,$n0
+ $LDX $n0,$rp,$cnt
+ adde $acc4,$acc5,$t0
+ adde $acc5,$acc6,$t1
+ adde $acc6,$acc7,$t2
+ adde $acc7,$carry,$t3
+ #addze $carry,$zero # moved above
+ bne .Lsqr8x_tail
+ # note that carry flag is guaranteed
+ # to be zero at this point
+ $LD $a0,$SIZE_T*1($tp)
+ $POP $carry,$SIZE_T*10($sp) # pull top-most carry in case we break
+ $UCMP $np_end,$np # done yet?
+ $LD $a1,$SIZE_T*2($tp)
+ sub $t2,$np_end,$num # rewinded np
+ $LD $a2,$SIZE_T*3($tp)
+ $LD $a3,$SIZE_T*4($tp)
+ $LD $a4,$SIZE_T*5($tp)
+ $LD $a5,$SIZE_T*6($tp)
+ $LD $a6,$SIZE_T*7($tp)
+ $LD $a7,$SIZE_T*8($tp)
+ beq .Lsqr8x_tail_break
+
+ addc $acc0,$acc0,$a0
+ $LD $a0,$SIZE_T*1($np)
+ adde $acc1,$acc1,$a1
+ $LD $a1,$SIZE_T*2($np)
+ adde $acc2,$acc2,$a2
+ $LD $a2,$SIZE_T*3($np)
+ adde $acc3,$acc3,$a3
+ $LD $a3,$SIZE_T*4($np)
+ adde $acc4,$acc4,$a4
+ $LD $a4,$SIZE_T*5($np)
+ adde $acc5,$acc5,$a5
+ $LD $a5,$SIZE_T*6($np)
+ adde $acc6,$acc6,$a6
+ $LD $a6,$SIZE_T*7($np)
+ adde $acc7,$acc7,$a7
+ $LDU $a7,$SIZE_T*8($np)
+ #addze $carry,$zero # moved above
+ b .Lsqr8x_tail
+
+.align 5
+.Lsqr8x_tail_break:
+ $POP $n0,$SIZE_T*8($sp) # pull n0
+ $POP $t3,$SIZE_T*9($sp) # &tp[2*num-1]
+ addi $cnt,$tp,$SIZE_T*8 # end of current t[num] window
+
+ addic $carry,$carry,-1 # "move" top-most carry to carry bit
+ adde $t0,$acc0,$a0
+ $LD $acc0,$SIZE_T*8($rp)
+ $LD $a0,$SIZE_T*1($t2) # recall that $t2 is &n[-1]
+ adde $t1,$acc1,$a1
+ $LD $acc1,$SIZE_T*9($rp)
+ $LD $a1,$SIZE_T*2($t2)
+ adde $acc2,$acc2,$a2
+ $LD $a2,$SIZE_T*3($t2)
+ adde $acc3,$acc3,$a3
+ $LD $a3,$SIZE_T*4($t2)
+ adde $acc4,$acc4,$a4
+ $LD $a4,$SIZE_T*5($t2)
+ adde $acc5,$acc5,$a5
+ $LD $a5,$SIZE_T*6($t2)
+ adde $acc6,$acc6,$a6
+ $LD $a6,$SIZE_T*7($t2)
+ adde $acc7,$acc7,$a7
+ $LD $a7,$SIZE_T*8($t2)
+ addi $np,$t2,$SIZE_T*8
+ addze $t2,$zero # top-most carry
+ $UMULL $na0,$n0,$acc0
+ $ST $t0,$SIZE_T*1($tp)
+ $UCMP $cnt,$t3 # did we hit the bottom?
+ $ST $t1,$SIZE_T*2($tp)
+ li $cnt,8
+ $ST $acc2,$SIZE_T*3($tp)
+ $LD $acc2,$SIZE_T*10($rp)
+ $ST $acc3,$SIZE_T*4($tp)
+ $LD $acc3,$SIZE_T*11($rp)
+ $ST $acc4,$SIZE_T*5($tp)
+ $LD $acc4,$SIZE_T*12($rp)
+ $ST $acc5,$SIZE_T*6($tp)
+ $LD $acc5,$SIZE_T*13($rp)
+ $ST $acc6,$SIZE_T*7($tp)
+ $LD $acc6,$SIZE_T*14($rp)
+ $ST $acc7,$SIZE_T*8($tp)
+ $LD $acc7,$SIZE_T*15($rp)
+ $PUSH $t2,$SIZE_T*10($sp) # off-load top-most carry
+ addi $tp,$rp,$SIZE_T*7 # slide the window
+ mtctr $cnt
+ bne .Lsqr8x_reduction
+
+ ################################################################
+ # Final step. We see if result is larger than modulus, and
+ # if it is, subtract the modulus. But comparison implies
+ # subtraction. So we subtract modulus, see if it borrowed,
+ # and conditionally copy original value.
+ $POP $rp,$SIZE_T*6($sp) # pull &rp[-1]
+ srwi $cnt,$num,`log($SIZE_T)/log(2)+3`
+ mr $n0,$tp # put tp aside
+ addi $tp,$tp,$SIZE_T*8
+ subi $cnt,$cnt,1
+ subfc $t0,$a0,$acc0
+ subfe $t1,$a1,$acc1
+ mr $carry,$t2
+ mr $ap_end,$rp # $rp copy
+
+ mtctr $cnt
+ b .Lsqr8x_sub
+
+.align 5
+.Lsqr8x_sub:
+ $LD $a0,$SIZE_T*1($np)
+ $LD $acc0,$SIZE_T*1($tp)
+ $LD $a1,$SIZE_T*2($np)
+ $LD $acc1,$SIZE_T*2($tp)
+ subfe $t2,$a2,$acc2
+ $LD $a2,$SIZE_T*3($np)
+ $LD $acc2,$SIZE_T*3($tp)
+ subfe $t3,$a3,$acc3
+ $LD $a3,$SIZE_T*4($np)
+ $LD $acc3,$SIZE_T*4($tp)
+ $ST $t0,$SIZE_T*1($rp)
+ subfe $t0,$a4,$acc4
+ $LD $a4,$SIZE_T*5($np)
+ $LD $acc4,$SIZE_T*5($tp)
+ $ST $t1,$SIZE_T*2($rp)
+ subfe $t1,$a5,$acc5
+ $LD $a5,$SIZE_T*6($np)
+ $LD $acc5,$SIZE_T*6($tp)
+ $ST $t2,$SIZE_T*3($rp)
+ subfe $t2,$a6,$acc6
+ $LD $a6,$SIZE_T*7($np)
+ $LD $acc6,$SIZE_T*7($tp)
+ $ST $t3,$SIZE_T*4($rp)
+ subfe $t3,$a7,$acc7
+ $LDU $a7,$SIZE_T*8($np)
+ $LDU $acc7,$SIZE_T*8($tp)
+ $ST $t0,$SIZE_T*5($rp)
+ subfe $t0,$a0,$acc0
+ $ST $t1,$SIZE_T*6($rp)
+ subfe $t1,$a1,$acc1
+ $ST $t2,$SIZE_T*7($rp)
+ $STU $t3,$SIZE_T*8($rp)
+ bdnz .Lsqr8x_sub
+
+ srwi $cnt,$num,`log($SIZE_T)/log(2)+2`
+ $LD $a0,$SIZE_T*1($ap_end) # original $rp
+ $LD $acc0,$SIZE_T*1($n0) # original $tp
+ subi $cnt,$cnt,1
+ $LD $a1,$SIZE_T*2($ap_end)
+ $LD $acc1,$SIZE_T*2($n0)
+ subfe $t2,$a2,$acc2
+ $LD $a2,$SIZE_T*3($ap_end)
+ $LD $acc2,$SIZE_T*3($n0)
+ subfe $t3,$a3,$acc3
+ $LD $a3,$SIZE_T*4($ap_end)
+ $LDU $acc3,$SIZE_T*4($n0)
+ $ST $t0,$SIZE_T*1($rp)
+ subfe $t0,$a4,$acc4
+ $ST $t1,$SIZE_T*2($rp)
+ subfe $t1,$a5,$acc5
+ $ST $t2,$SIZE_T*3($rp)
+ subfe $t2,$a6,$acc6
+ $ST $t3,$SIZE_T*4($rp)
+ subfe $t3,$a7,$acc7
+ $ST $t0,$SIZE_T*5($rp)
+ subfe $carry,$zero,$carry # did it borrow?
+ $ST $t1,$SIZE_T*6($rp)
+ $ST $t2,$SIZE_T*7($rp)
+ $ST $t3,$SIZE_T*8($rp)
+
+ addi $tp,$sp,$SIZE_T*11
+ mtctr $cnt
+
+.Lsqr4x_cond_copy:
+ andc $a0,$a0,$carry
+ $ST $zero,-$SIZE_T*3($n0) # wipe stack clean
+ and $acc0,$acc0,$carry
+ $ST $zero,-$SIZE_T*2($n0)
+ andc $a1,$a1,$carry
+ $ST $zero,-$SIZE_T*1($n0)
+ and $acc1,$acc1,$carry
+ $ST $zero,-$SIZE_T*0($n0)
+ andc $a2,$a2,$carry
+ $ST $zero,$SIZE_T*1($tp)
+ and $acc2,$acc2,$carry
+ $ST $zero,$SIZE_T*2($tp)
+ andc $a3,$a3,$carry
+ $ST $zero,$SIZE_T*3($tp)
+ and $acc3,$acc3,$carry
+ $STU $zero,$SIZE_T*4($tp)
+ or $t0,$a0,$acc0
+ $LD $a0,$SIZE_T*5($ap_end)
+ $LD $acc0,$SIZE_T*1($n0)
+ or $t1,$a1,$acc1
+ $LD $a1,$SIZE_T*6($ap_end)
+ $LD $acc1,$SIZE_T*2($n0)
+ or $t2,$a2,$acc2
+ $LD $a2,$SIZE_T*7($ap_end)
+ $LD $acc2,$SIZE_T*3($n0)
+ or $t3,$a3,$acc3
+ $LD $a3,$SIZE_T*8($ap_end)
+ $LDU $acc3,$SIZE_T*4($n0)
+ $ST $t0,$SIZE_T*1($ap_end)
+ $ST $t1,$SIZE_T*2($ap_end)
+ $ST $t2,$SIZE_T*3($ap_end)
+ $STU $t3,$SIZE_T*4($ap_end)
+ bdnz .Lsqr4x_cond_copy
+
+ $POP $ap,0($sp) # pull saved sp
+ andc $a0,$a0,$carry
+ and $acc0,$acc0,$carry
+ andc $a1,$a1,$carry
+ and $acc1,$acc1,$carry
+ andc $a2,$a2,$carry
+ and $acc2,$acc2,$carry
+ andc $a3,$a3,$carry
+ and $acc3,$acc3,$carry
+ or $t0,$a0,$acc0
+ or $t1,$a1,$acc1
+ or $t2,$a2,$acc2
+ or $t3,$a3,$acc3
+ $ST $t0,$SIZE_T*1($ap_end)
+ $ST $t1,$SIZE_T*2($ap_end)
+ $ST $t2,$SIZE_T*3($ap_end)
+ $ST $t3,$SIZE_T*4($ap_end)
+
+ b .Lsqr8x_done
+
+.align 5
+.Lsqr8x8_post_condition:
+ $POP $rp,$SIZE_T*6($sp) # pull rp
+ $POP $ap,0($sp) # pull saved sp
+ addze $carry,$zero
+
+ # $acc0-7,$carry hold result, $a0-7 hold modulus
+ subfc $acc0,$a0,$acc0
+ subfe $acc1,$a1,$acc1
+ $ST $zero,$SIZE_T*12($sp) # wipe stack clean
+ $ST $zero,$SIZE_T*13($sp)
+ subfe $acc2,$a2,$acc2
+ $ST $zero,$SIZE_T*14($sp)
+ $ST $zero,$SIZE_T*15($sp)
+ subfe $acc3,$a3,$acc3
+ $ST $zero,$SIZE_T*16($sp)
+ $ST $zero,$SIZE_T*17($sp)
+ subfe $acc4,$a4,$acc4
+ $ST $zero,$SIZE_T*18($sp)
+ $ST $zero,$SIZE_T*19($sp)
+ subfe $acc5,$a5,$acc5
+ $ST $zero,$SIZE_T*20($sp)
+ $ST $zero,$SIZE_T*21($sp)
+ subfe $acc6,$a6,$acc6
+ $ST $zero,$SIZE_T*22($sp)
+ $ST $zero,$SIZE_T*23($sp)
+ subfe $acc7,$a7,$acc7
+ $ST $zero,$SIZE_T*24($sp)
+ $ST $zero,$SIZE_T*25($sp)
+ subfe $carry,$zero,$carry # did it borrow?
+ $ST $zero,$SIZE_T*26($sp)
+ $ST $zero,$SIZE_T*27($sp)
+
+ and $a0,$a0,$carry
+ and $a1,$a1,$carry
+ addc $acc0,$acc0,$a0 # add modulus back if borrowed
+ and $a2,$a2,$carry
+ adde $acc1,$acc1,$a1
+ and $a3,$a3,$carry
+ adde $acc2,$acc2,$a2
+ and $a4,$a4,$carry
+ adde $acc3,$acc3,$a3
+ and $a5,$a5,$carry
+ adde $acc4,$acc4,$a4
+ and $a6,$a6,$carry
+ adde $acc5,$acc5,$a5
+ and $a7,$a7,$carry
+ adde $acc6,$acc6,$a6
+ adde $acc7,$acc7,$a7
+ $ST $acc0,$SIZE_T*1($rp)
+ $ST $acc1,$SIZE_T*2($rp)
+ $ST $acc2,$SIZE_T*3($rp)
+ $ST $acc3,$SIZE_T*4($rp)
+ $ST $acc4,$SIZE_T*5($rp)
+ $ST $acc5,$SIZE_T*6($rp)
+ $ST $acc6,$SIZE_T*7($rp)
+ $ST $acc7,$SIZE_T*8($rp)
+
+.Lsqr8x_done:
+ $PUSH $zero,$SIZE_T*8($sp)
+ $PUSH $zero,$SIZE_T*10($sp)
+
+ $POP r14,-$SIZE_T*18($ap)
+ li r3,1 # signal "done"
+ $POP r15,-$SIZE_T*17($ap)
+ $POP r16,-$SIZE_T*16($ap)
+ $POP r17,-$SIZE_T*15($ap)
+ $POP r18,-$SIZE_T*14($ap)
+ $POP r19,-$SIZE_T*13($ap)
+ $POP r20,-$SIZE_T*12($ap)
+ $POP r21,-$SIZE_T*11($ap)
+ $POP r22,-$SIZE_T*10($ap)
+ $POP r23,-$SIZE_T*9($ap)
+ $POP r24,-$SIZE_T*8($ap)
+ $POP r25,-$SIZE_T*7($ap)
+ $POP r26,-$SIZE_T*6($ap)
+ $POP r27,-$SIZE_T*5($ap)
+ $POP r28,-$SIZE_T*4($ap)
+ $POP r29,-$SIZE_T*3($ap)
+ $POP r30,-$SIZE_T*2($ap)
+ $POP r31,-$SIZE_T*1($ap)
+ mr $sp,$ap
+ blr
+ .long 0
+ .byte 0,12,4,0x20,0x80,18,6,0
+ .long 0
+.size __bn_sqr8x_mont,.-__bn_sqr8x_mont
+___
+}
+$code.=<<___;
+.asciz "Montgomery Multiplication for PPC, CRYPTOGAMS by <appro\@openssl.org>"
+___
+
+$code =~ s/\`([^\`]*)\`/eval $1/gem;
+print $code;
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/ppc.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/ppc.pl
new file mode 100644
index 0000000..a8d3f14
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/ppc.pl
@@ -0,0 +1,2012 @@
+#! /usr/bin/env perl
+# Copyright 2004-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+# Implemented as a Perl wrapper as we want to support several different
+# architectures with single file. We pick up the target based on the
+# file name we are asked to generate.
+#
+# It should be noted though that this perl code is nothing like
+# <openssl>/crypto/perlasm/x86*. In this case perl is used pretty much
+# as pre-processor to cover for platform differences in name decoration,
+# linker tables, 32-/64-bit instruction sets...
+#
+# As you might know there're several PowerPC ABI in use. Most notably
+# Linux and AIX use different 32-bit ABIs. Good news are that these ABIs
+# are similar enough to implement leaf(!) functions, which would be ABI
+# neutral. And that's what you find here: ABI neutral leaf functions.
+# In case you wonder what that is...
+#
+# AIX performance
+#
+# MEASUREMENTS WITH cc ON a 200 MhZ PowerPC 604e.
+#
+# The following is the performance of 32-bit compiler
+# generated code:
+#
+# OpenSSL 0.9.6c 21 dec 2001
+# built on: Tue Jun 11 11:06:51 EDT 2002
+# options:bn(64,32) ...
+#compiler: cc -DTHREADS -DAIX -DB_ENDIAN -DBN_LLONG -O3
+# sign verify sign/s verify/s
+#rsa 512 bits 0.0098s 0.0009s 102.0 1170.6
+#rsa 1024 bits 0.0507s 0.0026s 19.7 387.5
+#rsa 2048 bits 0.3036s 0.0085s 3.3 117.1
+#rsa 4096 bits 2.0040s 0.0299s 0.5 33.4
+#dsa 512 bits 0.0087s 0.0106s 114.3 94.5
+#dsa 1024 bits 0.0256s 0.0313s 39.0 32.0
+#
+# Same benchmark with this assembler code:
+#
+#rsa 512 bits 0.0056s 0.0005s 178.6 2049.2
+#rsa 1024 bits 0.0283s 0.0015s 35.3 674.1
+#rsa 2048 bits 0.1744s 0.0050s 5.7 201.2
+#rsa 4096 bits 1.1644s 0.0179s 0.9 55.7
+#dsa 512 bits 0.0052s 0.0062s 191.6 162.0
+#dsa 1024 bits 0.0149s 0.0180s 67.0 55.5
+#
+# Number of operations increases by at almost 75%
+#
+# Here are performance numbers for 64-bit compiler
+# generated code:
+#
+# OpenSSL 0.9.6g [engine] 9 Aug 2002
+# built on: Fri Apr 18 16:59:20 EDT 2003
+# options:bn(64,64) ...
+# compiler: cc -DTHREADS -D_REENTRANT -q64 -DB_ENDIAN -O3
+# sign verify sign/s verify/s
+#rsa 512 bits 0.0028s 0.0003s 357.1 3844.4
+#rsa 1024 bits 0.0148s 0.0008s 67.5 1239.7
+#rsa 2048 bits 0.0963s 0.0028s 10.4 353.0
+#rsa 4096 bits 0.6538s 0.0102s 1.5 98.1
+#dsa 512 bits 0.0026s 0.0032s 382.5 313.7
+#dsa 1024 bits 0.0081s 0.0099s 122.8 100.6
+#
+# Same benchmark with this assembler code:
+#
+#rsa 512 bits 0.0020s 0.0002s 510.4 6273.7
+#rsa 1024 bits 0.0088s 0.0005s 114.1 2128.3
+#rsa 2048 bits 0.0540s 0.0016s 18.5 622.5
+#rsa 4096 bits 0.3700s 0.0058s 2.7 171.0
+#dsa 512 bits 0.0016s 0.0020s 610.7 507.1
+#dsa 1024 bits 0.0047s 0.0058s 212.5 173.2
+#
+# Again, performance increases by at about 75%
+#
+# Mac OS X, Apple G5 1.8GHz (Note this is 32 bit code)
+# OpenSSL 0.9.7c 30 Sep 2003
+#
+# Original code.
+#
+#rsa 512 bits 0.0011s 0.0001s 906.1 11012.5
+#rsa 1024 bits 0.0060s 0.0003s 166.6 3363.1
+#rsa 2048 bits 0.0370s 0.0010s 27.1 982.4
+#rsa 4096 bits 0.2426s 0.0036s 4.1 280.4
+#dsa 512 bits 0.0010s 0.0012s 1038.1 841.5
+#dsa 1024 bits 0.0030s 0.0037s 329.6 269.7
+#dsa 2048 bits 0.0101s 0.0127s 98.9 78.6
+#
+# Same benchmark with this assembler code:
+#
+#rsa 512 bits 0.0007s 0.0001s 1416.2 16645.9
+#rsa 1024 bits 0.0036s 0.0002s 274.4 5380.6
+#rsa 2048 bits 0.0222s 0.0006s 45.1 1589.5
+#rsa 4096 bits 0.1469s 0.0022s 6.8 449.6
+#dsa 512 bits 0.0006s 0.0007s 1664.2 1376.2
+#dsa 1024 bits 0.0018s 0.0023s 545.0 442.2
+#dsa 2048 bits 0.0061s 0.0075s 163.5 132.8
+#
+# Performance increase of ~60%
+# Based on submission from Suresh N. Chari of IBM
+
+$flavour = shift;
+
+if ($flavour =~ /32/) {
+ $BITS= 32;
+ $BNSZ= $BITS/8;
+ $ISA= "\"ppc\"";
+
+ $LD= "lwz"; # load
+ $LDU= "lwzu"; # load and update
+ $ST= "stw"; # store
+ $STU= "stwu"; # store and update
+ $UMULL= "mullw"; # unsigned multiply low
+ $UMULH= "mulhwu"; # unsigned multiply high
+ $UDIV= "divwu"; # unsigned divide
+ $UCMPI= "cmplwi"; # unsigned compare with immediate
+ $UCMP= "cmplw"; # unsigned compare
+ $CNTLZ= "cntlzw"; # count leading zeros
+ $SHL= "slw"; # shift left
+ $SHR= "srw"; # unsigned shift right
+ $SHRI= "srwi"; # unsigned shift right by immediate
+ $SHLI= "slwi"; # shift left by immediate
+ $CLRU= "clrlwi"; # clear upper bits
+ $INSR= "insrwi"; # insert right
+ $ROTL= "rotlwi"; # rotate left by immediate
+ $TR= "tw"; # conditional trap
+} elsif ($flavour =~ /64/) {
+ $BITS= 64;
+ $BNSZ= $BITS/8;
+ $ISA= "\"ppc64\"";
+
+ # same as above, but 64-bit mnemonics...
+ $LD= "ld"; # load
+ $LDU= "ldu"; # load and update
+ $ST= "std"; # store
+ $STU= "stdu"; # store and update
+ $UMULL= "mulld"; # unsigned multiply low
+ $UMULH= "mulhdu"; # unsigned multiply high
+ $UDIV= "divdu"; # unsigned divide
+ $UCMPI= "cmpldi"; # unsigned compare with immediate
+ $UCMP= "cmpld"; # unsigned compare
+ $CNTLZ= "cntlzd"; # count leading zeros
+ $SHL= "sld"; # shift left
+ $SHR= "srd"; # unsigned shift right
+ $SHRI= "srdi"; # unsigned shift right by immediate
+ $SHLI= "sldi"; # shift left by immediate
+ $CLRU= "clrldi"; # clear upper bits
+ $INSR= "insrdi"; # insert right
+ $ROTL= "rotldi"; # rotate left by immediate
+ $TR= "td"; # conditional trap
+} else { die "nonsense $flavour"; }
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or
+( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or
+die "can't locate ppc-xlate.pl";
+
+open STDOUT,"| $^X $xlate $flavour ".shift || die "can't call $xlate: $!";
+
+$data=<<EOF;
+#--------------------------------------------------------------------
+#
+#
+#
+#
+# File: ppc32.s
+#
+# Created by: Suresh Chari
+# IBM Thomas J. Watson Research Library
+# Hawthorne, NY
+#
+#
+# Description: Optimized assembly routines for OpenSSL crypto
+# on the 32 bitPowerPC platform.
+#
+#
+# Version History
+#
+# 2. Fixed bn_add,bn_sub and bn_div_words, added comments,
+# cleaned up code. Also made a single version which can
+# be used for both the AIX and Linux compilers. See NOTE
+# below.
+# 12/05/03 Suresh Chari
+# (with lots of help from) Andy Polyakov
+##
+# 1. Initial version 10/20/02 Suresh Chari
+#
+#
+# The following file works for the xlc,cc
+# and gcc compilers.
+#
+# NOTE: To get the file to link correctly with the gcc compiler
+# you have to change the names of the routines and remove
+# the first .(dot) character. This should automatically
+# be done in the build process.
+#
+# Hand optimized assembly code for the following routines
+#
+# bn_sqr_comba4
+# bn_sqr_comba8
+# bn_mul_comba4
+# bn_mul_comba8
+# bn_sub_words
+# bn_add_words
+# bn_div_words
+# bn_sqr_words
+# bn_mul_words
+# bn_mul_add_words
+#
+# NOTE: It is possible to optimize this code more for
+# specific PowerPC or Power architectures. On the Northstar
+# architecture the optimizations in this file do
+# NOT provide much improvement.
+#
+# If you have comments or suggestions to improve code send
+# me a note at schari\@us.ibm.com
+#
+#--------------------------------------------------------------------------
+#
+# Defines to be used in the assembly code.
+#
+#.set r0,0 # we use it as storage for value of 0
+#.set SP,1 # preserved
+#.set RTOC,2 # preserved
+#.set r3,3 # 1st argument/return value
+#.set r4,4 # 2nd argument/volatile register
+#.set r5,5 # 3rd argument/volatile register
+#.set r6,6 # ...
+#.set r7,7
+#.set r8,8
+#.set r9,9
+#.set r10,10
+#.set r11,11
+#.set r12,12
+#.set r13,13 # not used, nor any other "below" it...
+
+# Declare function names to be global
+# NOTE: For gcc these names MUST be changed to remove
+# the first . i.e. for example change ".bn_sqr_comba4"
+# to "bn_sqr_comba4". This should be automatically done
+# in the build.
+
+ .globl .bn_sqr_comba4
+ .globl .bn_sqr_comba8
+ .globl .bn_mul_comba4
+ .globl .bn_mul_comba8
+ .globl .bn_sub_words
+ .globl .bn_add_words
+ .globl .bn_div_words
+ .globl .bn_sqr_words
+ .globl .bn_mul_words
+ .globl .bn_mul_add_words
+
+# .text section
+
+ .machine "any"
+ .text
+
+#
+# NOTE: The following label name should be changed to
+# "bn_sqr_comba4" i.e. remove the first dot
+# for the gcc compiler. This should be automatically
+# done in the build
+#
+
+.align 4
+.bn_sqr_comba4:
+#
+# Optimized version of bn_sqr_comba4.
+#
+# void bn_sqr_comba4(BN_ULONG *r, BN_ULONG *a)
+# r3 contains r
+# r4 contains a
+#
+# Freely use registers r5,r6,r7,r8,r9,r10,r11 as follows:
+#
+# r5,r6 are the two BN_ULONGs being multiplied.
+# r7,r8 are the results of the 32x32 giving 64 bit multiply.
+# r9,r10, r11 are the equivalents of c1,c2, c3.
+# Here's the assembly
+#
+#
+ xor r0,r0,r0 # set r0 = 0. Used in the addze
+ # instructions below
+
+ #sqr_add_c(a,0,c1,c2,c3)
+ $LD r5,`0*$BNSZ`(r4)
+ $UMULL r9,r5,r5
+ $UMULH r10,r5,r5 #in first iteration. No need
+ #to add since c1=c2=c3=0.
+ # Note c3(r11) is NOT set to 0
+ # but will be.
+
+ $ST r9,`0*$BNSZ`(r3) # r[0]=c1;
+ # sqr_add_c2(a,1,0,c2,c3,c1);
+ $LD r6,`1*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+
+ addc r7,r7,r7 # compute (r7,r8)=2*(r7,r8)
+ adde r8,r8,r8
+ addze r9,r0 # catch carry if any.
+ # r9= r0(=0) and carry
+
+ addc r10,r7,r10 # now add to temp result.
+ addze r11,r8 # r8 added to r11 which is 0
+ addze r9,r9
+
+ $ST r10,`1*$BNSZ`(r3) #r[1]=c2;
+ #sqr_add_c(a,1,c3,c1,c2)
+ $UMULL r7,r6,r6
+ $UMULH r8,r6,r6
+ addc r11,r7,r11
+ adde r9,r8,r9
+ addze r10,r0
+ #sqr_add_c2(a,2,0,c3,c1,c2)
+ $LD r6,`2*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+
+ addc r7,r7,r7
+ adde r8,r8,r8
+ addze r10,r10
+
+ addc r11,r7,r11
+ adde r9,r8,r9
+ addze r10,r10
+ $ST r11,`2*$BNSZ`(r3) #r[2]=c3
+ #sqr_add_c2(a,3,0,c1,c2,c3);
+ $LD r6,`3*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+ addc r7,r7,r7
+ adde r8,r8,r8
+ addze r11,r0
+
+ addc r9,r7,r9
+ adde r10,r8,r10
+ addze r11,r11
+ #sqr_add_c2(a,2,1,c1,c2,c3);
+ $LD r5,`1*$BNSZ`(r4)
+ $LD r6,`2*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+
+ addc r7,r7,r7
+ adde r8,r8,r8
+ addze r11,r11
+ addc r9,r7,r9
+ adde r10,r8,r10
+ addze r11,r11
+ $ST r9,`3*$BNSZ`(r3) #r[3]=c1
+ #sqr_add_c(a,2,c2,c3,c1);
+ $UMULL r7,r6,r6
+ $UMULH r8,r6,r6
+ addc r10,r7,r10
+ adde r11,r8,r11
+ addze r9,r0
+ #sqr_add_c2(a,3,1,c2,c3,c1);
+ $LD r6,`3*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+ addc r7,r7,r7
+ adde r8,r8,r8
+ addze r9,r9
+
+ addc r10,r7,r10
+ adde r11,r8,r11
+ addze r9,r9
+ $ST r10,`4*$BNSZ`(r3) #r[4]=c2
+ #sqr_add_c2(a,3,2,c3,c1,c2);
+ $LD r5,`2*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+ addc r7,r7,r7
+ adde r8,r8,r8
+ addze r10,r0
+
+ addc r11,r7,r11
+ adde r9,r8,r9
+ addze r10,r10
+ $ST r11,`5*$BNSZ`(r3) #r[5] = c3
+ #sqr_add_c(a,3,c1,c2,c3);
+ $UMULL r7,r6,r6
+ $UMULH r8,r6,r6
+ addc r9,r7,r9
+ adde r10,r8,r10
+
+ $ST r9,`6*$BNSZ`(r3) #r[6]=c1
+ $ST r10,`7*$BNSZ`(r3) #r[7]=c2
+ blr
+ .long 0
+ .byte 0,12,0x14,0,0,0,2,0
+ .long 0
+.size .bn_sqr_comba4,.-.bn_sqr_comba4
+
+#
+# NOTE: The following label name should be changed to
+# "bn_sqr_comba8" i.e. remove the first dot
+# for the gcc compiler. This should be automatically
+# done in the build
+#
+
+.align 4
+.bn_sqr_comba8:
+#
+# This is an optimized version of the bn_sqr_comba8 routine.
+# Tightly uses the adde instruction
+#
+#
+# void bn_sqr_comba8(BN_ULONG *r, BN_ULONG *a)
+# r3 contains r
+# r4 contains a
+#
+# Freely use registers r5,r6,r7,r8,r9,r10,r11 as follows:
+#
+# r5,r6 are the two BN_ULONGs being multiplied.
+# r7,r8 are the results of the 32x32 giving 64 bit multiply.
+# r9,r10, r11 are the equivalents of c1,c2, c3.
+#
+# Possible optimization of loading all 8 longs of a into registers
+# doesn't provide any speedup
+#
+
+ xor r0,r0,r0 #set r0 = 0.Used in addze
+ #instructions below.
+
+ #sqr_add_c(a,0,c1,c2,c3);
+ $LD r5,`0*$BNSZ`(r4)
+ $UMULL r9,r5,r5 #1st iteration: no carries.
+ $UMULH r10,r5,r5
+ $ST r9,`0*$BNSZ`(r3) # r[0]=c1;
+ #sqr_add_c2(a,1,0,c2,c3,c1);
+ $LD r6,`1*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+
+ addc r10,r7,r10 #add the two register number
+ adde r11,r8,r0 # (r8,r7) to the three register
+ addze r9,r0 # number (r9,r11,r10).NOTE:r0=0
+
+ addc r10,r7,r10 #add the two register number
+ adde r11,r8,r11 # (r8,r7) to the three register
+ addze r9,r9 # number (r9,r11,r10).
+
+ $ST r10,`1*$BNSZ`(r3) # r[1]=c2
+
+ #sqr_add_c(a,1,c3,c1,c2);
+ $UMULL r7,r6,r6
+ $UMULH r8,r6,r6
+ addc r11,r7,r11
+ adde r9,r8,r9
+ addze r10,r0
+ #sqr_add_c2(a,2,0,c3,c1,c2);
+ $LD r6,`2*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+
+ addc r11,r7,r11
+ adde r9,r8,r9
+ addze r10,r10
+
+ addc r11,r7,r11
+ adde r9,r8,r9
+ addze r10,r10
+
+ $ST r11,`2*$BNSZ`(r3) #r[2]=c3
+ #sqr_add_c2(a,3,0,c1,c2,c3);
+ $LD r6,`3*$BNSZ`(r4) #r6 = a[3]. r5 is already a[0].
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+
+ addc r9,r7,r9
+ adde r10,r8,r10
+ addze r11,r0
+
+ addc r9,r7,r9
+ adde r10,r8,r10
+ addze r11,r11
+ #sqr_add_c2(a,2,1,c1,c2,c3);
+ $LD r5,`1*$BNSZ`(r4)
+ $LD r6,`2*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+
+ addc r9,r7,r9
+ adde r10,r8,r10
+ addze r11,r11
+
+ addc r9,r7,r9
+ adde r10,r8,r10
+ addze r11,r11
+
+ $ST r9,`3*$BNSZ`(r3) #r[3]=c1;
+ #sqr_add_c(a,2,c2,c3,c1);
+ $UMULL r7,r6,r6
+ $UMULH r8,r6,r6
+
+ addc r10,r7,r10
+ adde r11,r8,r11
+ addze r9,r0
+ #sqr_add_c2(a,3,1,c2,c3,c1);
+ $LD r6,`3*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+
+ addc r10,r7,r10
+ adde r11,r8,r11
+ addze r9,r9
+
+ addc r10,r7,r10
+ adde r11,r8,r11
+ addze r9,r9
+ #sqr_add_c2(a,4,0,c2,c3,c1);
+ $LD r5,`0*$BNSZ`(r4)
+ $LD r6,`4*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+
+ addc r10,r7,r10
+ adde r11,r8,r11
+ addze r9,r9
+
+ addc r10,r7,r10
+ adde r11,r8,r11
+ addze r9,r9
+ $ST r10,`4*$BNSZ`(r3) #r[4]=c2;
+ #sqr_add_c2(a,5,0,c3,c1,c2);
+ $LD r6,`5*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+
+ addc r11,r7,r11
+ adde r9,r8,r9
+ addze r10,r0
+
+ addc r11,r7,r11
+ adde r9,r8,r9
+ addze r10,r10
+ #sqr_add_c2(a,4,1,c3,c1,c2);
+ $LD r5,`1*$BNSZ`(r4)
+ $LD r6,`4*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+
+ addc r11,r7,r11
+ adde r9,r8,r9
+ addze r10,r10
+
+ addc r11,r7,r11
+ adde r9,r8,r9
+ addze r10,r10
+ #sqr_add_c2(a,3,2,c3,c1,c2);
+ $LD r5,`2*$BNSZ`(r4)
+ $LD r6,`3*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+
+ addc r11,r7,r11
+ adde r9,r8,r9
+ addze r10,r10
+
+ addc r11,r7,r11
+ adde r9,r8,r9
+ addze r10,r10
+ $ST r11,`5*$BNSZ`(r3) #r[5]=c3;
+ #sqr_add_c(a,3,c1,c2,c3);
+ $UMULL r7,r6,r6
+ $UMULH r8,r6,r6
+ addc r9,r7,r9
+ adde r10,r8,r10
+ addze r11,r0
+ #sqr_add_c2(a,4,2,c1,c2,c3);
+ $LD r6,`4*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+
+ addc r9,r7,r9
+ adde r10,r8,r10
+ addze r11,r11
+
+ addc r9,r7,r9
+ adde r10,r8,r10
+ addze r11,r11
+ #sqr_add_c2(a,5,1,c1,c2,c3);
+ $LD r5,`1*$BNSZ`(r4)
+ $LD r6,`5*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+
+ addc r9,r7,r9
+ adde r10,r8,r10
+ addze r11,r11
+
+ addc r9,r7,r9
+ adde r10,r8,r10
+ addze r11,r11
+ #sqr_add_c2(a,6,0,c1,c2,c3);
+ $LD r5,`0*$BNSZ`(r4)
+ $LD r6,`6*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+ addc r9,r7,r9
+ adde r10,r8,r10
+ addze r11,r11
+ addc r9,r7,r9
+ adde r10,r8,r10
+ addze r11,r11
+ $ST r9,`6*$BNSZ`(r3) #r[6]=c1;
+ #sqr_add_c2(a,7,0,c2,c3,c1);
+ $LD r6,`7*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+
+ addc r10,r7,r10
+ adde r11,r8,r11
+ addze r9,r0
+ addc r10,r7,r10
+ adde r11,r8,r11
+ addze r9,r9
+ #sqr_add_c2(a,6,1,c2,c3,c1);
+ $LD r5,`1*$BNSZ`(r4)
+ $LD r6,`6*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+
+ addc r10,r7,r10
+ adde r11,r8,r11
+ addze r9,r9
+ addc r10,r7,r10
+ adde r11,r8,r11
+ addze r9,r9
+ #sqr_add_c2(a,5,2,c2,c3,c1);
+ $LD r5,`2*$BNSZ`(r4)
+ $LD r6,`5*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+ addc r10,r7,r10
+ adde r11,r8,r11
+ addze r9,r9
+ addc r10,r7,r10
+ adde r11,r8,r11
+ addze r9,r9
+ #sqr_add_c2(a,4,3,c2,c3,c1);
+ $LD r5,`3*$BNSZ`(r4)
+ $LD r6,`4*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+
+ addc r10,r7,r10
+ adde r11,r8,r11
+ addze r9,r9
+ addc r10,r7,r10
+ adde r11,r8,r11
+ addze r9,r9
+ $ST r10,`7*$BNSZ`(r3) #r[7]=c2;
+ #sqr_add_c(a,4,c3,c1,c2);
+ $UMULL r7,r6,r6
+ $UMULH r8,r6,r6
+ addc r11,r7,r11
+ adde r9,r8,r9
+ addze r10,r0
+ #sqr_add_c2(a,5,3,c3,c1,c2);
+ $LD r6,`5*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+ addc r11,r7,r11
+ adde r9,r8,r9
+ addze r10,r10
+ addc r11,r7,r11
+ adde r9,r8,r9
+ addze r10,r10
+ #sqr_add_c2(a,6,2,c3,c1,c2);
+ $LD r5,`2*$BNSZ`(r4)
+ $LD r6,`6*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+ addc r11,r7,r11
+ adde r9,r8,r9
+ addze r10,r10
+
+ addc r11,r7,r11
+ adde r9,r8,r9
+ addze r10,r10
+ #sqr_add_c2(a,7,1,c3,c1,c2);
+ $LD r5,`1*$BNSZ`(r4)
+ $LD r6,`7*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+ addc r11,r7,r11
+ adde r9,r8,r9
+ addze r10,r10
+ addc r11,r7,r11
+ adde r9,r8,r9
+ addze r10,r10
+ $ST r11,`8*$BNSZ`(r3) #r[8]=c3;
+ #sqr_add_c2(a,7,2,c1,c2,c3);
+ $LD r5,`2*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+
+ addc r9,r7,r9
+ adde r10,r8,r10
+ addze r11,r0
+ addc r9,r7,r9
+ adde r10,r8,r10
+ addze r11,r11
+ #sqr_add_c2(a,6,3,c1,c2,c3);
+ $LD r5,`3*$BNSZ`(r4)
+ $LD r6,`6*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+ addc r9,r7,r9
+ adde r10,r8,r10
+ addze r11,r11
+ addc r9,r7,r9
+ adde r10,r8,r10
+ addze r11,r11
+ #sqr_add_c2(a,5,4,c1,c2,c3);
+ $LD r5,`4*$BNSZ`(r4)
+ $LD r6,`5*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+ addc r9,r7,r9
+ adde r10,r8,r10
+ addze r11,r11
+ addc r9,r7,r9
+ adde r10,r8,r10
+ addze r11,r11
+ $ST r9,`9*$BNSZ`(r3) #r[9]=c1;
+ #sqr_add_c(a,5,c2,c3,c1);
+ $UMULL r7,r6,r6
+ $UMULH r8,r6,r6
+ addc r10,r7,r10
+ adde r11,r8,r11
+ addze r9,r0
+ #sqr_add_c2(a,6,4,c2,c3,c1);
+ $LD r6,`6*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+ addc r10,r7,r10
+ adde r11,r8,r11
+ addze r9,r9
+ addc r10,r7,r10
+ adde r11,r8,r11
+ addze r9,r9
+ #sqr_add_c2(a,7,3,c2,c3,c1);
+ $LD r5,`3*$BNSZ`(r4)
+ $LD r6,`7*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+ addc r10,r7,r10
+ adde r11,r8,r11
+ addze r9,r9
+ addc r10,r7,r10
+ adde r11,r8,r11
+ addze r9,r9
+ $ST r10,`10*$BNSZ`(r3) #r[10]=c2;
+ #sqr_add_c2(a,7,4,c3,c1,c2);
+ $LD r5,`4*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+ addc r11,r7,r11
+ adde r9,r8,r9
+ addze r10,r0
+ addc r11,r7,r11
+ adde r9,r8,r9
+ addze r10,r10
+ #sqr_add_c2(a,6,5,c3,c1,c2);
+ $LD r5,`5*$BNSZ`(r4)
+ $LD r6,`6*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+ addc r11,r7,r11
+ adde r9,r8,r9
+ addze r10,r10
+ addc r11,r7,r11
+ adde r9,r8,r9
+ addze r10,r10
+ $ST r11,`11*$BNSZ`(r3) #r[11]=c3;
+ #sqr_add_c(a,6,c1,c2,c3);
+ $UMULL r7,r6,r6
+ $UMULH r8,r6,r6
+ addc r9,r7,r9
+ adde r10,r8,r10
+ addze r11,r0
+ #sqr_add_c2(a,7,5,c1,c2,c3)
+ $LD r6,`7*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+ addc r9,r7,r9
+ adde r10,r8,r10
+ addze r11,r11
+ addc r9,r7,r9
+ adde r10,r8,r10
+ addze r11,r11
+ $ST r9,`12*$BNSZ`(r3) #r[12]=c1;
+
+ #sqr_add_c2(a,7,6,c2,c3,c1)
+ $LD r5,`6*$BNSZ`(r4)
+ $UMULL r7,r5,r6
+ $UMULH r8,r5,r6
+ addc r10,r7,r10
+ adde r11,r8,r11
+ addze r9,r0
+ addc r10,r7,r10
+ adde r11,r8,r11
+ addze r9,r9
+ $ST r10,`13*$BNSZ`(r3) #r[13]=c2;
+ #sqr_add_c(a,7,c3,c1,c2);
+ $UMULL r7,r6,r6
+ $UMULH r8,r6,r6
+ addc r11,r7,r11
+ adde r9,r8,r9
+ $ST r11,`14*$BNSZ`(r3) #r[14]=c3;
+ $ST r9, `15*$BNSZ`(r3) #r[15]=c1;
+
+
+ blr
+ .long 0
+ .byte 0,12,0x14,0,0,0,2,0
+ .long 0
+.size .bn_sqr_comba8,.-.bn_sqr_comba8
+
+#
+# NOTE: The following label name should be changed to
+# "bn_mul_comba4" i.e. remove the first dot
+# for the gcc compiler. This should be automatically
+# done in the build
+#
+
+.align 4
+.bn_mul_comba4:
+#
+# This is an optimized version of the bn_mul_comba4 routine.
+#
+# void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
+# r3 contains r
+# r4 contains a
+# r5 contains b
+# r6, r7 are the 2 BN_ULONGs being multiplied.
+# r8, r9 are the results of the 32x32 giving 64 multiply.
+# r10, r11, r12 are the equivalents of c1, c2, and c3.
+#
+ xor r0,r0,r0 #r0=0. Used in addze below.
+ #mul_add_c(a[0],b[0],c1,c2,c3);
+ $LD r6,`0*$BNSZ`(r4)
+ $LD r7,`0*$BNSZ`(r5)
+ $UMULL r10,r6,r7
+ $UMULH r11,r6,r7
+ $ST r10,`0*$BNSZ`(r3) #r[0]=c1
+ #mul_add_c(a[0],b[1],c2,c3,c1);
+ $LD r7,`1*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r8,r11
+ adde r12,r9,r0
+ addze r10,r0
+ #mul_add_c(a[1],b[0],c2,c3,c1);
+ $LD r6, `1*$BNSZ`(r4)
+ $LD r7, `0*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r8,r11
+ adde r12,r9,r12
+ addze r10,r10
+ $ST r11,`1*$BNSZ`(r3) #r[1]=c2
+ #mul_add_c(a[2],b[0],c3,c1,c2);
+ $LD r6,`2*$BNSZ`(r4)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r8,r12
+ adde r10,r9,r10
+ addze r11,r0
+ #mul_add_c(a[1],b[1],c3,c1,c2);
+ $LD r6,`1*$BNSZ`(r4)
+ $LD r7,`1*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r8,r12
+ adde r10,r9,r10
+ addze r11,r11
+ #mul_add_c(a[0],b[2],c3,c1,c2);
+ $LD r6,`0*$BNSZ`(r4)
+ $LD r7,`2*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r8,r12
+ adde r10,r9,r10
+ addze r11,r11
+ $ST r12,`2*$BNSZ`(r3) #r[2]=c3
+ #mul_add_c(a[0],b[3],c1,c2,c3);
+ $LD r7,`3*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r8,r10
+ adde r11,r9,r11
+ addze r12,r0
+ #mul_add_c(a[1],b[2],c1,c2,c3);
+ $LD r6,`1*$BNSZ`(r4)
+ $LD r7,`2*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r8,r10
+ adde r11,r9,r11
+ addze r12,r12
+ #mul_add_c(a[2],b[1],c1,c2,c3);
+ $LD r6,`2*$BNSZ`(r4)
+ $LD r7,`1*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r8,r10
+ adde r11,r9,r11
+ addze r12,r12
+ #mul_add_c(a[3],b[0],c1,c2,c3);
+ $LD r6,`3*$BNSZ`(r4)
+ $LD r7,`0*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r8,r10
+ adde r11,r9,r11
+ addze r12,r12
+ $ST r10,`3*$BNSZ`(r3) #r[3]=c1
+ #mul_add_c(a[3],b[1],c2,c3,c1);
+ $LD r7,`1*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r8,r11
+ adde r12,r9,r12
+ addze r10,r0
+ #mul_add_c(a[2],b[2],c2,c3,c1);
+ $LD r6,`2*$BNSZ`(r4)
+ $LD r7,`2*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r8,r11
+ adde r12,r9,r12
+ addze r10,r10
+ #mul_add_c(a[1],b[3],c2,c3,c1);
+ $LD r6,`1*$BNSZ`(r4)
+ $LD r7,`3*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r8,r11
+ adde r12,r9,r12
+ addze r10,r10
+ $ST r11,`4*$BNSZ`(r3) #r[4]=c2
+ #mul_add_c(a[2],b[3],c3,c1,c2);
+ $LD r6,`2*$BNSZ`(r4)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r8,r12
+ adde r10,r9,r10
+ addze r11,r0
+ #mul_add_c(a[3],b[2],c3,c1,c2);
+ $LD r6,`3*$BNSZ`(r4)
+ $LD r7,`2*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r8,r12
+ adde r10,r9,r10
+ addze r11,r11
+ $ST r12,`5*$BNSZ`(r3) #r[5]=c3
+ #mul_add_c(a[3],b[3],c1,c2,c3);
+ $LD r7,`3*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r8,r10
+ adde r11,r9,r11
+
+ $ST r10,`6*$BNSZ`(r3) #r[6]=c1
+ $ST r11,`7*$BNSZ`(r3) #r[7]=c2
+ blr
+ .long 0
+ .byte 0,12,0x14,0,0,0,3,0
+ .long 0
+.size .bn_mul_comba4,.-.bn_mul_comba4
+
+#
+# NOTE: The following label name should be changed to
+# "bn_mul_comba8" i.e. remove the first dot
+# for the gcc compiler. This should be automatically
+# done in the build
+#
+
+.align 4
+.bn_mul_comba8:
+#
+# Optimized version of the bn_mul_comba8 routine.
+#
+# void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
+# r3 contains r
+# r4 contains a
+# r5 contains b
+# r6, r7 are the 2 BN_ULONGs being multiplied.
+# r8, r9 are the results of the 32x32 giving 64 multiply.
+# r10, r11, r12 are the equivalents of c1, c2, and c3.
+#
+ xor r0,r0,r0 #r0=0. Used in addze below.
+
+ #mul_add_c(a[0],b[0],c1,c2,c3);
+ $LD r6,`0*$BNSZ`(r4) #a[0]
+ $LD r7,`0*$BNSZ`(r5) #b[0]
+ $UMULL r10,r6,r7
+ $UMULH r11,r6,r7
+ $ST r10,`0*$BNSZ`(r3) #r[0]=c1;
+ #mul_add_c(a[0],b[1],c2,c3,c1);
+ $LD r7,`1*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r11,r8
+ addze r12,r9 # since we didn't set r12 to zero before.
+ addze r10,r0
+ #mul_add_c(a[1],b[0],c2,c3,c1);
+ $LD r6,`1*$BNSZ`(r4)
+ $LD r7,`0*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r11,r8
+ adde r12,r12,r9
+ addze r10,r10
+ $ST r11,`1*$BNSZ`(r3) #r[1]=c2;
+ #mul_add_c(a[2],b[0],c3,c1,c2);
+ $LD r6,`2*$BNSZ`(r4)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r12,r8
+ adde r10,r10,r9
+ addze r11,r0
+ #mul_add_c(a[1],b[1],c3,c1,c2);
+ $LD r6,`1*$BNSZ`(r4)
+ $LD r7,`1*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r12,r8
+ adde r10,r10,r9
+ addze r11,r11
+ #mul_add_c(a[0],b[2],c3,c1,c2);
+ $LD r6,`0*$BNSZ`(r4)
+ $LD r7,`2*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r12,r8
+ adde r10,r10,r9
+ addze r11,r11
+ $ST r12,`2*$BNSZ`(r3) #r[2]=c3;
+ #mul_add_c(a[0],b[3],c1,c2,c3);
+ $LD r7,`3*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r10,r8
+ adde r11,r11,r9
+ addze r12,r0
+ #mul_add_c(a[1],b[2],c1,c2,c3);
+ $LD r6,`1*$BNSZ`(r4)
+ $LD r7,`2*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r10,r8
+ adde r11,r11,r9
+ addze r12,r12
+
+ #mul_add_c(a[2],b[1],c1,c2,c3);
+ $LD r6,`2*$BNSZ`(r4)
+ $LD r7,`1*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r10,r8
+ adde r11,r11,r9
+ addze r12,r12
+ #mul_add_c(a[3],b[0],c1,c2,c3);
+ $LD r6,`3*$BNSZ`(r4)
+ $LD r7,`0*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r10,r8
+ adde r11,r11,r9
+ addze r12,r12
+ $ST r10,`3*$BNSZ`(r3) #r[3]=c1;
+ #mul_add_c(a[4],b[0],c2,c3,c1);
+ $LD r6,`4*$BNSZ`(r4)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r11,r8
+ adde r12,r12,r9
+ addze r10,r0
+ #mul_add_c(a[3],b[1],c2,c3,c1);
+ $LD r6,`3*$BNSZ`(r4)
+ $LD r7,`1*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r11,r8
+ adde r12,r12,r9
+ addze r10,r10
+ #mul_add_c(a[2],b[2],c2,c3,c1);
+ $LD r6,`2*$BNSZ`(r4)
+ $LD r7,`2*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r11,r8
+ adde r12,r12,r9
+ addze r10,r10
+ #mul_add_c(a[1],b[3],c2,c3,c1);
+ $LD r6,`1*$BNSZ`(r4)
+ $LD r7,`3*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r11,r8
+ adde r12,r12,r9
+ addze r10,r10
+ #mul_add_c(a[0],b[4],c2,c3,c1);
+ $LD r6,`0*$BNSZ`(r4)
+ $LD r7,`4*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r11,r8
+ adde r12,r12,r9
+ addze r10,r10
+ $ST r11,`4*$BNSZ`(r3) #r[4]=c2;
+ #mul_add_c(a[0],b[5],c3,c1,c2);
+ $LD r7,`5*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r12,r8
+ adde r10,r10,r9
+ addze r11,r0
+ #mul_add_c(a[1],b[4],c3,c1,c2);
+ $LD r6,`1*$BNSZ`(r4)
+ $LD r7,`4*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r12,r8
+ adde r10,r10,r9
+ addze r11,r11
+ #mul_add_c(a[2],b[3],c3,c1,c2);
+ $LD r6,`2*$BNSZ`(r4)
+ $LD r7,`3*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r12,r8
+ adde r10,r10,r9
+ addze r11,r11
+ #mul_add_c(a[3],b[2],c3,c1,c2);
+ $LD r6,`3*$BNSZ`(r4)
+ $LD r7,`2*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r12,r8
+ adde r10,r10,r9
+ addze r11,r11
+ #mul_add_c(a[4],b[1],c3,c1,c2);
+ $LD r6,`4*$BNSZ`(r4)
+ $LD r7,`1*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r12,r8
+ adde r10,r10,r9
+ addze r11,r11
+ #mul_add_c(a[5],b[0],c3,c1,c2);
+ $LD r6,`5*$BNSZ`(r4)
+ $LD r7,`0*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r12,r8
+ adde r10,r10,r9
+ addze r11,r11
+ $ST r12,`5*$BNSZ`(r3) #r[5]=c3;
+ #mul_add_c(a[6],b[0],c1,c2,c3);
+ $LD r6,`6*$BNSZ`(r4)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r10,r8
+ adde r11,r11,r9
+ addze r12,r0
+ #mul_add_c(a[5],b[1],c1,c2,c3);
+ $LD r6,`5*$BNSZ`(r4)
+ $LD r7,`1*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r10,r8
+ adde r11,r11,r9
+ addze r12,r12
+ #mul_add_c(a[4],b[2],c1,c2,c3);
+ $LD r6,`4*$BNSZ`(r4)
+ $LD r7,`2*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r10,r8
+ adde r11,r11,r9
+ addze r12,r12
+ #mul_add_c(a[3],b[3],c1,c2,c3);
+ $LD r6,`3*$BNSZ`(r4)
+ $LD r7,`3*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r10,r8
+ adde r11,r11,r9
+ addze r12,r12
+ #mul_add_c(a[2],b[4],c1,c2,c3);
+ $LD r6,`2*$BNSZ`(r4)
+ $LD r7,`4*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r10,r8
+ adde r11,r11,r9
+ addze r12,r12
+ #mul_add_c(a[1],b[5],c1,c2,c3);
+ $LD r6,`1*$BNSZ`(r4)
+ $LD r7,`5*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r10,r8
+ adde r11,r11,r9
+ addze r12,r12
+ #mul_add_c(a[0],b[6],c1,c2,c3);
+ $LD r6,`0*$BNSZ`(r4)
+ $LD r7,`6*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r10,r8
+ adde r11,r11,r9
+ addze r12,r12
+ $ST r10,`6*$BNSZ`(r3) #r[6]=c1;
+ #mul_add_c(a[0],b[7],c2,c3,c1);
+ $LD r7,`7*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r11,r8
+ adde r12,r12,r9
+ addze r10,r0
+ #mul_add_c(a[1],b[6],c2,c3,c1);
+ $LD r6,`1*$BNSZ`(r4)
+ $LD r7,`6*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r11,r8
+ adde r12,r12,r9
+ addze r10,r10
+ #mul_add_c(a[2],b[5],c2,c3,c1);
+ $LD r6,`2*$BNSZ`(r4)
+ $LD r7,`5*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r11,r8
+ adde r12,r12,r9
+ addze r10,r10
+ #mul_add_c(a[3],b[4],c2,c3,c1);
+ $LD r6,`3*$BNSZ`(r4)
+ $LD r7,`4*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r11,r8
+ adde r12,r12,r9
+ addze r10,r10
+ #mul_add_c(a[4],b[3],c2,c3,c1);
+ $LD r6,`4*$BNSZ`(r4)
+ $LD r7,`3*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r11,r8
+ adde r12,r12,r9
+ addze r10,r10
+ #mul_add_c(a[5],b[2],c2,c3,c1);
+ $LD r6,`5*$BNSZ`(r4)
+ $LD r7,`2*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r11,r8
+ adde r12,r12,r9
+ addze r10,r10
+ #mul_add_c(a[6],b[1],c2,c3,c1);
+ $LD r6,`6*$BNSZ`(r4)
+ $LD r7,`1*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r11,r8
+ adde r12,r12,r9
+ addze r10,r10
+ #mul_add_c(a[7],b[0],c2,c3,c1);
+ $LD r6,`7*$BNSZ`(r4)
+ $LD r7,`0*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r11,r8
+ adde r12,r12,r9
+ addze r10,r10
+ $ST r11,`7*$BNSZ`(r3) #r[7]=c2;
+ #mul_add_c(a[7],b[1],c3,c1,c2);
+ $LD r7,`1*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r12,r8
+ adde r10,r10,r9
+ addze r11,r0
+ #mul_add_c(a[6],b[2],c3,c1,c2);
+ $LD r6,`6*$BNSZ`(r4)
+ $LD r7,`2*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r12,r8
+ adde r10,r10,r9
+ addze r11,r11
+ #mul_add_c(a[5],b[3],c3,c1,c2);
+ $LD r6,`5*$BNSZ`(r4)
+ $LD r7,`3*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r12,r8
+ adde r10,r10,r9
+ addze r11,r11
+ #mul_add_c(a[4],b[4],c3,c1,c2);
+ $LD r6,`4*$BNSZ`(r4)
+ $LD r7,`4*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r12,r8
+ adde r10,r10,r9
+ addze r11,r11
+ #mul_add_c(a[3],b[5],c3,c1,c2);
+ $LD r6,`3*$BNSZ`(r4)
+ $LD r7,`5*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r12,r8
+ adde r10,r10,r9
+ addze r11,r11
+ #mul_add_c(a[2],b[6],c3,c1,c2);
+ $LD r6,`2*$BNSZ`(r4)
+ $LD r7,`6*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r12,r8
+ adde r10,r10,r9
+ addze r11,r11
+ #mul_add_c(a[1],b[7],c3,c1,c2);
+ $LD r6,`1*$BNSZ`(r4)
+ $LD r7,`7*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r12,r8
+ adde r10,r10,r9
+ addze r11,r11
+ $ST r12,`8*$BNSZ`(r3) #r[8]=c3;
+ #mul_add_c(a[2],b[7],c1,c2,c3);
+ $LD r6,`2*$BNSZ`(r4)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r10,r8
+ adde r11,r11,r9
+ addze r12,r0
+ #mul_add_c(a[3],b[6],c1,c2,c3);
+ $LD r6,`3*$BNSZ`(r4)
+ $LD r7,`6*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r10,r8
+ adde r11,r11,r9
+ addze r12,r12
+ #mul_add_c(a[4],b[5],c1,c2,c3);
+ $LD r6,`4*$BNSZ`(r4)
+ $LD r7,`5*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r10,r8
+ adde r11,r11,r9
+ addze r12,r12
+ #mul_add_c(a[5],b[4],c1,c2,c3);
+ $LD r6,`5*$BNSZ`(r4)
+ $LD r7,`4*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r10,r8
+ adde r11,r11,r9
+ addze r12,r12
+ #mul_add_c(a[6],b[3],c1,c2,c3);
+ $LD r6,`6*$BNSZ`(r4)
+ $LD r7,`3*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r10,r8
+ adde r11,r11,r9
+ addze r12,r12
+ #mul_add_c(a[7],b[2],c1,c2,c3);
+ $LD r6,`7*$BNSZ`(r4)
+ $LD r7,`2*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r10,r8
+ adde r11,r11,r9
+ addze r12,r12
+ $ST r10,`9*$BNSZ`(r3) #r[9]=c1;
+ #mul_add_c(a[7],b[3],c2,c3,c1);
+ $LD r7,`3*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r11,r8
+ adde r12,r12,r9
+ addze r10,r0
+ #mul_add_c(a[6],b[4],c2,c3,c1);
+ $LD r6,`6*$BNSZ`(r4)
+ $LD r7,`4*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r11,r8
+ adde r12,r12,r9
+ addze r10,r10
+ #mul_add_c(a[5],b[5],c2,c3,c1);
+ $LD r6,`5*$BNSZ`(r4)
+ $LD r7,`5*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r11,r8
+ adde r12,r12,r9
+ addze r10,r10
+ #mul_add_c(a[4],b[6],c2,c3,c1);
+ $LD r6,`4*$BNSZ`(r4)
+ $LD r7,`6*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r11,r8
+ adde r12,r12,r9
+ addze r10,r10
+ #mul_add_c(a[3],b[7],c2,c3,c1);
+ $LD r6,`3*$BNSZ`(r4)
+ $LD r7,`7*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r11,r8
+ adde r12,r12,r9
+ addze r10,r10
+ $ST r11,`10*$BNSZ`(r3) #r[10]=c2;
+ #mul_add_c(a[4],b[7],c3,c1,c2);
+ $LD r6,`4*$BNSZ`(r4)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r12,r8
+ adde r10,r10,r9
+ addze r11,r0
+ #mul_add_c(a[5],b[6],c3,c1,c2);
+ $LD r6,`5*$BNSZ`(r4)
+ $LD r7,`6*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r12,r8
+ adde r10,r10,r9
+ addze r11,r11
+ #mul_add_c(a[6],b[5],c3,c1,c2);
+ $LD r6,`6*$BNSZ`(r4)
+ $LD r7,`5*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r12,r8
+ adde r10,r10,r9
+ addze r11,r11
+ #mul_add_c(a[7],b[4],c3,c1,c2);
+ $LD r6,`7*$BNSZ`(r4)
+ $LD r7,`4*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r12,r8
+ adde r10,r10,r9
+ addze r11,r11
+ $ST r12,`11*$BNSZ`(r3) #r[11]=c3;
+ #mul_add_c(a[7],b[5],c1,c2,c3);
+ $LD r7,`5*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r10,r8
+ adde r11,r11,r9
+ addze r12,r0
+ #mul_add_c(a[6],b[6],c1,c2,c3);
+ $LD r6,`6*$BNSZ`(r4)
+ $LD r7,`6*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r10,r8
+ adde r11,r11,r9
+ addze r12,r12
+ #mul_add_c(a[5],b[7],c1,c2,c3);
+ $LD r6,`5*$BNSZ`(r4)
+ $LD r7,`7*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r10,r10,r8
+ adde r11,r11,r9
+ addze r12,r12
+ $ST r10,`12*$BNSZ`(r3) #r[12]=c1;
+ #mul_add_c(a[6],b[7],c2,c3,c1);
+ $LD r6,`6*$BNSZ`(r4)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r11,r8
+ adde r12,r12,r9
+ addze r10,r0
+ #mul_add_c(a[7],b[6],c2,c3,c1);
+ $LD r6,`7*$BNSZ`(r4)
+ $LD r7,`6*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r11,r11,r8
+ adde r12,r12,r9
+ addze r10,r10
+ $ST r11,`13*$BNSZ`(r3) #r[13]=c2;
+ #mul_add_c(a[7],b[7],c3,c1,c2);
+ $LD r7,`7*$BNSZ`(r5)
+ $UMULL r8,r6,r7
+ $UMULH r9,r6,r7
+ addc r12,r12,r8
+ adde r10,r10,r9
+ $ST r12,`14*$BNSZ`(r3) #r[14]=c3;
+ $ST r10,`15*$BNSZ`(r3) #r[15]=c1;
+ blr
+ .long 0
+ .byte 0,12,0x14,0,0,0,3,0
+ .long 0
+.size .bn_mul_comba8,.-.bn_mul_comba8
+
+#
+# NOTE: The following label name should be changed to
+# "bn_sub_words" i.e. remove the first dot
+# for the gcc compiler. This should be automatically
+# done in the build
+#
+#
+.align 4
+.bn_sub_words:
+#
+# Handcoded version of bn_sub_words
+#
+#BN_ULONG bn_sub_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
+#
+# r3 = r
+# r4 = a
+# r5 = b
+# r6 = n
+#
+# Note: No loop unrolling done since this is not a performance
+# critical loop.
+
+ xor r0,r0,r0 #set r0 = 0
+#
+# check for r6 = 0 AND set carry bit.
+#
+ subfc. r7,r0,r6 # If r6 is 0 then result is 0.
+ # if r6 > 0 then result !=0
+ # In either case carry bit is set.
+ beq Lppcasm_sub_adios
+ addi r4,r4,-$BNSZ
+ addi r3,r3,-$BNSZ
+ addi r5,r5,-$BNSZ
+ mtctr r6
+Lppcasm_sub_mainloop:
+ $LDU r7,$BNSZ(r4)
+ $LDU r8,$BNSZ(r5)
+ subfe r6,r8,r7 # r6 = r7+carry bit + onescomplement(r8)
+ # if carry = 1 this is r7-r8. Else it
+ # is r7-r8 -1 as we need.
+ $STU r6,$BNSZ(r3)
+ bdnz Lppcasm_sub_mainloop
+Lppcasm_sub_adios:
+ subfze r3,r0 # if carry bit is set then r3 = 0 else -1
+ andi. r3,r3,1 # keep only last bit.
+ blr
+ .long 0
+ .byte 0,12,0x14,0,0,0,4,0
+ .long 0
+.size .bn_sub_words,.-.bn_sub_words
+
+#
+# NOTE: The following label name should be changed to
+# "bn_add_words" i.e. remove the first dot
+# for the gcc compiler. This should be automatically
+# done in the build
+#
+
+.align 4
+.bn_add_words:
+#
+# Handcoded version of bn_add_words
+#
+#BN_ULONG bn_add_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
+#
+# r3 = r
+# r4 = a
+# r5 = b
+# r6 = n
+#
+# Note: No loop unrolling done since this is not a performance
+# critical loop.
+
+ xor r0,r0,r0
+#
+# check for r6 = 0. Is this needed?
+#
+ addic. r6,r6,0 #test r6 and clear carry bit.
+ beq Lppcasm_add_adios
+ addi r4,r4,-$BNSZ
+ addi r3,r3,-$BNSZ
+ addi r5,r5,-$BNSZ
+ mtctr r6
+Lppcasm_add_mainloop:
+ $LDU r7,$BNSZ(r4)
+ $LDU r8,$BNSZ(r5)
+ adde r8,r7,r8
+ $STU r8,$BNSZ(r3)
+ bdnz Lppcasm_add_mainloop
+Lppcasm_add_adios:
+ addze r3,r0 #return carry bit.
+ blr
+ .long 0
+ .byte 0,12,0x14,0,0,0,4,0
+ .long 0
+.size .bn_add_words,.-.bn_add_words
+
+#
+# NOTE: The following label name should be changed to
+# "bn_div_words" i.e. remove the first dot
+# for the gcc compiler. This should be automatically
+# done in the build
+#
+
+.align 4
+.bn_div_words:
+#
+# This is a cleaned up version of code generated by
+# the AIX compiler. The only optimization is to use
+# the PPC instruction to count leading zeros instead
+# of call to num_bits_word. Since this was compiled
+# only at level -O2 we can possibly squeeze it more?
+#
+# r3 = h
+# r4 = l
+# r5 = d
+
+ $UCMPI 0,r5,0 # compare r5 and 0
+ bne Lppcasm_div1 # proceed if d!=0
+ li r3,-1 # d=0 return -1
+ blr
+Lppcasm_div1:
+ xor r0,r0,r0 #r0=0
+ li r8,$BITS
+ $CNTLZ. r7,r5 #r7 = num leading 0s in d.
+ beq Lppcasm_div2 #proceed if no leading zeros
+ subf r8,r7,r8 #r8 = BN_num_bits_word(d)
+ $SHR. r9,r3,r8 #are there any bits above r8'th?
+ $TR 16,r9,r0 #if there're, signal to dump core...
+Lppcasm_div2:
+ $UCMP 0,r3,r5 #h>=d?
+ blt Lppcasm_div3 #goto Lppcasm_div3 if not
+ subf r3,r5,r3 #h-=d ;
+Lppcasm_div3: #r7 = BN_BITS2-i. so r7=i
+ cmpi 0,0,r7,0 # is (i == 0)?
+ beq Lppcasm_div4
+ $SHL r3,r3,r7 # h = (h<< i)
+ $SHR r8,r4,r8 # r8 = (l >> BN_BITS2 -i)
+ $SHL r5,r5,r7 # d<<=i
+ or r3,r3,r8 # h = (h<<i)|(l>>(BN_BITS2-i))
+ $SHL r4,r4,r7 # l <<=i
+Lppcasm_div4:
+ $SHRI r9,r5,`$BITS/2` # r9 = dh
+ # dl will be computed when needed
+ # as it saves registers.
+ li r6,2 #r6=2
+ mtctr r6 #counter will be in count.
+Lppcasm_divouterloop:
+ $SHRI r8,r3,`$BITS/2` #r8 = (h>>BN_BITS4)
+ $SHRI r11,r4,`$BITS/2` #r11= (l&BN_MASK2h)>>BN_BITS4
+ # compute here for innerloop.
+ $UCMP 0,r8,r9 # is (h>>BN_BITS4)==dh
+ bne Lppcasm_div5 # goto Lppcasm_div5 if not
+
+ li r8,-1
+ $CLRU r8,r8,`$BITS/2` #q = BN_MASK2l
+ b Lppcasm_div6
+Lppcasm_div5:
+ $UDIV r8,r3,r9 #q = h/dh
+Lppcasm_div6:
+ $UMULL r12,r9,r8 #th = q*dh
+ $CLRU r10,r5,`$BITS/2` #r10=dl
+ $UMULL r6,r8,r10 #tl = q*dl
+
+Lppcasm_divinnerloop:
+ subf r10,r12,r3 #t = h -th
+ $SHRI r7,r10,`$BITS/2` #r7= (t &BN_MASK2H), sort of...
+ addic. r7,r7,0 #test if r7 == 0. used below.
+ # now want to compute
+ # r7 = (t<<BN_BITS4)|((l&BN_MASK2h)>>BN_BITS4)
+ # the following 2 instructions do that
+ $SHLI r7,r10,`$BITS/2` # r7 = (t<<BN_BITS4)
+ or r7,r7,r11 # r7|=((l&BN_MASK2h)>>BN_BITS4)
+ $UCMP cr1,r6,r7 # compare (tl <= r7)
+ bne Lppcasm_divinnerexit
+ ble cr1,Lppcasm_divinnerexit
+ addi r8,r8,-1 #q--
+ subf r12,r9,r12 #th -=dh
+ $CLRU r10,r5,`$BITS/2` #r10=dl. t is no longer needed in loop.
+ subf r6,r10,r6 #tl -=dl
+ b Lppcasm_divinnerloop
+Lppcasm_divinnerexit:
+ $SHRI r10,r6,`$BITS/2` #t=(tl>>BN_BITS4)
+ $SHLI r11,r6,`$BITS/2` #tl=(tl<<BN_BITS4)&BN_MASK2h;
+ $UCMP cr1,r4,r11 # compare l and tl
+ add r12,r12,r10 # th+=t
+ bge cr1,Lppcasm_div7 # if (l>=tl) goto Lppcasm_div7
+ addi r12,r12,1 # th++
+Lppcasm_div7:
+ subf r11,r11,r4 #r11=l-tl
+ $UCMP cr1,r3,r12 #compare h and th
+ bge cr1,Lppcasm_div8 #if (h>=th) goto Lppcasm_div8
+ addi r8,r8,-1 # q--
+ add r3,r5,r3 # h+=d
+Lppcasm_div8:
+ subf r12,r12,r3 #r12 = h-th
+ $SHLI r4,r11,`$BITS/2` #l=(l&BN_MASK2l)<<BN_BITS4
+ # want to compute
+ # h = ((h<<BN_BITS4)|(l>>BN_BITS4))&BN_MASK2
+ # the following 2 instructions will do this.
+ $INSR r11,r12,`$BITS/2`,`$BITS/2` # r11 is the value we want rotated $BITS/2.
+ $ROTL r3,r11,`$BITS/2` # rotate by $BITS/2 and store in r3
+ bdz Lppcasm_div9 #if (count==0) break ;
+ $SHLI r0,r8,`$BITS/2` #ret =q<<BN_BITS4
+ b Lppcasm_divouterloop
+Lppcasm_div9:
+ or r3,r8,r0
+ blr
+ .long 0
+ .byte 0,12,0x14,0,0,0,3,0
+ .long 0
+.size .bn_div_words,.-.bn_div_words
+
+#
+# NOTE: The following label name should be changed to
+# "bn_sqr_words" i.e. remove the first dot
+# for the gcc compiler. This should be automatically
+# done in the build
+#
+.align 4
+.bn_sqr_words:
+#
+# Optimized version of bn_sqr_words
+#
+# void bn_sqr_words(BN_ULONG *r, BN_ULONG *a, int n)
+#
+# r3 = r
+# r4 = a
+# r5 = n
+#
+# r6 = a[i].
+# r7,r8 = product.
+#
+# No unrolling done here. Not performance critical.
+
+ addic. r5,r5,0 #test r5.
+ beq Lppcasm_sqr_adios
+ addi r4,r4,-$BNSZ
+ addi r3,r3,-$BNSZ
+ mtctr r5
+Lppcasm_sqr_mainloop:
+ #sqr(r[0],r[1],a[0]);
+ $LDU r6,$BNSZ(r4)
+ $UMULL r7,r6,r6
+ $UMULH r8,r6,r6
+ $STU r7,$BNSZ(r3)
+ $STU r8,$BNSZ(r3)
+ bdnz Lppcasm_sqr_mainloop
+Lppcasm_sqr_adios:
+ blr
+ .long 0
+ .byte 0,12,0x14,0,0,0,3,0
+ .long 0
+.size .bn_sqr_words,.-.bn_sqr_words
+
+#
+# NOTE: The following label name should be changed to
+# "bn_mul_words" i.e. remove the first dot
+# for the gcc compiler. This should be automatically
+# done in the build
+#
+
+.align 4
+.bn_mul_words:
+#
+# BN_ULONG bn_mul_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w)
+#
+# r3 = rp
+# r4 = ap
+# r5 = num
+# r6 = w
+ xor r0,r0,r0
+ xor r12,r12,r12 # used for carry
+ rlwinm. r7,r5,30,2,31 # num >> 2
+ beq Lppcasm_mw_REM
+ mtctr r7
+Lppcasm_mw_LOOP:
+ #mul(rp[0],ap[0],w,c1);
+ $LD r8,`0*$BNSZ`(r4)
+ $UMULL r9,r6,r8
+ $UMULH r10,r6,r8
+ addc r9,r9,r12
+ #addze r10,r10 #carry is NOT ignored.
+ #will be taken care of
+ #in second spin below
+ #using adde.
+ $ST r9,`0*$BNSZ`(r3)
+ #mul(rp[1],ap[1],w,c1);
+ $LD r8,`1*$BNSZ`(r4)
+ $UMULL r11,r6,r8
+ $UMULH r12,r6,r8
+ adde r11,r11,r10
+ #addze r12,r12
+ $ST r11,`1*$BNSZ`(r3)
+ #mul(rp[2],ap[2],w,c1);
+ $LD r8,`2*$BNSZ`(r4)
+ $UMULL r9,r6,r8
+ $UMULH r10,r6,r8
+ adde r9,r9,r12
+ #addze r10,r10
+ $ST r9,`2*$BNSZ`(r3)
+ #mul_add(rp[3],ap[3],w,c1);
+ $LD r8,`3*$BNSZ`(r4)
+ $UMULL r11,r6,r8
+ $UMULH r12,r6,r8
+ adde r11,r11,r10
+ addze r12,r12 #this spin we collect carry into
+ #r12
+ $ST r11,`3*$BNSZ`(r3)
+
+ addi r3,r3,`4*$BNSZ`
+ addi r4,r4,`4*$BNSZ`
+ bdnz Lppcasm_mw_LOOP
+
+Lppcasm_mw_REM:
+ andi. r5,r5,0x3
+ beq Lppcasm_mw_OVER
+ #mul(rp[0],ap[0],w,c1);
+ $LD r8,`0*$BNSZ`(r4)
+ $UMULL r9,r6,r8
+ $UMULH r10,r6,r8
+ addc r9,r9,r12
+ addze r10,r10
+ $ST r9,`0*$BNSZ`(r3)
+ addi r12,r10,0
+
+ addi r5,r5,-1
+ cmpli 0,0,r5,0
+ beq Lppcasm_mw_OVER
+
+
+ #mul(rp[1],ap[1],w,c1);
+ $LD r8,`1*$BNSZ`(r4)
+ $UMULL r9,r6,r8
+ $UMULH r10,r6,r8
+ addc r9,r9,r12
+ addze r10,r10
+ $ST r9,`1*$BNSZ`(r3)
+ addi r12,r10,0
+
+ addi r5,r5,-1
+ cmpli 0,0,r5,0
+ beq Lppcasm_mw_OVER
+
+ #mul_add(rp[2],ap[2],w,c1);
+ $LD r8,`2*$BNSZ`(r4)
+ $UMULL r9,r6,r8
+ $UMULH r10,r6,r8
+ addc r9,r9,r12
+ addze r10,r10
+ $ST r9,`2*$BNSZ`(r3)
+ addi r12,r10,0
+
+Lppcasm_mw_OVER:
+ addi r3,r12,0
+ blr
+ .long 0
+ .byte 0,12,0x14,0,0,0,4,0
+ .long 0
+.size .bn_mul_words,.-.bn_mul_words
+
+#
+# NOTE: The following label name should be changed to
+# "bn_mul_add_words" i.e. remove the first dot
+# for the gcc compiler. This should be automatically
+# done in the build
+#
+
+.align 4
+.bn_mul_add_words:
+#
+# BN_ULONG bn_mul_add_words(BN_ULONG *rp, BN_ULONG *ap, int num, BN_ULONG w)
+#
+# r3 = rp
+# r4 = ap
+# r5 = num
+# r6 = w
+#
+# empirical evidence suggests that unrolled version performs best!!
+#
+ xor r0,r0,r0 #r0 = 0
+ xor r12,r12,r12 #r12 = 0 . used for carry
+ rlwinm. r7,r5,30,2,31 # num >> 2
+ beq Lppcasm_maw_leftover # if (num < 4) go LPPCASM_maw_leftover
+ mtctr r7
+Lppcasm_maw_mainloop:
+ #mul_add(rp[0],ap[0],w,c1);
+ $LD r8,`0*$BNSZ`(r4)
+ $LD r11,`0*$BNSZ`(r3)
+ $UMULL r9,r6,r8
+ $UMULH r10,r6,r8
+ addc r9,r9,r12 #r12 is carry.
+ addze r10,r10
+ addc r9,r9,r11
+ #addze r10,r10
+ #the above instruction addze
+ #is NOT needed. Carry will NOT
+ #be ignored. It's not affected
+ #by multiply and will be collected
+ #in the next spin
+ $ST r9,`0*$BNSZ`(r3)
+
+ #mul_add(rp[1],ap[1],w,c1);
+ $LD r8,`1*$BNSZ`(r4)
+ $LD r9,`1*$BNSZ`(r3)
+ $UMULL r11,r6,r8
+ $UMULH r12,r6,r8
+ adde r11,r11,r10 #r10 is carry.
+ addze r12,r12
+ addc r11,r11,r9
+ #addze r12,r12
+ $ST r11,`1*$BNSZ`(r3)
+
+ #mul_add(rp[2],ap[2],w,c1);
+ $LD r8,`2*$BNSZ`(r4)
+ $UMULL r9,r6,r8
+ $LD r11,`2*$BNSZ`(r3)
+ $UMULH r10,r6,r8
+ adde r9,r9,r12
+ addze r10,r10
+ addc r9,r9,r11
+ #addze r10,r10
+ $ST r9,`2*$BNSZ`(r3)
+
+ #mul_add(rp[3],ap[3],w,c1);
+ $LD r8,`3*$BNSZ`(r4)
+ $UMULL r11,r6,r8
+ $LD r9,`3*$BNSZ`(r3)
+ $UMULH r12,r6,r8
+ adde r11,r11,r10
+ addze r12,r12
+ addc r11,r11,r9
+ addze r12,r12
+ $ST r11,`3*$BNSZ`(r3)
+ addi r3,r3,`4*$BNSZ`
+ addi r4,r4,`4*$BNSZ`
+ bdnz Lppcasm_maw_mainloop
+
+Lppcasm_maw_leftover:
+ andi. r5,r5,0x3
+ beq Lppcasm_maw_adios
+ addi r3,r3,-$BNSZ
+ addi r4,r4,-$BNSZ
+ #mul_add(rp[0],ap[0],w,c1);
+ mtctr r5
+ $LDU r8,$BNSZ(r4)
+ $UMULL r9,r6,r8
+ $UMULH r10,r6,r8
+ $LDU r11,$BNSZ(r3)
+ addc r9,r9,r11
+ addze r10,r10
+ addc r9,r9,r12
+ addze r12,r10
+ $ST r9,0(r3)
+
+ bdz Lppcasm_maw_adios
+ #mul_add(rp[1],ap[1],w,c1);
+ $LDU r8,$BNSZ(r4)
+ $UMULL r9,r6,r8
+ $UMULH r10,r6,r8
+ $LDU r11,$BNSZ(r3)
+ addc r9,r9,r11
+ addze r10,r10
+ addc r9,r9,r12
+ addze r12,r10
+ $ST r9,0(r3)
+
+ bdz Lppcasm_maw_adios
+ #mul_add(rp[2],ap[2],w,c1);
+ $LDU r8,$BNSZ(r4)
+ $UMULL r9,r6,r8
+ $UMULH r10,r6,r8
+ $LDU r11,$BNSZ(r3)
+ addc r9,r9,r11
+ addze r10,r10
+ addc r9,r9,r12
+ addze r12,r10
+ $ST r9,0(r3)
+
+Lppcasm_maw_adios:
+ addi r3,r12,0
+ blr
+ .long 0
+ .byte 0,12,0x14,0,0,0,4,0
+ .long 0
+.size .bn_mul_add_words,.-.bn_mul_add_words
+ .align 4
+EOF
+$data =~ s/\`([^\`]*)\`/eval $1/gem;
+print $data;
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/ppc64-mont.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/ppc64-mont.pl
new file mode 100644
index 0000000..b55fd46
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/ppc64-mont.pl
@@ -0,0 +1,1652 @@
+#! /usr/bin/env perl
+# Copyright 2007-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+
+# December 2007
+
+# The reason for undertaken effort is basically following. Even though
+# Power 6 CPU operates at incredible 4.7GHz clock frequency, its PKI
+# performance was observed to be less than impressive, essentially as
+# fast as 1.8GHz PPC970, or 2.6 times(!) slower than one would hope.
+# Well, it's not surprising that IBM had to make some sacrifices to
+# boost the clock frequency that much, but no overall improvement?
+# Having observed how much difference did switching to FPU make on
+# UltraSPARC, playing same stunt on Power 6 appeared appropriate...
+# Unfortunately the resulting performance improvement is not as
+# impressive, ~30%, and in absolute terms is still very far from what
+# one would expect from 4.7GHz CPU. There is a chance that I'm doing
+# something wrong, but in the lack of assembler level micro-profiling
+# data or at least decent platform guide I can't tell... Or better
+# results might be achieved with VMX... Anyway, this module provides
+# *worse* performance on other PowerPC implementations, ~40-15% slower
+# on PPC970 depending on key length and ~40% slower on Power 5 for all
+# key lengths. As it's obviously inappropriate as "best all-round"
+# alternative, it has to be complemented with run-time CPU family
+# detection. Oh! It should also be noted that unlike other PowerPC
+# implementation IALU ppc-mont.pl module performs *suboptimally* on
+# >=1024-bit key lengths on Power 6. It should also be noted that
+# *everything* said so far applies to 64-bit builds! As far as 32-bit
+# application executed on 64-bit CPU goes, this module is likely to
+# become preferred choice, because it's easy to adapt it for such
+# case and *is* faster than 32-bit ppc-mont.pl on *all* processors.
+
+# February 2008
+
+# Micro-profiling assisted optimization results in ~15% improvement
+# over original ppc64-mont.pl version, or overall ~50% improvement
+# over ppc.pl module on Power 6. If compared to ppc-mont.pl on same
+# Power 6 CPU, this module is 5-150% faster depending on key length,
+# [hereafter] more for longer keys. But if compared to ppc-mont.pl
+# on 1.8GHz PPC970, it's only 5-55% faster. Still far from impressive
+# in absolute terms, but it's apparently the way Power 6 is...
+
+# December 2009
+
+# Adapted for 32-bit build this module delivers 25-120%, yes, more
+# than *twice* for longer keys, performance improvement over 32-bit
+# ppc-mont.pl on 1.8GHz PPC970. However! This implementation utilizes
+# even 64-bit integer operations and the trouble is that most PPC
+# operating systems don't preserve upper halves of general purpose
+# registers upon 32-bit signal delivery. They do preserve them upon
+# context switch, but not signalling:-( This means that asynchronous
+# signals have to be blocked upon entry to this subroutine. Signal
+# masking (and of course complementary unmasking) has quite an impact
+# on performance, naturally larger for shorter keys. It's so severe
+# that 512-bit key performance can be as low as 1/3 of expected one.
+# This is why this routine can be engaged for longer key operations
+# only on these OSes, see crypto/ppccap.c for further details. MacOS X
+# is an exception from this and doesn't require signal masking, and
+# that's where above improvement coefficients were collected. For
+# others alternative would be to break dependence on upper halves of
+# GPRs by sticking to 32-bit integer operations...
+
+# December 2012
+
+# Remove above mentioned dependence on GPRs' upper halves in 32-bit
+# build. No signal masking overhead, but integer instructions are
+# *more* numerous... It's still "universally" faster than 32-bit
+# ppc-mont.pl, but improvement coefficient is not as impressive
+# for longer keys...
+
+$flavour = shift;
+
+if ($flavour =~ /32/) {
+ $SIZE_T=4;
+ $RZONE= 224;
+ $fname= "bn_mul_mont_fpu64";
+
+ $STUX= "stwux"; # store indexed and update
+ $PUSH= "stw";
+ $POP= "lwz";
+} elsif ($flavour =~ /64/) {
+ $SIZE_T=8;
+ $RZONE= 288;
+ $fname= "bn_mul_mont_fpu64";
+
+ # same as above, but 64-bit mnemonics...
+ $STUX= "stdux"; # store indexed and update
+ $PUSH= "std";
+ $POP= "ld";
+} else { die "nonsense $flavour"; }
+
+$LITTLE_ENDIAN = ($flavour=~/le$/) ? 4 : 0;
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or
+( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or
+die "can't locate ppc-xlate.pl";
+
+open STDOUT,"| $^X $xlate $flavour ".shift || die "can't call $xlate: $!";
+
+$FRAME=64; # padded frame header
+$TRANSFER=16*8;
+
+$carry="r0";
+$sp="r1";
+$toc="r2";
+$rp="r3"; $ovf="r3";
+$ap="r4";
+$bp="r5";
+$np="r6";
+$n0="r7";
+$num="r8";
+$rp="r9"; # $rp is reassigned
+$tp="r10";
+$j="r11";
+$i="r12";
+# non-volatile registers
+$c1="r19";
+$n1="r20";
+$a1="r21";
+$nap_d="r22"; # interleaved ap and np in double format
+$a0="r23"; # ap[0]
+$t0="r24"; # temporary registers
+$t1="r25";
+$t2="r26";
+$t3="r27";
+$t4="r28";
+$t5="r29";
+$t6="r30";
+$t7="r31";
+
+# PPC offers enough register bank capacity to unroll inner loops twice
+#
+# ..A3A2A1A0
+# dcba
+# -----------
+# A0a
+# A0b
+# A0c
+# A0d
+# A1a
+# A1b
+# A1c
+# A1d
+# A2a
+# A2b
+# A2c
+# A2d
+# A3a
+# A3b
+# A3c
+# A3d
+# ..a
+# ..b
+#
+$ba="f0"; $bb="f1"; $bc="f2"; $bd="f3";
+$na="f4"; $nb="f5"; $nc="f6"; $nd="f7";
+$dota="f8"; $dotb="f9";
+$A0="f10"; $A1="f11"; $A2="f12"; $A3="f13";
+$N0="f20"; $N1="f21"; $N2="f22"; $N3="f23";
+$T0a="f24"; $T0b="f25";
+$T1a="f26"; $T1b="f27";
+$T2a="f28"; $T2b="f29";
+$T3a="f30"; $T3b="f31";
+
+# sp----------->+-------------------------------+
+# | saved sp |
+# +-------------------------------+
+# . .
+# +64 +-------------------------------+
+# | 16 gpr<->fpr transfer zone |
+# . .
+# . .
+# +16*8 +-------------------------------+
+# | __int64 tmp[-1] |
+# +-------------------------------+
+# | __int64 tmp[num] |
+# . .
+# . .
+# . .
+# +(num+1)*8 +-------------------------------+
+# | padding to 64 byte boundary |
+# . .
+# +X +-------------------------------+
+# | double nap_d[4*num] |
+# . .
+# . .
+# . .
+# +-------------------------------+
+# . .
+# -13*size_t +-------------------------------+
+# | 13 saved gpr, r19-r31 |
+# . .
+# . .
+# -12*8 +-------------------------------+
+# | 12 saved fpr, f20-f31 |
+# . .
+# . .
+# +-------------------------------+
+
+$code=<<___;
+.machine "any"
+.text
+
+.globl .$fname
+.align 5
+.$fname:
+ cmpwi $num,`3*8/$SIZE_T`
+ mr $rp,r3 ; $rp is reassigned
+ li r3,0 ; possible "not handled" return code
+ bltlr-
+ andi. r0,$num,`16/$SIZE_T-1` ; $num has to be "even"
+ bnelr-
+
+ slwi $num,$num,`log($SIZE_T)/log(2)` ; num*=sizeof(BN_LONG)
+ li $i,-4096
+ slwi $tp,$num,2 ; place for {an}p_{lh}[num], i.e. 4*num
+ add $tp,$tp,$num ; place for tp[num+1]
+ addi $tp,$tp,`$FRAME+$TRANSFER+8+64+$RZONE`
+ subf $tp,$tp,$sp ; $sp-$tp
+ and $tp,$tp,$i ; minimize TLB usage
+ subf $tp,$sp,$tp ; $tp-$sp
+ mr $i,$sp
+ $STUX $sp,$sp,$tp ; alloca
+
+ $PUSH r19,`-12*8-13*$SIZE_T`($i)
+ $PUSH r20,`-12*8-12*$SIZE_T`($i)
+ $PUSH r21,`-12*8-11*$SIZE_T`($i)
+ $PUSH r22,`-12*8-10*$SIZE_T`($i)
+ $PUSH r23,`-12*8-9*$SIZE_T`($i)
+ $PUSH r24,`-12*8-8*$SIZE_T`($i)
+ $PUSH r25,`-12*8-7*$SIZE_T`($i)
+ $PUSH r26,`-12*8-6*$SIZE_T`($i)
+ $PUSH r27,`-12*8-5*$SIZE_T`($i)
+ $PUSH r28,`-12*8-4*$SIZE_T`($i)
+ $PUSH r29,`-12*8-3*$SIZE_T`($i)
+ $PUSH r30,`-12*8-2*$SIZE_T`($i)
+ $PUSH r31,`-12*8-1*$SIZE_T`($i)
+ stfd f20,`-12*8`($i)
+ stfd f21,`-11*8`($i)
+ stfd f22,`-10*8`($i)
+ stfd f23,`-9*8`($i)
+ stfd f24,`-8*8`($i)
+ stfd f25,`-7*8`($i)
+ stfd f26,`-6*8`($i)
+ stfd f27,`-5*8`($i)
+ stfd f28,`-4*8`($i)
+ stfd f29,`-3*8`($i)
+ stfd f30,`-2*8`($i)
+ stfd f31,`-1*8`($i)
+
+ addi $tp,$sp,`$FRAME+$TRANSFER+8+64`
+ li $i,-64
+ add $nap_d,$tp,$num
+ and $nap_d,$nap_d,$i ; align to 64 bytes
+ ; nap_d is off by 1, because it's used with stfdu/lfdu
+ addi $nap_d,$nap_d,-8
+ srwi $j,$num,`3+1` ; counter register, num/2
+ addi $j,$j,-1
+ addi $tp,$sp,`$FRAME+$TRANSFER-8`
+ li $carry,0
+ mtctr $j
+___
+
+$code.=<<___ if ($SIZE_T==8);
+ ld $a0,0($ap) ; pull ap[0] value
+ ld $t3,0($bp) ; bp[0]
+ ld $n0,0($n0) ; pull n0[0] value
+
+ mulld $t7,$a0,$t3 ; ap[0]*bp[0]
+ ; transfer bp[0] to FPU as 4x16-bit values
+ extrdi $t0,$t3,16,48
+ extrdi $t1,$t3,16,32
+ extrdi $t2,$t3,16,16
+ extrdi $t3,$t3,16,0
+ std $t0,`$FRAME+0`($sp)
+ std $t1,`$FRAME+8`($sp)
+ std $t2,`$FRAME+16`($sp)
+ std $t3,`$FRAME+24`($sp)
+
+ mulld $t7,$t7,$n0 ; tp[0]*n0
+ ; transfer (ap[0]*bp[0])*n0 to FPU as 4x16-bit values
+ extrdi $t4,$t7,16,48
+ extrdi $t5,$t7,16,32
+ extrdi $t6,$t7,16,16
+ extrdi $t7,$t7,16,0
+ std $t4,`$FRAME+32`($sp)
+ std $t5,`$FRAME+40`($sp)
+ std $t6,`$FRAME+48`($sp)
+ std $t7,`$FRAME+56`($sp)
+
+ extrdi $t0,$a0,32,32 ; lwz $t0,4($ap)
+ extrdi $t1,$a0,32,0 ; lwz $t1,0($ap)
+ lwz $t2,`12^$LITTLE_ENDIAN`($ap) ; load a[1] as 32-bit word pair
+ lwz $t3,`8^$LITTLE_ENDIAN`($ap)
+ lwz $t4,`4^$LITTLE_ENDIAN`($np) ; load n[0] as 32-bit word pair
+ lwz $t5,`0^$LITTLE_ENDIAN`($np)
+ lwz $t6,`12^$LITTLE_ENDIAN`($np) ; load n[1] as 32-bit word pair
+ lwz $t7,`8^$LITTLE_ENDIAN`($np)
+___
+$code.=<<___ if ($SIZE_T==4);
+ lwz $a0,0($ap) ; pull ap[0,1] value
+ mr $n1,$n0
+ lwz $a1,4($ap)
+ li $c1,0
+ lwz $t1,0($bp) ; bp[0,1]
+ lwz $t3,4($bp)
+ lwz $n0,0($n1) ; pull n0[0,1] value
+ lwz $n1,4($n1)
+
+ mullw $t4,$a0,$t1 ; mulld ap[0]*bp[0]
+ mulhwu $t5,$a0,$t1
+ mullw $t6,$a1,$t1
+ mullw $t7,$a0,$t3
+ add $t5,$t5,$t6
+ add $t5,$t5,$t7
+ ; transfer bp[0] to FPU as 4x16-bit values
+ extrwi $t0,$t1,16,16
+ extrwi $t1,$t1,16,0
+ extrwi $t2,$t3,16,16
+ extrwi $t3,$t3,16,0
+ std $t0,`$FRAME+0`($sp) ; yes, std in 32-bit build
+ std $t1,`$FRAME+8`($sp)
+ std $t2,`$FRAME+16`($sp)
+ std $t3,`$FRAME+24`($sp)
+
+ mullw $t0,$t4,$n0 ; mulld tp[0]*n0
+ mulhwu $t1,$t4,$n0
+ mullw $t2,$t5,$n0
+ mullw $t3,$t4,$n1
+ add $t1,$t1,$t2
+ add $t1,$t1,$t3
+ ; transfer (ap[0]*bp[0])*n0 to FPU as 4x16-bit values
+ extrwi $t4,$t0,16,16
+ extrwi $t5,$t0,16,0
+ extrwi $t6,$t1,16,16
+ extrwi $t7,$t1,16,0
+ std $t4,`$FRAME+32`($sp) ; yes, std in 32-bit build
+ std $t5,`$FRAME+40`($sp)
+ std $t6,`$FRAME+48`($sp)
+ std $t7,`$FRAME+56`($sp)
+
+ mr $t0,$a0 ; lwz $t0,0($ap)
+ mr $t1,$a1 ; lwz $t1,4($ap)
+ lwz $t2,8($ap) ; load a[j..j+3] as 32-bit word pairs
+ lwz $t3,12($ap)
+ lwz $t4,0($np) ; load n[j..j+3] as 32-bit word pairs
+ lwz $t5,4($np)
+ lwz $t6,8($np)
+ lwz $t7,12($np)
+___
+$code.=<<___;
+ lfd $ba,`$FRAME+0`($sp)
+ lfd $bb,`$FRAME+8`($sp)
+ lfd $bc,`$FRAME+16`($sp)
+ lfd $bd,`$FRAME+24`($sp)
+ lfd $na,`$FRAME+32`($sp)
+ lfd $nb,`$FRAME+40`($sp)
+ lfd $nc,`$FRAME+48`($sp)
+ lfd $nd,`$FRAME+56`($sp)
+ std $t0,`$FRAME+64`($sp) ; yes, std even in 32-bit build
+ std $t1,`$FRAME+72`($sp)
+ std $t2,`$FRAME+80`($sp)
+ std $t3,`$FRAME+88`($sp)
+ std $t4,`$FRAME+96`($sp)
+ std $t5,`$FRAME+104`($sp)
+ std $t6,`$FRAME+112`($sp)
+ std $t7,`$FRAME+120`($sp)
+ fcfid $ba,$ba
+ fcfid $bb,$bb
+ fcfid $bc,$bc
+ fcfid $bd,$bd
+ fcfid $na,$na
+ fcfid $nb,$nb
+ fcfid $nc,$nc
+ fcfid $nd,$nd
+
+ lfd $A0,`$FRAME+64`($sp)
+ lfd $A1,`$FRAME+72`($sp)
+ lfd $A2,`$FRAME+80`($sp)
+ lfd $A3,`$FRAME+88`($sp)
+ lfd $N0,`$FRAME+96`($sp)
+ lfd $N1,`$FRAME+104`($sp)
+ lfd $N2,`$FRAME+112`($sp)
+ lfd $N3,`$FRAME+120`($sp)
+ fcfid $A0,$A0
+ fcfid $A1,$A1
+ fcfid $A2,$A2
+ fcfid $A3,$A3
+ fcfid $N0,$N0
+ fcfid $N1,$N1
+ fcfid $N2,$N2
+ fcfid $N3,$N3
+ addi $ap,$ap,16
+ addi $np,$np,16
+
+ fmul $T1a,$A1,$ba
+ fmul $T1b,$A1,$bb
+ stfd $A0,8($nap_d) ; save a[j] in double format
+ stfd $A1,16($nap_d)
+ fmul $T2a,$A2,$ba
+ fmul $T2b,$A2,$bb
+ stfd $A2,24($nap_d) ; save a[j+1] in double format
+ stfd $A3,32($nap_d)
+ fmul $T3a,$A3,$ba
+ fmul $T3b,$A3,$bb
+ stfd $N0,40($nap_d) ; save n[j] in double format
+ stfd $N1,48($nap_d)
+ fmul $T0a,$A0,$ba
+ fmul $T0b,$A0,$bb
+ stfd $N2,56($nap_d) ; save n[j+1] in double format
+ stfdu $N3,64($nap_d)
+
+ fmadd $T1a,$A0,$bc,$T1a
+ fmadd $T1b,$A0,$bd,$T1b
+ fmadd $T2a,$A1,$bc,$T2a
+ fmadd $T2b,$A1,$bd,$T2b
+ fmadd $T3a,$A2,$bc,$T3a
+ fmadd $T3b,$A2,$bd,$T3b
+ fmul $dota,$A3,$bc
+ fmul $dotb,$A3,$bd
+
+ fmadd $T1a,$N1,$na,$T1a
+ fmadd $T1b,$N1,$nb,$T1b
+ fmadd $T2a,$N2,$na,$T2a
+ fmadd $T2b,$N2,$nb,$T2b
+ fmadd $T3a,$N3,$na,$T3a
+ fmadd $T3b,$N3,$nb,$T3b
+ fmadd $T0a,$N0,$na,$T0a
+ fmadd $T0b,$N0,$nb,$T0b
+
+ fmadd $T1a,$N0,$nc,$T1a
+ fmadd $T1b,$N0,$nd,$T1b
+ fmadd $T2a,$N1,$nc,$T2a
+ fmadd $T2b,$N1,$nd,$T2b
+ fmadd $T3a,$N2,$nc,$T3a
+ fmadd $T3b,$N2,$nd,$T3b
+ fmadd $dota,$N3,$nc,$dota
+ fmadd $dotb,$N3,$nd,$dotb
+
+ fctid $T0a,$T0a
+ fctid $T0b,$T0b
+ fctid $T1a,$T1a
+ fctid $T1b,$T1b
+ fctid $T2a,$T2a
+ fctid $T2b,$T2b
+ fctid $T3a,$T3a
+ fctid $T3b,$T3b
+
+ stfd $T0a,`$FRAME+0`($sp)
+ stfd $T0b,`$FRAME+8`($sp)
+ stfd $T1a,`$FRAME+16`($sp)
+ stfd $T1b,`$FRAME+24`($sp)
+ stfd $T2a,`$FRAME+32`($sp)
+ stfd $T2b,`$FRAME+40`($sp)
+ stfd $T3a,`$FRAME+48`($sp)
+ stfd $T3b,`$FRAME+56`($sp)
+
+.align 5
+L1st:
+___
+$code.=<<___ if ($SIZE_T==8);
+ lwz $t0,`4^$LITTLE_ENDIAN`($ap) ; load a[j] as 32-bit word pair
+ lwz $t1,`0^$LITTLE_ENDIAN`($ap)
+ lwz $t2,`12^$LITTLE_ENDIAN`($ap) ; load a[j+1] as 32-bit word pair
+ lwz $t3,`8^$LITTLE_ENDIAN`($ap)
+ lwz $t4,`4^$LITTLE_ENDIAN`($np) ; load n[j] as 32-bit word pair
+ lwz $t5,`0^$LITTLE_ENDIAN`($np)
+ lwz $t6,`12^$LITTLE_ENDIAN`($np) ; load n[j+1] as 32-bit word pair
+ lwz $t7,`8^$LITTLE_ENDIAN`($np)
+___
+$code.=<<___ if ($SIZE_T==4);
+ lwz $t0,0($ap) ; load a[j..j+3] as 32-bit word pairs
+ lwz $t1,4($ap)
+ lwz $t2,8($ap)
+ lwz $t3,12($ap)
+ lwz $t4,0($np) ; load n[j..j+3] as 32-bit word pairs
+ lwz $t5,4($np)
+ lwz $t6,8($np)
+ lwz $t7,12($np)
+___
+$code.=<<___;
+ std $t0,`$FRAME+64`($sp) ; yes, std even in 32-bit build
+ std $t1,`$FRAME+72`($sp)
+ std $t2,`$FRAME+80`($sp)
+ std $t3,`$FRAME+88`($sp)
+ std $t4,`$FRAME+96`($sp)
+ std $t5,`$FRAME+104`($sp)
+ std $t6,`$FRAME+112`($sp)
+ std $t7,`$FRAME+120`($sp)
+___
+if ($SIZE_T==8 or $flavour =~ /osx/) {
+$code.=<<___;
+ ld $t0,`$FRAME+0`($sp)
+ ld $t1,`$FRAME+8`($sp)
+ ld $t2,`$FRAME+16`($sp)
+ ld $t3,`$FRAME+24`($sp)
+ ld $t4,`$FRAME+32`($sp)
+ ld $t5,`$FRAME+40`($sp)
+ ld $t6,`$FRAME+48`($sp)
+ ld $t7,`$FRAME+56`($sp)
+___
+} else {
+$code.=<<___;
+ lwz $t1,`$FRAME+0^$LITTLE_ENDIAN`($sp)
+ lwz $t0,`$FRAME+4^$LITTLE_ENDIAN`($sp)
+ lwz $t3,`$FRAME+8^$LITTLE_ENDIAN`($sp)
+ lwz $t2,`$FRAME+12^$LITTLE_ENDIAN`($sp)
+ lwz $t5,`$FRAME+16^$LITTLE_ENDIAN`($sp)
+ lwz $t4,`$FRAME+20^$LITTLE_ENDIAN`($sp)
+ lwz $t7,`$FRAME+24^$LITTLE_ENDIAN`($sp)
+ lwz $t6,`$FRAME+28^$LITTLE_ENDIAN`($sp)
+___
+}
+$code.=<<___;
+ lfd $A0,`$FRAME+64`($sp)
+ lfd $A1,`$FRAME+72`($sp)
+ lfd $A2,`$FRAME+80`($sp)
+ lfd $A3,`$FRAME+88`($sp)
+ lfd $N0,`$FRAME+96`($sp)
+ lfd $N1,`$FRAME+104`($sp)
+ lfd $N2,`$FRAME+112`($sp)
+ lfd $N3,`$FRAME+120`($sp)
+ fcfid $A0,$A0
+ fcfid $A1,$A1
+ fcfid $A2,$A2
+ fcfid $A3,$A3
+ fcfid $N0,$N0
+ fcfid $N1,$N1
+ fcfid $N2,$N2
+ fcfid $N3,$N3
+ addi $ap,$ap,16
+ addi $np,$np,16
+
+ fmul $T1a,$A1,$ba
+ fmul $T1b,$A1,$bb
+ fmul $T2a,$A2,$ba
+ fmul $T2b,$A2,$bb
+ stfd $A0,8($nap_d) ; save a[j] in double format
+ stfd $A1,16($nap_d)
+ fmul $T3a,$A3,$ba
+ fmul $T3b,$A3,$bb
+ fmadd $T0a,$A0,$ba,$dota
+ fmadd $T0b,$A0,$bb,$dotb
+ stfd $A2,24($nap_d) ; save a[j+1] in double format
+ stfd $A3,32($nap_d)
+___
+if ($SIZE_T==8 or $flavour =~ /osx/) {
+$code.=<<___;
+ fmadd $T1a,$A0,$bc,$T1a
+ fmadd $T1b,$A0,$bd,$T1b
+ fmadd $T2a,$A1,$bc,$T2a
+ fmadd $T2b,$A1,$bd,$T2b
+ stfd $N0,40($nap_d) ; save n[j] in double format
+ stfd $N1,48($nap_d)
+ fmadd $T3a,$A2,$bc,$T3a
+ fmadd $T3b,$A2,$bd,$T3b
+ add $t0,$t0,$carry ; can not overflow
+ fmul $dota,$A3,$bc
+ fmul $dotb,$A3,$bd
+ stfd $N2,56($nap_d) ; save n[j+1] in double format
+ stfdu $N3,64($nap_d)
+ srdi $carry,$t0,16
+ add $t1,$t1,$carry
+ srdi $carry,$t1,16
+
+ fmadd $T1a,$N1,$na,$T1a
+ fmadd $T1b,$N1,$nb,$T1b
+ insrdi $t0,$t1,16,32
+ fmadd $T2a,$N2,$na,$T2a
+ fmadd $T2b,$N2,$nb,$T2b
+ add $t2,$t2,$carry
+ fmadd $T3a,$N3,$na,$T3a
+ fmadd $T3b,$N3,$nb,$T3b
+ srdi $carry,$t2,16
+ fmadd $T0a,$N0,$na,$T0a
+ fmadd $T0b,$N0,$nb,$T0b
+ insrdi $t0,$t2,16,16
+ add $t3,$t3,$carry
+ srdi $carry,$t3,16
+
+ fmadd $T1a,$N0,$nc,$T1a
+ fmadd $T1b,$N0,$nd,$T1b
+ insrdi $t0,$t3,16,0 ; 0..63 bits
+ fmadd $T2a,$N1,$nc,$T2a
+ fmadd $T2b,$N1,$nd,$T2b
+ add $t4,$t4,$carry
+ fmadd $T3a,$N2,$nc,$T3a
+ fmadd $T3b,$N2,$nd,$T3b
+ srdi $carry,$t4,16
+ fmadd $dota,$N3,$nc,$dota
+ fmadd $dotb,$N3,$nd,$dotb
+ add $t5,$t5,$carry
+ srdi $carry,$t5,16
+ insrdi $t4,$t5,16,32
+
+ fctid $T0a,$T0a
+ fctid $T0b,$T0b
+ add $t6,$t6,$carry
+ fctid $T1a,$T1a
+ fctid $T1b,$T1b
+ srdi $carry,$t6,16
+ fctid $T2a,$T2a
+ fctid $T2b,$T2b
+ insrdi $t4,$t6,16,16
+ fctid $T3a,$T3a
+ fctid $T3b,$T3b
+ add $t7,$t7,$carry
+ insrdi $t4,$t7,16,0 ; 64..127 bits
+ srdi $carry,$t7,16 ; upper 33 bits
+
+ stfd $T0a,`$FRAME+0`($sp)
+ stfd $T0b,`$FRAME+8`($sp)
+ stfd $T1a,`$FRAME+16`($sp)
+ stfd $T1b,`$FRAME+24`($sp)
+ stfd $T2a,`$FRAME+32`($sp)
+ stfd $T2b,`$FRAME+40`($sp)
+ stfd $T3a,`$FRAME+48`($sp)
+ stfd $T3b,`$FRAME+56`($sp)
+ std $t0,8($tp) ; tp[j-1]
+ stdu $t4,16($tp) ; tp[j]
+___
+} else {
+$code.=<<___;
+ fmadd $T1a,$A0,$bc,$T1a
+ fmadd $T1b,$A0,$bd,$T1b
+ addc $t0,$t0,$carry
+ adde $t1,$t1,$c1
+ srwi $carry,$t0,16
+ fmadd $T2a,$A1,$bc,$T2a
+ fmadd $T2b,$A1,$bd,$T2b
+ stfd $N0,40($nap_d) ; save n[j] in double format
+ stfd $N1,48($nap_d)
+ srwi $c1,$t1,16
+ insrwi $carry,$t1,16,0
+ fmadd $T3a,$A2,$bc,$T3a
+ fmadd $T3b,$A2,$bd,$T3b
+ addc $t2,$t2,$carry
+ adde $t3,$t3,$c1
+ srwi $carry,$t2,16
+ fmul $dota,$A3,$bc
+ fmul $dotb,$A3,$bd
+ stfd $N2,56($nap_d) ; save n[j+1] in double format
+ stfdu $N3,64($nap_d)
+ insrwi $t0,$t2,16,0 ; 0..31 bits
+ srwi $c1,$t3,16
+ insrwi $carry,$t3,16,0
+
+ fmadd $T1a,$N1,$na,$T1a
+ fmadd $T1b,$N1,$nb,$T1b
+ lwz $t3,`$FRAME+32^$LITTLE_ENDIAN`($sp) ; permuted $t1
+ lwz $t2,`$FRAME+36^$LITTLE_ENDIAN`($sp) ; permuted $t0
+ addc $t4,$t4,$carry
+ adde $t5,$t5,$c1
+ srwi $carry,$t4,16
+ fmadd $T2a,$N2,$na,$T2a
+ fmadd $T2b,$N2,$nb,$T2b
+ srwi $c1,$t5,16
+ insrwi $carry,$t5,16,0
+ fmadd $T3a,$N3,$na,$T3a
+ fmadd $T3b,$N3,$nb,$T3b
+ addc $t6,$t6,$carry
+ adde $t7,$t7,$c1
+ srwi $carry,$t6,16
+ fmadd $T0a,$N0,$na,$T0a
+ fmadd $T0b,$N0,$nb,$T0b
+ insrwi $t4,$t6,16,0 ; 32..63 bits
+ srwi $c1,$t7,16
+ insrwi $carry,$t7,16,0
+
+ fmadd $T1a,$N0,$nc,$T1a
+ fmadd $T1b,$N0,$nd,$T1b
+ lwz $t7,`$FRAME+40^$LITTLE_ENDIAN`($sp) ; permuted $t3
+ lwz $t6,`$FRAME+44^$LITTLE_ENDIAN`($sp) ; permuted $t2
+ addc $t2,$t2,$carry
+ adde $t3,$t3,$c1
+ srwi $carry,$t2,16
+ fmadd $T2a,$N1,$nc,$T2a
+ fmadd $T2b,$N1,$nd,$T2b
+ stw $t0,12($tp) ; tp[j-1]
+ stw $t4,8($tp)
+ srwi $c1,$t3,16
+ insrwi $carry,$t3,16,0
+ fmadd $T3a,$N2,$nc,$T3a
+ fmadd $T3b,$N2,$nd,$T3b
+ lwz $t1,`$FRAME+48^$LITTLE_ENDIAN`($sp) ; permuted $t5
+ lwz $t0,`$FRAME+52^$LITTLE_ENDIAN`($sp) ; permuted $t4
+ addc $t6,$t6,$carry
+ adde $t7,$t7,$c1
+ srwi $carry,$t6,16
+ fmadd $dota,$N3,$nc,$dota
+ fmadd $dotb,$N3,$nd,$dotb
+ insrwi $t2,$t6,16,0 ; 64..95 bits
+ srwi $c1,$t7,16
+ insrwi $carry,$t7,16,0
+
+ fctid $T0a,$T0a
+ fctid $T0b,$T0b
+ lwz $t5,`$FRAME+56^$LITTLE_ENDIAN`($sp) ; permuted $t7
+ lwz $t4,`$FRAME+60^$LITTLE_ENDIAN`($sp) ; permuted $t6
+ addc $t0,$t0,$carry
+ adde $t1,$t1,$c1
+ srwi $carry,$t0,16
+ fctid $T1a,$T1a
+ fctid $T1b,$T1b
+ srwi $c1,$t1,16
+ insrwi $carry,$t1,16,0
+ fctid $T2a,$T2a
+ fctid $T2b,$T2b
+ addc $t4,$t4,$carry
+ adde $t5,$t5,$c1
+ srwi $carry,$t4,16
+ fctid $T3a,$T3a
+ fctid $T3b,$T3b
+ insrwi $t0,$t4,16,0 ; 96..127 bits
+ srwi $c1,$t5,16
+ insrwi $carry,$t5,16,0
+
+ stfd $T0a,`$FRAME+0`($sp)
+ stfd $T0b,`$FRAME+8`($sp)
+ stfd $T1a,`$FRAME+16`($sp)
+ stfd $T1b,`$FRAME+24`($sp)
+ stfd $T2a,`$FRAME+32`($sp)
+ stfd $T2b,`$FRAME+40`($sp)
+ stfd $T3a,`$FRAME+48`($sp)
+ stfd $T3b,`$FRAME+56`($sp)
+ stw $t2,20($tp) ; tp[j]
+ stwu $t0,16($tp)
+___
+}
+$code.=<<___;
+ bdnz L1st
+
+ fctid $dota,$dota
+ fctid $dotb,$dotb
+___
+if ($SIZE_T==8 or $flavour =~ /osx/) {
+$code.=<<___;
+ ld $t0,`$FRAME+0`($sp)
+ ld $t1,`$FRAME+8`($sp)
+ ld $t2,`$FRAME+16`($sp)
+ ld $t3,`$FRAME+24`($sp)
+ ld $t4,`$FRAME+32`($sp)
+ ld $t5,`$FRAME+40`($sp)
+ ld $t6,`$FRAME+48`($sp)
+ ld $t7,`$FRAME+56`($sp)
+ stfd $dota,`$FRAME+64`($sp)
+ stfd $dotb,`$FRAME+72`($sp)
+
+ add $t0,$t0,$carry ; can not overflow
+ srdi $carry,$t0,16
+ add $t1,$t1,$carry
+ srdi $carry,$t1,16
+ insrdi $t0,$t1,16,32
+ add $t2,$t2,$carry
+ srdi $carry,$t2,16
+ insrdi $t0,$t2,16,16
+ add $t3,$t3,$carry
+ srdi $carry,$t3,16
+ insrdi $t0,$t3,16,0 ; 0..63 bits
+ add $t4,$t4,$carry
+ srdi $carry,$t4,16
+ add $t5,$t5,$carry
+ srdi $carry,$t5,16
+ insrdi $t4,$t5,16,32
+ add $t6,$t6,$carry
+ srdi $carry,$t6,16
+ insrdi $t4,$t6,16,16
+ add $t7,$t7,$carry
+ insrdi $t4,$t7,16,0 ; 64..127 bits
+ srdi $carry,$t7,16 ; upper 33 bits
+ ld $t6,`$FRAME+64`($sp)
+ ld $t7,`$FRAME+72`($sp)
+
+ std $t0,8($tp) ; tp[j-1]
+ stdu $t4,16($tp) ; tp[j]
+
+ add $t6,$t6,$carry ; can not overflow
+ srdi $carry,$t6,16
+ add $t7,$t7,$carry
+ insrdi $t6,$t7,48,0
+ srdi $ovf,$t7,48
+ std $t6,8($tp) ; tp[num-1]
+___
+} else {
+$code.=<<___;
+ lwz $t1,`$FRAME+0^$LITTLE_ENDIAN`($sp)
+ lwz $t0,`$FRAME+4^$LITTLE_ENDIAN`($sp)
+ lwz $t3,`$FRAME+8^$LITTLE_ENDIAN`($sp)
+ lwz $t2,`$FRAME+12^$LITTLE_ENDIAN`($sp)
+ lwz $t5,`$FRAME+16^$LITTLE_ENDIAN`($sp)
+ lwz $t4,`$FRAME+20^$LITTLE_ENDIAN`($sp)
+ lwz $t7,`$FRAME+24^$LITTLE_ENDIAN`($sp)
+ lwz $t6,`$FRAME+28^$LITTLE_ENDIAN`($sp)
+ stfd $dota,`$FRAME+64`($sp)
+ stfd $dotb,`$FRAME+72`($sp)
+
+ addc $t0,$t0,$carry
+ adde $t1,$t1,$c1
+ srwi $carry,$t0,16
+ insrwi $carry,$t1,16,0
+ srwi $c1,$t1,16
+ addc $t2,$t2,$carry
+ adde $t3,$t3,$c1
+ srwi $carry,$t2,16
+ insrwi $t0,$t2,16,0 ; 0..31 bits
+ insrwi $carry,$t3,16,0
+ srwi $c1,$t3,16
+ addc $t4,$t4,$carry
+ adde $t5,$t5,$c1
+ srwi $carry,$t4,16
+ insrwi $carry,$t5,16,0
+ srwi $c1,$t5,16
+ addc $t6,$t6,$carry
+ adde $t7,$t7,$c1
+ srwi $carry,$t6,16
+ insrwi $t4,$t6,16,0 ; 32..63 bits
+ insrwi $carry,$t7,16,0
+ srwi $c1,$t7,16
+ stw $t0,12($tp) ; tp[j-1]
+ stw $t4,8($tp)
+
+ lwz $t3,`$FRAME+32^$LITTLE_ENDIAN`($sp) ; permuted $t1
+ lwz $t2,`$FRAME+36^$LITTLE_ENDIAN`($sp) ; permuted $t0
+ lwz $t7,`$FRAME+40^$LITTLE_ENDIAN`($sp) ; permuted $t3
+ lwz $t6,`$FRAME+44^$LITTLE_ENDIAN`($sp) ; permuted $t2
+ lwz $t1,`$FRAME+48^$LITTLE_ENDIAN`($sp) ; permuted $t5
+ lwz $t0,`$FRAME+52^$LITTLE_ENDIAN`($sp) ; permuted $t4
+ lwz $t5,`$FRAME+56^$LITTLE_ENDIAN`($sp) ; permuted $t7
+ lwz $t4,`$FRAME+60^$LITTLE_ENDIAN`($sp) ; permuted $t6
+
+ addc $t2,$t2,$carry
+ adde $t3,$t3,$c1
+ srwi $carry,$t2,16
+ insrwi $carry,$t3,16,0
+ srwi $c1,$t3,16
+ addc $t6,$t6,$carry
+ adde $t7,$t7,$c1
+ srwi $carry,$t6,16
+ insrwi $t2,$t6,16,0 ; 64..95 bits
+ insrwi $carry,$t7,16,0
+ srwi $c1,$t7,16
+ addc $t0,$t0,$carry
+ adde $t1,$t1,$c1
+ srwi $carry,$t0,16
+ insrwi $carry,$t1,16,0
+ srwi $c1,$t1,16
+ addc $t4,$t4,$carry
+ adde $t5,$t5,$c1
+ srwi $carry,$t4,16
+ insrwi $t0,$t4,16,0 ; 96..127 bits
+ insrwi $carry,$t5,16,0
+ srwi $c1,$t5,16
+ stw $t2,20($tp) ; tp[j]
+ stwu $t0,16($tp)
+
+ lwz $t7,`$FRAME+64^$LITTLE_ENDIAN`($sp)
+ lwz $t6,`$FRAME+68^$LITTLE_ENDIAN`($sp)
+ lwz $t5,`$FRAME+72^$LITTLE_ENDIAN`($sp)
+ lwz $t4,`$FRAME+76^$LITTLE_ENDIAN`($sp)
+
+ addc $t6,$t6,$carry
+ adde $t7,$t7,$c1
+ srwi $carry,$t6,16
+ insrwi $carry,$t7,16,0
+ srwi $c1,$t7,16
+ addc $t4,$t4,$carry
+ adde $t5,$t5,$c1
+
+ insrwi $t6,$t4,16,0
+ srwi $t4,$t4,16
+ insrwi $t4,$t5,16,0
+ srwi $ovf,$t5,16
+ stw $t6,12($tp) ; tp[num-1]
+ stw $t4,8($tp)
+___
+}
+$code.=<<___;
+ slwi $t7,$num,2
+ subf $nap_d,$t7,$nap_d ; rewind pointer
+
+ li $i,8 ; i=1
+.align 5
+Louter:
+ addi $tp,$sp,`$FRAME+$TRANSFER`
+ li $carry,0
+ mtctr $j
+___
+$code.=<<___ if ($SIZE_T==8);
+ ldx $t3,$bp,$i ; bp[i]
+
+ ld $t6,`$FRAME+$TRANSFER+8`($sp) ; tp[0]
+ mulld $t7,$a0,$t3 ; ap[0]*bp[i]
+ add $t7,$t7,$t6 ; ap[0]*bp[i]+tp[0]
+ ; transfer bp[i] to FPU as 4x16-bit values
+ extrdi $t0,$t3,16,48
+ extrdi $t1,$t3,16,32
+ extrdi $t2,$t3,16,16
+ extrdi $t3,$t3,16,0
+ std $t0,`$FRAME+0`($sp)
+ std $t1,`$FRAME+8`($sp)
+ std $t2,`$FRAME+16`($sp)
+ std $t3,`$FRAME+24`($sp)
+
+ mulld $t7,$t7,$n0 ; tp[0]*n0
+ ; transfer (ap[0]*bp[i]+tp[0])*n0 to FPU as 4x16-bit values
+ extrdi $t4,$t7,16,48
+ extrdi $t5,$t7,16,32
+ extrdi $t6,$t7,16,16
+ extrdi $t7,$t7,16,0
+ std $t4,`$FRAME+32`($sp)
+ std $t5,`$FRAME+40`($sp)
+ std $t6,`$FRAME+48`($sp)
+ std $t7,`$FRAME+56`($sp)
+___
+$code.=<<___ if ($SIZE_T==4);
+ add $t0,$bp,$i
+ li $c1,0
+ lwz $t1,0($t0) ; bp[i,i+1]
+ lwz $t3,4($t0)
+
+ mullw $t4,$a0,$t1 ; ap[0]*bp[i]
+ lwz $t0,`$FRAME+$TRANSFER+8+4`($sp) ; tp[0]
+ mulhwu $t5,$a0,$t1
+ lwz $t2,`$FRAME+$TRANSFER+8`($sp) ; tp[0]
+ mullw $t6,$a1,$t1
+ mullw $t7,$a0,$t3
+ add $t5,$t5,$t6
+ add $t5,$t5,$t7
+ addc $t4,$t4,$t0 ; ap[0]*bp[i]+tp[0]
+ adde $t5,$t5,$t2
+ ; transfer bp[i] to FPU as 4x16-bit values
+ extrwi $t0,$t1,16,16
+ extrwi $t1,$t1,16,0
+ extrwi $t2,$t3,16,16
+ extrwi $t3,$t3,16,0
+ std $t0,`$FRAME+0`($sp) ; yes, std in 32-bit build
+ std $t1,`$FRAME+8`($sp)
+ std $t2,`$FRAME+16`($sp)
+ std $t3,`$FRAME+24`($sp)
+
+ mullw $t0,$t4,$n0 ; mulld tp[0]*n0
+ mulhwu $t1,$t4,$n0
+ mullw $t2,$t5,$n0
+ mullw $t3,$t4,$n1
+ add $t1,$t1,$t2
+ add $t1,$t1,$t3
+ ; transfer (ap[0]*bp[i]+tp[0])*n0 to FPU as 4x16-bit values
+ extrwi $t4,$t0,16,16
+ extrwi $t5,$t0,16,0
+ extrwi $t6,$t1,16,16
+ extrwi $t7,$t1,16,0
+ std $t4,`$FRAME+32`($sp) ; yes, std in 32-bit build
+ std $t5,`$FRAME+40`($sp)
+ std $t6,`$FRAME+48`($sp)
+ std $t7,`$FRAME+56`($sp)
+___
+$code.=<<___;
+ lfd $A0,8($nap_d) ; load a[j] in double format
+ lfd $A1,16($nap_d)
+ lfd $A2,24($nap_d) ; load a[j+1] in double format
+ lfd $A3,32($nap_d)
+ lfd $N0,40($nap_d) ; load n[j] in double format
+ lfd $N1,48($nap_d)
+ lfd $N2,56($nap_d) ; load n[j+1] in double format
+ lfdu $N3,64($nap_d)
+
+ lfd $ba,`$FRAME+0`($sp)
+ lfd $bb,`$FRAME+8`($sp)
+ lfd $bc,`$FRAME+16`($sp)
+ lfd $bd,`$FRAME+24`($sp)
+ lfd $na,`$FRAME+32`($sp)
+ lfd $nb,`$FRAME+40`($sp)
+ lfd $nc,`$FRAME+48`($sp)
+ lfd $nd,`$FRAME+56`($sp)
+
+ fcfid $ba,$ba
+ fcfid $bb,$bb
+ fcfid $bc,$bc
+ fcfid $bd,$bd
+ fcfid $na,$na
+ fcfid $nb,$nb
+ fcfid $nc,$nc
+ fcfid $nd,$nd
+
+ fmul $T1a,$A1,$ba
+ fmul $T1b,$A1,$bb
+ fmul $T2a,$A2,$ba
+ fmul $T2b,$A2,$bb
+ fmul $T3a,$A3,$ba
+ fmul $T3b,$A3,$bb
+ fmul $T0a,$A0,$ba
+ fmul $T0b,$A0,$bb
+
+ fmadd $T1a,$A0,$bc,$T1a
+ fmadd $T1b,$A0,$bd,$T1b
+ fmadd $T2a,$A1,$bc,$T2a
+ fmadd $T2b,$A1,$bd,$T2b
+ fmadd $T3a,$A2,$bc,$T3a
+ fmadd $T3b,$A2,$bd,$T3b
+ fmul $dota,$A3,$bc
+ fmul $dotb,$A3,$bd
+
+ fmadd $T1a,$N1,$na,$T1a
+ fmadd $T1b,$N1,$nb,$T1b
+ lfd $A0,8($nap_d) ; load a[j] in double format
+ lfd $A1,16($nap_d)
+ fmadd $T2a,$N2,$na,$T2a
+ fmadd $T2b,$N2,$nb,$T2b
+ lfd $A2,24($nap_d) ; load a[j+1] in double format
+ lfd $A3,32($nap_d)
+ fmadd $T3a,$N3,$na,$T3a
+ fmadd $T3b,$N3,$nb,$T3b
+ fmadd $T0a,$N0,$na,$T0a
+ fmadd $T0b,$N0,$nb,$T0b
+
+ fmadd $T1a,$N0,$nc,$T1a
+ fmadd $T1b,$N0,$nd,$T1b
+ fmadd $T2a,$N1,$nc,$T2a
+ fmadd $T2b,$N1,$nd,$T2b
+ fmadd $T3a,$N2,$nc,$T3a
+ fmadd $T3b,$N2,$nd,$T3b
+ fmadd $dota,$N3,$nc,$dota
+ fmadd $dotb,$N3,$nd,$dotb
+
+ fctid $T0a,$T0a
+ fctid $T0b,$T0b
+ fctid $T1a,$T1a
+ fctid $T1b,$T1b
+ fctid $T2a,$T2a
+ fctid $T2b,$T2b
+ fctid $T3a,$T3a
+ fctid $T3b,$T3b
+
+ stfd $T0a,`$FRAME+0`($sp)
+ stfd $T0b,`$FRAME+8`($sp)
+ stfd $T1a,`$FRAME+16`($sp)
+ stfd $T1b,`$FRAME+24`($sp)
+ stfd $T2a,`$FRAME+32`($sp)
+ stfd $T2b,`$FRAME+40`($sp)
+ stfd $T3a,`$FRAME+48`($sp)
+ stfd $T3b,`$FRAME+56`($sp)
+
+.align 5
+Linner:
+ fmul $T1a,$A1,$ba
+ fmul $T1b,$A1,$bb
+ fmul $T2a,$A2,$ba
+ fmul $T2b,$A2,$bb
+ lfd $N0,40($nap_d) ; load n[j] in double format
+ lfd $N1,48($nap_d)
+ fmul $T3a,$A3,$ba
+ fmul $T3b,$A3,$bb
+ fmadd $T0a,$A0,$ba,$dota
+ fmadd $T0b,$A0,$bb,$dotb
+ lfd $N2,56($nap_d) ; load n[j+1] in double format
+ lfdu $N3,64($nap_d)
+
+ fmadd $T1a,$A0,$bc,$T1a
+ fmadd $T1b,$A0,$bd,$T1b
+ fmadd $T2a,$A1,$bc,$T2a
+ fmadd $T2b,$A1,$bd,$T2b
+ lfd $A0,8($nap_d) ; load a[j] in double format
+ lfd $A1,16($nap_d)
+ fmadd $T3a,$A2,$bc,$T3a
+ fmadd $T3b,$A2,$bd,$T3b
+ fmul $dota,$A3,$bc
+ fmul $dotb,$A3,$bd
+ lfd $A2,24($nap_d) ; load a[j+1] in double format
+ lfd $A3,32($nap_d)
+___
+if ($SIZE_T==8 or $flavour =~ /osx/) {
+$code.=<<___;
+ fmadd $T1a,$N1,$na,$T1a
+ fmadd $T1b,$N1,$nb,$T1b
+ ld $t0,`$FRAME+0`($sp)
+ ld $t1,`$FRAME+8`($sp)
+ fmadd $T2a,$N2,$na,$T2a
+ fmadd $T2b,$N2,$nb,$T2b
+ ld $t2,`$FRAME+16`($sp)
+ ld $t3,`$FRAME+24`($sp)
+ fmadd $T3a,$N3,$na,$T3a
+ fmadd $T3b,$N3,$nb,$T3b
+ add $t0,$t0,$carry ; can not overflow
+ ld $t4,`$FRAME+32`($sp)
+ ld $t5,`$FRAME+40`($sp)
+ fmadd $T0a,$N0,$na,$T0a
+ fmadd $T0b,$N0,$nb,$T0b
+ srdi $carry,$t0,16
+ add $t1,$t1,$carry
+ srdi $carry,$t1,16
+ ld $t6,`$FRAME+48`($sp)
+ ld $t7,`$FRAME+56`($sp)
+
+ fmadd $T1a,$N0,$nc,$T1a
+ fmadd $T1b,$N0,$nd,$T1b
+ insrdi $t0,$t1,16,32
+ ld $t1,8($tp) ; tp[j]
+ fmadd $T2a,$N1,$nc,$T2a
+ fmadd $T2b,$N1,$nd,$T2b
+ add $t2,$t2,$carry
+ fmadd $T3a,$N2,$nc,$T3a
+ fmadd $T3b,$N2,$nd,$T3b
+ srdi $carry,$t2,16
+ insrdi $t0,$t2,16,16
+ fmadd $dota,$N3,$nc,$dota
+ fmadd $dotb,$N3,$nd,$dotb
+ add $t3,$t3,$carry
+ ldu $t2,16($tp) ; tp[j+1]
+ srdi $carry,$t3,16
+ insrdi $t0,$t3,16,0 ; 0..63 bits
+ add $t4,$t4,$carry
+
+ fctid $T0a,$T0a
+ fctid $T0b,$T0b
+ srdi $carry,$t4,16
+ fctid $T1a,$T1a
+ fctid $T1b,$T1b
+ add $t5,$t5,$carry
+ fctid $T2a,$T2a
+ fctid $T2b,$T2b
+ srdi $carry,$t5,16
+ insrdi $t4,$t5,16,32
+ fctid $T3a,$T3a
+ fctid $T3b,$T3b
+ add $t6,$t6,$carry
+ srdi $carry,$t6,16
+ insrdi $t4,$t6,16,16
+
+ stfd $T0a,`$FRAME+0`($sp)
+ stfd $T0b,`$FRAME+8`($sp)
+ add $t7,$t7,$carry
+ addc $t3,$t0,$t1
+___
+$code.=<<___ if ($SIZE_T==4); # adjust XER[CA]
+ extrdi $t0,$t0,32,0
+ extrdi $t1,$t1,32,0
+ adde $t0,$t0,$t1
+___
+$code.=<<___;
+ stfd $T1a,`$FRAME+16`($sp)
+ stfd $T1b,`$FRAME+24`($sp)
+ insrdi $t4,$t7,16,0 ; 64..127 bits
+ srdi $carry,$t7,16 ; upper 33 bits
+ stfd $T2a,`$FRAME+32`($sp)
+ stfd $T2b,`$FRAME+40`($sp)
+ adde $t5,$t4,$t2
+___
+$code.=<<___ if ($SIZE_T==4); # adjust XER[CA]
+ extrdi $t4,$t4,32,0
+ extrdi $t2,$t2,32,0
+ adde $t4,$t4,$t2
+___
+$code.=<<___;
+ stfd $T3a,`$FRAME+48`($sp)
+ stfd $T3b,`$FRAME+56`($sp)
+ addze $carry,$carry
+ std $t3,-16($tp) ; tp[j-1]
+ std $t5,-8($tp) ; tp[j]
+___
+} else {
+$code.=<<___;
+ fmadd $T1a,$N1,$na,$T1a
+ fmadd $T1b,$N1,$nb,$T1b
+ lwz $t1,`$FRAME+0^$LITTLE_ENDIAN`($sp)
+ lwz $t0,`$FRAME+4^$LITTLE_ENDIAN`($sp)
+ fmadd $T2a,$N2,$na,$T2a
+ fmadd $T2b,$N2,$nb,$T2b
+ lwz $t3,`$FRAME+8^$LITTLE_ENDIAN`($sp)
+ lwz $t2,`$FRAME+12^$LITTLE_ENDIAN`($sp)
+ fmadd $T3a,$N3,$na,$T3a
+ fmadd $T3b,$N3,$nb,$T3b
+ lwz $t5,`$FRAME+16^$LITTLE_ENDIAN`($sp)
+ lwz $t4,`$FRAME+20^$LITTLE_ENDIAN`($sp)
+ addc $t0,$t0,$carry
+ adde $t1,$t1,$c1
+ srwi $carry,$t0,16
+ fmadd $T0a,$N0,$na,$T0a
+ fmadd $T0b,$N0,$nb,$T0b
+ lwz $t7,`$FRAME+24^$LITTLE_ENDIAN`($sp)
+ lwz $t6,`$FRAME+28^$LITTLE_ENDIAN`($sp)
+ srwi $c1,$t1,16
+ insrwi $carry,$t1,16,0
+
+ fmadd $T1a,$N0,$nc,$T1a
+ fmadd $T1b,$N0,$nd,$T1b
+ addc $t2,$t2,$carry
+ adde $t3,$t3,$c1
+ srwi $carry,$t2,16
+ fmadd $T2a,$N1,$nc,$T2a
+ fmadd $T2b,$N1,$nd,$T2b
+ insrwi $t0,$t2,16,0 ; 0..31 bits
+ srwi $c1,$t3,16
+ insrwi $carry,$t3,16,0
+ fmadd $T3a,$N2,$nc,$T3a
+ fmadd $T3b,$N2,$nd,$T3b
+ lwz $t2,12($tp) ; tp[j]
+ lwz $t3,8($tp)
+ addc $t4,$t4,$carry
+ adde $t5,$t5,$c1
+ srwi $carry,$t4,16
+ fmadd $dota,$N3,$nc,$dota
+ fmadd $dotb,$N3,$nd,$dotb
+ srwi $c1,$t5,16
+ insrwi $carry,$t5,16,0
+
+ fctid $T0a,$T0a
+ addc $t6,$t6,$carry
+ adde $t7,$t7,$c1
+ srwi $carry,$t6,16
+ fctid $T0b,$T0b
+ insrwi $t4,$t6,16,0 ; 32..63 bits
+ srwi $c1,$t7,16
+ insrwi $carry,$t7,16,0
+ fctid $T1a,$T1a
+ addc $t0,$t0,$t2
+ adde $t4,$t4,$t3
+ lwz $t3,`$FRAME+32^$LITTLE_ENDIAN`($sp) ; permuted $t1
+ lwz $t2,`$FRAME+36^$LITTLE_ENDIAN`($sp) ; permuted $t0
+ fctid $T1b,$T1b
+ addze $carry,$carry
+ addze $c1,$c1
+ stw $t0,4($tp) ; tp[j-1]
+ stw $t4,0($tp)
+ fctid $T2a,$T2a
+ addc $t2,$t2,$carry
+ adde $t3,$t3,$c1
+ srwi $carry,$t2,16
+ lwz $t7,`$FRAME+40^$LITTLE_ENDIAN`($sp) ; permuted $t3
+ lwz $t6,`$FRAME+44^$LITTLE_ENDIAN`($sp) ; permuted $t2
+ fctid $T2b,$T2b
+ srwi $c1,$t3,16
+ insrwi $carry,$t3,16,0
+ lwz $t1,`$FRAME+48^$LITTLE_ENDIAN`($sp) ; permuted $t5
+ lwz $t0,`$FRAME+52^$LITTLE_ENDIAN`($sp) ; permuted $t4
+ fctid $T3a,$T3a
+ addc $t6,$t6,$carry
+ adde $t7,$t7,$c1
+ srwi $carry,$t6,16
+ lwz $t5,`$FRAME+56^$LITTLE_ENDIAN`($sp) ; permuted $t7
+ lwz $t4,`$FRAME+60^$LITTLE_ENDIAN`($sp) ; permuted $t6
+ fctid $T3b,$T3b
+
+ insrwi $t2,$t6,16,0 ; 64..95 bits
+ insrwi $carry,$t7,16,0
+ srwi $c1,$t7,16
+ lwz $t6,20($tp)
+ lwzu $t7,16($tp)
+ addc $t0,$t0,$carry
+ stfd $T0a,`$FRAME+0`($sp)
+ adde $t1,$t1,$c1
+ srwi $carry,$t0,16
+ stfd $T0b,`$FRAME+8`($sp)
+ insrwi $carry,$t1,16,0
+ srwi $c1,$t1,16
+ addc $t4,$t4,$carry
+ stfd $T1a,`$FRAME+16`($sp)
+ adde $t5,$t5,$c1
+ srwi $carry,$t4,16
+ insrwi $t0,$t4,16,0 ; 96..127 bits
+ stfd $T1b,`$FRAME+24`($sp)
+ insrwi $carry,$t5,16,0
+ srwi $c1,$t5,16
+
+ addc $t2,$t2,$t6
+ stfd $T2a,`$FRAME+32`($sp)
+ adde $t0,$t0,$t7
+ stfd $T2b,`$FRAME+40`($sp)
+ addze $carry,$carry
+ stfd $T3a,`$FRAME+48`($sp)
+ addze $c1,$c1
+ stfd $T3b,`$FRAME+56`($sp)
+ stw $t2,-4($tp) ; tp[j]
+ stw $t0,-8($tp)
+___
+}
+$code.=<<___;
+ bdnz Linner
+
+ fctid $dota,$dota
+ fctid $dotb,$dotb
+___
+if ($SIZE_T==8 or $flavour =~ /osx/) {
+$code.=<<___;
+ ld $t0,`$FRAME+0`($sp)
+ ld $t1,`$FRAME+8`($sp)
+ ld $t2,`$FRAME+16`($sp)
+ ld $t3,`$FRAME+24`($sp)
+ ld $t4,`$FRAME+32`($sp)
+ ld $t5,`$FRAME+40`($sp)
+ ld $t6,`$FRAME+48`($sp)
+ ld $t7,`$FRAME+56`($sp)
+ stfd $dota,`$FRAME+64`($sp)
+ stfd $dotb,`$FRAME+72`($sp)
+
+ add $t0,$t0,$carry ; can not overflow
+ srdi $carry,$t0,16
+ add $t1,$t1,$carry
+ srdi $carry,$t1,16
+ insrdi $t0,$t1,16,32
+ add $t2,$t2,$carry
+ ld $t1,8($tp) ; tp[j]
+ srdi $carry,$t2,16
+ insrdi $t0,$t2,16,16
+ add $t3,$t3,$carry
+ ldu $t2,16($tp) ; tp[j+1]
+ srdi $carry,$t3,16
+ insrdi $t0,$t3,16,0 ; 0..63 bits
+ add $t4,$t4,$carry
+ srdi $carry,$t4,16
+ add $t5,$t5,$carry
+ srdi $carry,$t5,16
+ insrdi $t4,$t5,16,32
+ add $t6,$t6,$carry
+ srdi $carry,$t6,16
+ insrdi $t4,$t6,16,16
+ add $t7,$t7,$carry
+ insrdi $t4,$t7,16,0 ; 64..127 bits
+ srdi $carry,$t7,16 ; upper 33 bits
+ ld $t6,`$FRAME+64`($sp)
+ ld $t7,`$FRAME+72`($sp)
+
+ addc $t3,$t0,$t1
+___
+$code.=<<___ if ($SIZE_T==4); # adjust XER[CA]
+ extrdi $t0,$t0,32,0
+ extrdi $t1,$t1,32,0
+ adde $t0,$t0,$t1
+___
+$code.=<<___;
+ adde $t5,$t4,$t2
+___
+$code.=<<___ if ($SIZE_T==4); # adjust XER[CA]
+ extrdi $t4,$t4,32,0
+ extrdi $t2,$t2,32,0
+ adde $t4,$t4,$t2
+___
+$code.=<<___;
+ addze $carry,$carry
+
+ std $t3,-16($tp) ; tp[j-1]
+ std $t5,-8($tp) ; tp[j]
+
+ add $carry,$carry,$ovf ; consume upmost overflow
+ add $t6,$t6,$carry ; can not overflow
+ srdi $carry,$t6,16
+ add $t7,$t7,$carry
+ insrdi $t6,$t7,48,0
+ srdi $ovf,$t7,48
+ std $t6,0($tp) ; tp[num-1]
+___
+} else {
+$code.=<<___;
+ lwz $t1,`$FRAME+0^$LITTLE_ENDIAN`($sp)
+ lwz $t0,`$FRAME+4^$LITTLE_ENDIAN`($sp)
+ lwz $t3,`$FRAME+8^$LITTLE_ENDIAN`($sp)
+ lwz $t2,`$FRAME+12^$LITTLE_ENDIAN`($sp)
+ lwz $t5,`$FRAME+16^$LITTLE_ENDIAN`($sp)
+ lwz $t4,`$FRAME+20^$LITTLE_ENDIAN`($sp)
+ lwz $t7,`$FRAME+24^$LITTLE_ENDIAN`($sp)
+ lwz $t6,`$FRAME+28^$LITTLE_ENDIAN`($sp)
+ stfd $dota,`$FRAME+64`($sp)
+ stfd $dotb,`$FRAME+72`($sp)
+
+ addc $t0,$t0,$carry
+ adde $t1,$t1,$c1
+ srwi $carry,$t0,16
+ insrwi $carry,$t1,16,0
+ srwi $c1,$t1,16
+ addc $t2,$t2,$carry
+ adde $t3,$t3,$c1
+ srwi $carry,$t2,16
+ insrwi $t0,$t2,16,0 ; 0..31 bits
+ lwz $t2,12($tp) ; tp[j]
+ insrwi $carry,$t3,16,0
+ srwi $c1,$t3,16
+ lwz $t3,8($tp)
+ addc $t4,$t4,$carry
+ adde $t5,$t5,$c1
+ srwi $carry,$t4,16
+ insrwi $carry,$t5,16,0
+ srwi $c1,$t5,16
+ addc $t6,$t6,$carry
+ adde $t7,$t7,$c1
+ srwi $carry,$t6,16
+ insrwi $t4,$t6,16,0 ; 32..63 bits
+ insrwi $carry,$t7,16,0
+ srwi $c1,$t7,16
+
+ addc $t0,$t0,$t2
+ adde $t4,$t4,$t3
+ addze $carry,$carry
+ addze $c1,$c1
+ stw $t0,4($tp) ; tp[j-1]
+ stw $t4,0($tp)
+
+ lwz $t3,`$FRAME+32^$LITTLE_ENDIAN`($sp) ; permuted $t1
+ lwz $t2,`$FRAME+36^$LITTLE_ENDIAN`($sp) ; permuted $t0
+ lwz $t7,`$FRAME+40^$LITTLE_ENDIAN`($sp) ; permuted $t3
+ lwz $t6,`$FRAME+44^$LITTLE_ENDIAN`($sp) ; permuted $t2
+ lwz $t1,`$FRAME+48^$LITTLE_ENDIAN`($sp) ; permuted $t5
+ lwz $t0,`$FRAME+52^$LITTLE_ENDIAN`($sp) ; permuted $t4
+ lwz $t5,`$FRAME+56^$LITTLE_ENDIAN`($sp) ; permuted $t7
+ lwz $t4,`$FRAME+60^$LITTLE_ENDIAN`($sp) ; permuted $t6
+
+ addc $t2,$t2,$carry
+ adde $t3,$t3,$c1
+ srwi $carry,$t2,16
+ insrwi $carry,$t3,16,0
+ srwi $c1,$t3,16
+ addc $t6,$t6,$carry
+ adde $t7,$t7,$c1
+ srwi $carry,$t6,16
+ insrwi $t2,$t6,16,0 ; 64..95 bits
+ lwz $t6,20($tp)
+ insrwi $carry,$t7,16,0
+ srwi $c1,$t7,16
+ lwzu $t7,16($tp)
+ addc $t0,$t0,$carry
+ adde $t1,$t1,$c1
+ srwi $carry,$t0,16
+ insrwi $carry,$t1,16,0
+ srwi $c1,$t1,16
+ addc $t4,$t4,$carry
+ adde $t5,$t5,$c1
+ srwi $carry,$t4,16
+ insrwi $t0,$t4,16,0 ; 96..127 bits
+ insrwi $carry,$t5,16,0
+ srwi $c1,$t5,16
+
+ addc $t2,$t2,$t6
+ adde $t0,$t0,$t7
+ lwz $t7,`$FRAME+64^$LITTLE_ENDIAN`($sp)
+ lwz $t6,`$FRAME+68^$LITTLE_ENDIAN`($sp)
+ addze $carry,$carry
+ addze $c1,$c1
+ lwz $t5,`$FRAME+72^$LITTLE_ENDIAN`($sp)
+ lwz $t4,`$FRAME+76^$LITTLE_ENDIAN`($sp)
+
+ addc $t6,$t6,$carry
+ adde $t7,$t7,$c1
+ stw $t2,-4($tp) ; tp[j]
+ stw $t0,-8($tp)
+ addc $t6,$t6,$ovf
+ addze $t7,$t7
+ srwi $carry,$t6,16
+ insrwi $carry,$t7,16,0
+ srwi $c1,$t7,16
+ addc $t4,$t4,$carry
+ adde $t5,$t5,$c1
+
+ insrwi $t6,$t4,16,0
+ srwi $t4,$t4,16
+ insrwi $t4,$t5,16,0
+ srwi $ovf,$t5,16
+ stw $t6,4($tp) ; tp[num-1]
+ stw $t4,0($tp)
+___
+}
+$code.=<<___;
+ slwi $t7,$num,2
+ addi $i,$i,8
+ subf $nap_d,$t7,$nap_d ; rewind pointer
+ cmpw $i,$num
+ blt- Louter
+___
+
+$code.=<<___ if ($SIZE_T==8);
+ subf $np,$num,$np ; rewind np
+ addi $j,$j,1 ; restore counter
+ subfc $i,$i,$i ; j=0 and "clear" XER[CA]
+ addi $tp,$sp,`$FRAME+$TRANSFER+8`
+ addi $t4,$sp,`$FRAME+$TRANSFER+16`
+ addi $t5,$np,8
+ addi $t6,$rp,8
+ mtctr $j
+
+.align 4
+Lsub: ldx $t0,$tp,$i
+ ldx $t1,$np,$i
+ ldx $t2,$t4,$i
+ ldx $t3,$t5,$i
+ subfe $t0,$t1,$t0 ; tp[j]-np[j]
+ subfe $t2,$t3,$t2 ; tp[j+1]-np[j+1]
+ stdx $t0,$rp,$i
+ stdx $t2,$t6,$i
+ addi $i,$i,16
+ bdnz Lsub
+
+ li $i,0
+ subfe $ovf,$i,$ovf ; handle upmost overflow bit
+ mtctr $j
+
+.align 4
+Lcopy: ; conditional copy
+ ldx $t0,$tp,$i
+ ldx $t1,$t4,$i
+ ldx $t2,$rp,$i
+ ldx $t3,$t6,$i
+ std $i,8($nap_d) ; zap nap_d
+ std $i,16($nap_d)
+ std $i,24($nap_d)
+ std $i,32($nap_d)
+ std $i,40($nap_d)
+ std $i,48($nap_d)
+ std $i,56($nap_d)
+ stdu $i,64($nap_d)
+ and $t0,$t0,$ovf
+ and $t1,$t1,$ovf
+ andc $t2,$t2,$ovf
+ andc $t3,$t3,$ovf
+ or $t0,$t0,$t2
+ or $t1,$t1,$t3
+ stdx $t0,$rp,$i
+ stdx $t1,$t6,$i
+ stdx $i,$tp,$i ; zap tp at once
+ stdx $i,$t4,$i
+ addi $i,$i,16
+ bdnz Lcopy
+___
+$code.=<<___ if ($SIZE_T==4);
+ subf $np,$num,$np ; rewind np
+ addi $j,$j,1 ; restore counter
+ subfc $i,$i,$i ; j=0 and "clear" XER[CA]
+ addi $tp,$sp,`$FRAME+$TRANSFER`
+ addi $np,$np,-4
+ addi $rp,$rp,-4
+ addi $ap,$sp,`$FRAME+$TRANSFER+4`
+ mtctr $j
+
+.align 4
+Lsub: lwz $t0,12($tp) ; load tp[j..j+3] in 64-bit word order
+ lwz $t1,8($tp)
+ lwz $t2,20($tp)
+ lwzu $t3,16($tp)
+ lwz $t4,4($np) ; load np[j..j+3] in 32-bit word order
+ lwz $t5,8($np)
+ lwz $t6,12($np)
+ lwzu $t7,16($np)
+ subfe $t4,$t4,$t0 ; tp[j]-np[j]
+ stw $t0,4($ap) ; save tp[j..j+3] in 32-bit word order
+ subfe $t5,$t5,$t1 ; tp[j+1]-np[j+1]
+ stw $t1,8($ap)
+ subfe $t6,$t6,$t2 ; tp[j+2]-np[j+2]
+ stw $t2,12($ap)
+ subfe $t7,$t7,$t3 ; tp[j+3]-np[j+3]
+ stwu $t3,16($ap)
+ stw $t4,4($rp)
+ stw $t5,8($rp)
+ stw $t6,12($rp)
+ stwu $t7,16($rp)
+ bdnz Lsub
+
+ li $i,0
+ subfe $ovf,$i,$ovf ; handle upmost overflow bit
+ addi $ap,$sp,`$FRAME+$TRANSFER+4`
+ subf $rp,$num,$rp ; rewind rp
+ addi $tp,$sp,`$FRAME+$TRANSFER`
+ mtctr $j
+
+.align 4
+Lcopy: ; conditional copy
+ lwz $t0,4($ap)
+ lwz $t1,8($ap)
+ lwz $t2,12($ap)
+ lwzu $t3,16($ap)
+ lwz $t4,4($rp)
+ lwz $t5,8($rp)
+ lwz $t6,12($rp)
+ lwz $t7,16($rp)
+ std $i,8($nap_d) ; zap nap_d
+ std $i,16($nap_d)
+ std $i,24($nap_d)
+ std $i,32($nap_d)
+ std $i,40($nap_d)
+ std $i,48($nap_d)
+ std $i,56($nap_d)
+ stdu $i,64($nap_d)
+ and $t0,$t0,$ovf
+ and $t1,$t1,$ovf
+ and $t2,$t2,$ovf
+ and $t3,$t3,$ovf
+ andc $t4,$t4,$ovf
+ andc $t5,$t5,$ovf
+ andc $t6,$t6,$ovf
+ andc $t7,$t7,$ovf
+ or $t0,$t0,$t4
+ or $t1,$t1,$t5
+ or $t2,$t2,$t6
+ or $t3,$t3,$t7
+ stw $t0,4($rp)
+ stw $t1,8($rp)
+ stw $t2,12($rp)
+ stwu $t3,16($rp)
+ std $i,8($tp) ; zap tp at once
+ stdu $i,16($tp)
+ bdnz Lcopy
+___
+
+$code.=<<___;
+ $POP $i,0($sp)
+ li r3,1 ; signal "handled"
+ $POP r19,`-12*8-13*$SIZE_T`($i)
+ $POP r20,`-12*8-12*$SIZE_T`($i)
+ $POP r21,`-12*8-11*$SIZE_T`($i)
+ $POP r22,`-12*8-10*$SIZE_T`($i)
+ $POP r23,`-12*8-9*$SIZE_T`($i)
+ $POP r24,`-12*8-8*$SIZE_T`($i)
+ $POP r25,`-12*8-7*$SIZE_T`($i)
+ $POP r26,`-12*8-6*$SIZE_T`($i)
+ $POP r27,`-12*8-5*$SIZE_T`($i)
+ $POP r28,`-12*8-4*$SIZE_T`($i)
+ $POP r29,`-12*8-3*$SIZE_T`($i)
+ $POP r30,`-12*8-2*$SIZE_T`($i)
+ $POP r31,`-12*8-1*$SIZE_T`($i)
+ lfd f20,`-12*8`($i)
+ lfd f21,`-11*8`($i)
+ lfd f22,`-10*8`($i)
+ lfd f23,`-9*8`($i)
+ lfd f24,`-8*8`($i)
+ lfd f25,`-7*8`($i)
+ lfd f26,`-6*8`($i)
+ lfd f27,`-5*8`($i)
+ lfd f28,`-4*8`($i)
+ lfd f29,`-3*8`($i)
+ lfd f30,`-2*8`($i)
+ lfd f31,`-1*8`($i)
+ mr $sp,$i
+ blr
+ .long 0
+ .byte 0,12,4,0,0x8c,13,6,0
+ .long 0
+.size .$fname,.-.$fname
+
+.asciz "Montgomery Multiplication for PPC64, CRYPTOGAMS by <appro\@openssl.org>"
+___
+
+$code =~ s/\`([^\`]*)\`/eval $1/gem;
+print $code;
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/rsaz-avx2.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/rsaz-avx2.pl
new file mode 100755
index 0000000..0be771f
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/rsaz-avx2.pl
@@ -0,0 +1,1982 @@
+#! /usr/bin/env perl
+# Copyright 2013-2020 The OpenSSL Project Authors. All Rights Reserved.
+# Copyright (c) 2012, Intel Corporation. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+#
+# Originally written by Shay Gueron (1, 2), and Vlad Krasnov (1)
+# (1) Intel Corporation, Israel Development Center, Haifa, Israel
+# (2) University of Haifa, Israel
+#
+# References:
+# [1] S. Gueron, V. Krasnov: "Software Implementation of Modular
+# Exponentiation, Using Advanced Vector Instructions Architectures",
+# F. Ozbudak and F. Rodriguez-Henriquez (Eds.): WAIFI 2012, LNCS 7369,
+# pp. 119?135, 2012. Springer-Verlag Berlin Heidelberg 2012
+# [2] S. Gueron: "Efficient Software Implementations of Modular
+# Exponentiation", Journal of Cryptographic Engineering 2:31-43 (2012).
+# [3] S. Gueron, V. Krasnov: "Speeding up Big-numbers Squaring",IEEE
+# Proceedings of 9th International Conference on Information Technology:
+# New Generations (ITNG 2012), pp.821-823 (2012)
+# [4] S. Gueron, V. Krasnov: "[PATCH] Efficient and side channel analysis
+# resistant 1024-bit modular exponentiation, for optimizing RSA2048
+# on AVX2 capable x86_64 platforms",
+# http://rt.openssl.org/Ticket/Display.html?id=2850&user=guest&pass=guest
+#
+# +13% improvement over original submission by <appro@openssl.org>
+#
+# rsa2048 sign/sec OpenSSL 1.0.1 scalar(*) this
+# 2.3GHz Haswell 621 765/+23% 1113/+79%
+# 2.3GHz Broadwell(**) 688 1200(***)/+74% 1120/+63%
+#
+# (*) if system doesn't support AVX2, for reference purposes;
+# (**) scaled to 2.3GHz to simplify comparison;
+# (***) scalar AD*X code is faster than AVX2 and is preferred code
+# path for Broadwell;
+
+$flavour = shift;
+$output = shift;
+if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
+
+$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or
+( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or
+die "can't locate x86_64-xlate.pl";
+
+if (`$ENV{CC} -Wa,-v -c -o /dev/null -x assembler /dev/null 2>&1`
+ =~ /GNU assembler version ([2-9]\.[0-9]+)/) {
+ $avx = ($1>=2.19) + ($1>=2.22);
+ $addx = ($1>=2.23);
+}
+
+if (!$avx && $win64 && ($flavour =~ /nasm/ || $ENV{ASM} =~ /nasm/) &&
+ `nasm -v 2>&1` =~ /NASM version ([2-9]\.[0-9]+)/) {
+ $avx = ($1>=2.09) + ($1>=2.10);
+ $addx = ($1>=2.10);
+}
+
+if (!$avx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
+ `ml64 2>&1` =~ /Version ([0-9]+)\./) {
+ $avx = ($1>=10) + ($1>=11);
+ $addx = ($1>=11);
+}
+
+if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|based on LLVM) ([0-9]+)\.([0-9]+)/) {
+ my $ver = $2 + $3/100.0; # 3.1->3.01, 3.10->3.10
+ $avx = ($ver>=3.0) + ($ver>=3.01);
+ $addx = ($ver>=3.03);
+}
+
+open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\"";
+*STDOUT = *OUT;
+
+if ($avx>1) {{{
+{ # void AMS_WW(
+my $rp="%rdi"; # BN_ULONG *rp,
+my $ap="%rsi"; # const BN_ULONG *ap,
+my $np="%rdx"; # const BN_ULONG *np,
+my $n0="%ecx"; # const BN_ULONG n0,
+my $rep="%r8d"; # int repeat);
+
+# The registers that hold the accumulated redundant result
+# The AMM works on 1024 bit operands, and redundant word size is 29
+# Therefore: ceil(1024/29)/4 = 9
+my $ACC0="%ymm0";
+my $ACC1="%ymm1";
+my $ACC2="%ymm2";
+my $ACC3="%ymm3";
+my $ACC4="%ymm4";
+my $ACC5="%ymm5";
+my $ACC6="%ymm6";
+my $ACC7="%ymm7";
+my $ACC8="%ymm8";
+my $ACC9="%ymm9";
+# Registers that hold the broadcasted words of bp, currently used
+my $B1="%ymm10";
+my $B2="%ymm11";
+# Registers that hold the broadcasted words of Y, currently used
+my $Y1="%ymm12";
+my $Y2="%ymm13";
+# Helper registers
+my $TEMP1="%ymm14";
+my $AND_MASK="%ymm15";
+# alu registers that hold the first words of the ACC
+my $r0="%r9";
+my $r1="%r10";
+my $r2="%r11";
+my $r3="%r12";
+
+my $i="%r14d"; # loop counter
+my $tmp = "%r15";
+
+my $FrameSize=32*18+32*8; # place for A^2 and 2*A
+
+my $aap=$r0;
+my $tp0="%rbx";
+my $tp1=$r3;
+my $tpa=$tmp;
+
+$np="%r13"; # reassigned argument
+
+$code.=<<___;
+.text
+
+.globl rsaz_1024_sqr_avx2
+.type rsaz_1024_sqr_avx2,\@function,5
+.align 64
+rsaz_1024_sqr_avx2: # 702 cycles, 14% faster than rsaz_1024_mul_avx2
+.cfi_startproc
+ lea (%rsp), %rax
+.cfi_def_cfa_register %rax
+ push %rbx
+.cfi_push %rbx
+ push %rbp
+.cfi_push %rbp
+ push %r12
+.cfi_push %r12
+ push %r13
+.cfi_push %r13
+ push %r14
+.cfi_push %r14
+ push %r15
+.cfi_push %r15
+ vzeroupper
+___
+$code.=<<___ if ($win64);
+ lea -0xa8(%rsp),%rsp
+ vmovaps %xmm6,-0xd8(%rax)
+ vmovaps %xmm7,-0xc8(%rax)
+ vmovaps %xmm8,-0xb8(%rax)
+ vmovaps %xmm9,-0xa8(%rax)
+ vmovaps %xmm10,-0x98(%rax)
+ vmovaps %xmm11,-0x88(%rax)
+ vmovaps %xmm12,-0x78(%rax)
+ vmovaps %xmm13,-0x68(%rax)
+ vmovaps %xmm14,-0x58(%rax)
+ vmovaps %xmm15,-0x48(%rax)
+.Lsqr_1024_body:
+___
+$code.=<<___;
+ mov %rax,%rbp
+.cfi_def_cfa_register %rbp
+ mov %rdx, $np # reassigned argument
+ sub \$$FrameSize, %rsp
+ mov $np, $tmp
+ sub \$-128, $rp # size optimization
+ sub \$-128, $ap
+ sub \$-128, $np
+
+ and \$4095, $tmp # see if $np crosses page
+ add \$32*10, $tmp
+ shr \$12, $tmp
+ vpxor $ACC9,$ACC9,$ACC9
+ jz .Lsqr_1024_no_n_copy
+
+ # unaligned 256-bit load that crosses page boundary can
+ # cause >2x performance degradation here, so if $np does
+ # cross page boundary, copy it to stack and make sure stack
+ # frame doesn't...
+ sub \$32*10,%rsp
+ vmovdqu 32*0-128($np), $ACC0
+ and \$-2048, %rsp
+ vmovdqu 32*1-128($np), $ACC1
+ vmovdqu 32*2-128($np), $ACC2
+ vmovdqu 32*3-128($np), $ACC3
+ vmovdqu 32*4-128($np), $ACC4
+ vmovdqu 32*5-128($np), $ACC5
+ vmovdqu 32*6-128($np), $ACC6
+ vmovdqu 32*7-128($np), $ACC7
+ vmovdqu 32*8-128($np), $ACC8
+ lea $FrameSize+128(%rsp),$np
+ vmovdqu $ACC0, 32*0-128($np)
+ vmovdqu $ACC1, 32*1-128($np)
+ vmovdqu $ACC2, 32*2-128($np)
+ vmovdqu $ACC3, 32*3-128($np)
+ vmovdqu $ACC4, 32*4-128($np)
+ vmovdqu $ACC5, 32*5-128($np)
+ vmovdqu $ACC6, 32*6-128($np)
+ vmovdqu $ACC7, 32*7-128($np)
+ vmovdqu $ACC8, 32*8-128($np)
+ vmovdqu $ACC9, 32*9-128($np) # $ACC9 is zero
+
+.Lsqr_1024_no_n_copy:
+ and \$-1024, %rsp
+
+ vmovdqu 32*1-128($ap), $ACC1
+ vmovdqu 32*2-128($ap), $ACC2
+ vmovdqu 32*3-128($ap), $ACC3
+ vmovdqu 32*4-128($ap), $ACC4
+ vmovdqu 32*5-128($ap), $ACC5
+ vmovdqu 32*6-128($ap), $ACC6
+ vmovdqu 32*7-128($ap), $ACC7
+ vmovdqu 32*8-128($ap), $ACC8
+
+ lea 192(%rsp), $tp0 # 64+128=192
+ vmovdqu .Land_mask(%rip), $AND_MASK
+ jmp .LOOP_GRANDE_SQR_1024
+
+.align 32
+.LOOP_GRANDE_SQR_1024:
+ lea 32*18+128(%rsp), $aap # size optimization
+ lea 448(%rsp), $tp1 # 64+128+256=448
+
+ # the squaring is performed as described in Variant B of
+ # "Speeding up Big-Number Squaring", so start by calculating
+ # the A*2=A+A vector
+ vpaddq $ACC1, $ACC1, $ACC1
+ vpbroadcastq 32*0-128($ap), $B1
+ vpaddq $ACC2, $ACC2, $ACC2
+ vmovdqa $ACC1, 32*0-128($aap)
+ vpaddq $ACC3, $ACC3, $ACC3
+ vmovdqa $ACC2, 32*1-128($aap)
+ vpaddq $ACC4, $ACC4, $ACC4
+ vmovdqa $ACC3, 32*2-128($aap)
+ vpaddq $ACC5, $ACC5, $ACC5
+ vmovdqa $ACC4, 32*3-128($aap)
+ vpaddq $ACC6, $ACC6, $ACC6
+ vmovdqa $ACC5, 32*4-128($aap)
+ vpaddq $ACC7, $ACC7, $ACC7
+ vmovdqa $ACC6, 32*5-128($aap)
+ vpaddq $ACC8, $ACC8, $ACC8
+ vmovdqa $ACC7, 32*6-128($aap)
+ vpxor $ACC9, $ACC9, $ACC9
+ vmovdqa $ACC8, 32*7-128($aap)
+
+ vpmuludq 32*0-128($ap), $B1, $ACC0
+ vpbroadcastq 32*1-128($ap), $B2
+ vmovdqu $ACC9, 32*9-192($tp0) # zero upper half
+ vpmuludq $B1, $ACC1, $ACC1
+ vmovdqu $ACC9, 32*10-448($tp1)
+ vpmuludq $B1, $ACC2, $ACC2
+ vmovdqu $ACC9, 32*11-448($tp1)
+ vpmuludq $B1, $ACC3, $ACC3
+ vmovdqu $ACC9, 32*12-448($tp1)
+ vpmuludq $B1, $ACC4, $ACC4
+ vmovdqu $ACC9, 32*13-448($tp1)
+ vpmuludq $B1, $ACC5, $ACC5
+ vmovdqu $ACC9, 32*14-448($tp1)
+ vpmuludq $B1, $ACC6, $ACC6
+ vmovdqu $ACC9, 32*15-448($tp1)
+ vpmuludq $B1, $ACC7, $ACC7
+ vmovdqu $ACC9, 32*16-448($tp1)
+ vpmuludq $B1, $ACC8, $ACC8
+ vpbroadcastq 32*2-128($ap), $B1
+ vmovdqu $ACC9, 32*17-448($tp1)
+
+ mov $ap, $tpa
+ mov \$4, $i
+ jmp .Lsqr_entry_1024
+___
+$TEMP0=$Y1;
+$TEMP2=$Y2;
+$code.=<<___;
+.align 32
+.LOOP_SQR_1024:
+ vpbroadcastq 32*1-128($tpa), $B2
+ vpmuludq 32*0-128($ap), $B1, $ACC0
+ vpaddq 32*0-192($tp0), $ACC0, $ACC0
+ vpmuludq 32*0-128($aap), $B1, $ACC1
+ vpaddq 32*1-192($tp0), $ACC1, $ACC1
+ vpmuludq 32*1-128($aap), $B1, $ACC2
+ vpaddq 32*2-192($tp0), $ACC2, $ACC2
+ vpmuludq 32*2-128($aap), $B1, $ACC3
+ vpaddq 32*3-192($tp0), $ACC3, $ACC3
+ vpmuludq 32*3-128($aap), $B1, $ACC4
+ vpaddq 32*4-192($tp0), $ACC4, $ACC4
+ vpmuludq 32*4-128($aap), $B1, $ACC5
+ vpaddq 32*5-192($tp0), $ACC5, $ACC5
+ vpmuludq 32*5-128($aap), $B1, $ACC6
+ vpaddq 32*6-192($tp0), $ACC6, $ACC6
+ vpmuludq 32*6-128($aap), $B1, $ACC7
+ vpaddq 32*7-192($tp0), $ACC7, $ACC7
+ vpmuludq 32*7-128($aap), $B1, $ACC8
+ vpbroadcastq 32*2-128($tpa), $B1
+ vpaddq 32*8-192($tp0), $ACC8, $ACC8
+.Lsqr_entry_1024:
+ vmovdqu $ACC0, 32*0-192($tp0)
+ vmovdqu $ACC1, 32*1-192($tp0)
+
+ vpmuludq 32*1-128($ap), $B2, $TEMP0
+ vpaddq $TEMP0, $ACC2, $ACC2
+ vpmuludq 32*1-128($aap), $B2, $TEMP1
+ vpaddq $TEMP1, $ACC3, $ACC3
+ vpmuludq 32*2-128($aap), $B2, $TEMP2
+ vpaddq $TEMP2, $ACC4, $ACC4
+ vpmuludq 32*3-128($aap), $B2, $TEMP0
+ vpaddq $TEMP0, $ACC5, $ACC5
+ vpmuludq 32*4-128($aap), $B2, $TEMP1
+ vpaddq $TEMP1, $ACC6, $ACC6
+ vpmuludq 32*5-128($aap), $B2, $TEMP2
+ vpaddq $TEMP2, $ACC7, $ACC7
+ vpmuludq 32*6-128($aap), $B2, $TEMP0
+ vpaddq $TEMP0, $ACC8, $ACC8
+ vpmuludq 32*7-128($aap), $B2, $ACC0
+ vpbroadcastq 32*3-128($tpa), $B2
+ vpaddq 32*9-192($tp0), $ACC0, $ACC0
+
+ vmovdqu $ACC2, 32*2-192($tp0)
+ vmovdqu $ACC3, 32*3-192($tp0)
+
+ vpmuludq 32*2-128($ap), $B1, $TEMP2
+ vpaddq $TEMP2, $ACC4, $ACC4
+ vpmuludq 32*2-128($aap), $B1, $TEMP0
+ vpaddq $TEMP0, $ACC5, $ACC5
+ vpmuludq 32*3-128($aap), $B1, $TEMP1
+ vpaddq $TEMP1, $ACC6, $ACC6
+ vpmuludq 32*4-128($aap), $B1, $TEMP2
+ vpaddq $TEMP2, $ACC7, $ACC7
+ vpmuludq 32*5-128($aap), $B1, $TEMP0
+ vpaddq $TEMP0, $ACC8, $ACC8
+ vpmuludq 32*6-128($aap), $B1, $TEMP1
+ vpaddq $TEMP1, $ACC0, $ACC0
+ vpmuludq 32*7-128($aap), $B1, $ACC1
+ vpbroadcastq 32*4-128($tpa), $B1
+ vpaddq 32*10-448($tp1), $ACC1, $ACC1
+
+ vmovdqu $ACC4, 32*4-192($tp0)
+ vmovdqu $ACC5, 32*5-192($tp0)
+
+ vpmuludq 32*3-128($ap), $B2, $TEMP0
+ vpaddq $TEMP0, $ACC6, $ACC6
+ vpmuludq 32*3-128($aap), $B2, $TEMP1
+ vpaddq $TEMP1, $ACC7, $ACC7
+ vpmuludq 32*4-128($aap), $B2, $TEMP2
+ vpaddq $TEMP2, $ACC8, $ACC8
+ vpmuludq 32*5-128($aap), $B2, $TEMP0
+ vpaddq $TEMP0, $ACC0, $ACC0
+ vpmuludq 32*6-128($aap), $B2, $TEMP1
+ vpaddq $TEMP1, $ACC1, $ACC1
+ vpmuludq 32*7-128($aap), $B2, $ACC2
+ vpbroadcastq 32*5-128($tpa), $B2
+ vpaddq 32*11-448($tp1), $ACC2, $ACC2
+
+ vmovdqu $ACC6, 32*6-192($tp0)
+ vmovdqu $ACC7, 32*7-192($tp0)
+
+ vpmuludq 32*4-128($ap), $B1, $TEMP0
+ vpaddq $TEMP0, $ACC8, $ACC8
+ vpmuludq 32*4-128($aap), $B1, $TEMP1
+ vpaddq $TEMP1, $ACC0, $ACC0
+ vpmuludq 32*5-128($aap), $B1, $TEMP2
+ vpaddq $TEMP2, $ACC1, $ACC1
+ vpmuludq 32*6-128($aap), $B1, $TEMP0
+ vpaddq $TEMP0, $ACC2, $ACC2
+ vpmuludq 32*7-128($aap), $B1, $ACC3
+ vpbroadcastq 32*6-128($tpa), $B1
+ vpaddq 32*12-448($tp1), $ACC3, $ACC3
+
+ vmovdqu $ACC8, 32*8-192($tp0)
+ vmovdqu $ACC0, 32*9-192($tp0)
+ lea 8($tp0), $tp0
+
+ vpmuludq 32*5-128($ap), $B2, $TEMP2
+ vpaddq $TEMP2, $ACC1, $ACC1
+ vpmuludq 32*5-128($aap), $B2, $TEMP0
+ vpaddq $TEMP0, $ACC2, $ACC2
+ vpmuludq 32*6-128($aap), $B2, $TEMP1
+ vpaddq $TEMP1, $ACC3, $ACC3
+ vpmuludq 32*7-128($aap), $B2, $ACC4
+ vpbroadcastq 32*7-128($tpa), $B2
+ vpaddq 32*13-448($tp1), $ACC4, $ACC4
+
+ vmovdqu $ACC1, 32*10-448($tp1)
+ vmovdqu $ACC2, 32*11-448($tp1)
+
+ vpmuludq 32*6-128($ap), $B1, $TEMP0
+ vpaddq $TEMP0, $ACC3, $ACC3
+ vpmuludq 32*6-128($aap), $B1, $TEMP1
+ vpbroadcastq 32*8-128($tpa), $ACC0 # borrow $ACC0 for $B1
+ vpaddq $TEMP1, $ACC4, $ACC4
+ vpmuludq 32*7-128($aap), $B1, $ACC5
+ vpbroadcastq 32*0+8-128($tpa), $B1 # for next iteration
+ vpaddq 32*14-448($tp1), $ACC5, $ACC5
+
+ vmovdqu $ACC3, 32*12-448($tp1)
+ vmovdqu $ACC4, 32*13-448($tp1)
+ lea 8($tpa), $tpa
+
+ vpmuludq 32*7-128($ap), $B2, $TEMP0
+ vpaddq $TEMP0, $ACC5, $ACC5
+ vpmuludq 32*7-128($aap), $B2, $ACC6
+ vpaddq 32*15-448($tp1), $ACC6, $ACC6
+
+ vpmuludq 32*8-128($ap), $ACC0, $ACC7
+ vmovdqu $ACC5, 32*14-448($tp1)
+ vpaddq 32*16-448($tp1), $ACC7, $ACC7
+ vmovdqu $ACC6, 32*15-448($tp1)
+ vmovdqu $ACC7, 32*16-448($tp1)
+ lea 8($tp1), $tp1
+
+ dec $i
+ jnz .LOOP_SQR_1024
+___
+$ZERO = $ACC9;
+$TEMP0 = $B1;
+$TEMP2 = $B2;
+$TEMP3 = $Y1;
+$TEMP4 = $Y2;
+$code.=<<___;
+ # we need to fix indices 32-39 to avoid overflow
+ vmovdqu 32*8(%rsp), $ACC8 # 32*8-192($tp0),
+ vmovdqu 32*9(%rsp), $ACC1 # 32*9-192($tp0)
+ vmovdqu 32*10(%rsp), $ACC2 # 32*10-192($tp0)
+ lea 192(%rsp), $tp0 # 64+128=192
+
+ vpsrlq \$29, $ACC8, $TEMP1
+ vpand $AND_MASK, $ACC8, $ACC8
+ vpsrlq \$29, $ACC1, $TEMP2
+ vpand $AND_MASK, $ACC1, $ACC1
+
+ vpermq \$0x93, $TEMP1, $TEMP1
+ vpxor $ZERO, $ZERO, $ZERO
+ vpermq \$0x93, $TEMP2, $TEMP2
+
+ vpblendd \$3, $ZERO, $TEMP1, $TEMP0
+ vpblendd \$3, $TEMP1, $TEMP2, $TEMP1
+ vpaddq $TEMP0, $ACC8, $ACC8
+ vpblendd \$3, $TEMP2, $ZERO, $TEMP2
+ vpaddq $TEMP1, $ACC1, $ACC1
+ vpaddq $TEMP2, $ACC2, $ACC2
+ vmovdqu $ACC1, 32*9-192($tp0)
+ vmovdqu $ACC2, 32*10-192($tp0)
+
+ mov (%rsp), %rax
+ mov 8(%rsp), $r1
+ mov 16(%rsp), $r2
+ mov 24(%rsp), $r3
+ vmovdqu 32*1(%rsp), $ACC1
+ vmovdqu 32*2-192($tp0), $ACC2
+ vmovdqu 32*3-192($tp0), $ACC3
+ vmovdqu 32*4-192($tp0), $ACC4
+ vmovdqu 32*5-192($tp0), $ACC5
+ vmovdqu 32*6-192($tp0), $ACC6
+ vmovdqu 32*7-192($tp0), $ACC7
+
+ mov %rax, $r0
+ imull $n0, %eax
+ and \$0x1fffffff, %eax
+ vmovd %eax, $Y1
+
+ mov %rax, %rdx
+ imulq -128($np), %rax
+ vpbroadcastq $Y1, $Y1
+ add %rax, $r0
+ mov %rdx, %rax
+ imulq 8-128($np), %rax
+ shr \$29, $r0
+ add %rax, $r1
+ mov %rdx, %rax
+ imulq 16-128($np), %rax
+ add $r0, $r1
+ add %rax, $r2
+ imulq 24-128($np), %rdx
+ add %rdx, $r3
+
+ mov $r1, %rax
+ imull $n0, %eax
+ and \$0x1fffffff, %eax
+
+ mov \$9, $i
+ jmp .LOOP_REDUCE_1024
+
+.align 32
+.LOOP_REDUCE_1024:
+ vmovd %eax, $Y2
+ vpbroadcastq $Y2, $Y2
+
+ vpmuludq 32*1-128($np), $Y1, $TEMP0
+ mov %rax, %rdx
+ imulq -128($np), %rax
+ vpaddq $TEMP0, $ACC1, $ACC1
+ add %rax, $r1
+ vpmuludq 32*2-128($np), $Y1, $TEMP1
+ mov %rdx, %rax
+ imulq 8-128($np), %rax
+ vpaddq $TEMP1, $ACC2, $ACC2
+ vpmuludq 32*3-128($np), $Y1, $TEMP2
+ .byte 0x67
+ add %rax, $r2
+ .byte 0x67
+ mov %rdx, %rax
+ imulq 16-128($np), %rax
+ shr \$29, $r1
+ vpaddq $TEMP2, $ACC3, $ACC3
+ vpmuludq 32*4-128($np), $Y1, $TEMP0
+ add %rax, $r3
+ add $r1, $r2
+ vpaddq $TEMP0, $ACC4, $ACC4
+ vpmuludq 32*5-128($np), $Y1, $TEMP1
+ mov $r2, %rax
+ imull $n0, %eax
+ vpaddq $TEMP1, $ACC5, $ACC5
+ vpmuludq 32*6-128($np), $Y1, $TEMP2
+ and \$0x1fffffff, %eax
+ vpaddq $TEMP2, $ACC6, $ACC6
+ vpmuludq 32*7-128($np), $Y1, $TEMP0
+ vpaddq $TEMP0, $ACC7, $ACC7
+ vpmuludq 32*8-128($np), $Y1, $TEMP1
+ vmovd %eax, $Y1
+ #vmovdqu 32*1-8-128($np), $TEMP2 # moved below
+ vpaddq $TEMP1, $ACC8, $ACC8
+ #vmovdqu 32*2-8-128($np), $TEMP0 # moved below
+ vpbroadcastq $Y1, $Y1
+
+ vpmuludq 32*1-8-128($np), $Y2, $TEMP2 # see above
+ vmovdqu 32*3-8-128($np), $TEMP1
+ mov %rax, %rdx
+ imulq -128($np), %rax
+ vpaddq $TEMP2, $ACC1, $ACC1
+ vpmuludq 32*2-8-128($np), $Y2, $TEMP0 # see above
+ vmovdqu 32*4-8-128($np), $TEMP2
+ add %rax, $r2
+ mov %rdx, %rax
+ imulq 8-128($np), %rax
+ vpaddq $TEMP0, $ACC2, $ACC2
+ add $r3, %rax
+ shr \$29, $r2
+ vpmuludq $Y2, $TEMP1, $TEMP1
+ vmovdqu 32*5-8-128($np), $TEMP0
+ add $r2, %rax
+ vpaddq $TEMP1, $ACC3, $ACC3
+ vpmuludq $Y2, $TEMP2, $TEMP2
+ vmovdqu 32*6-8-128($np), $TEMP1
+ .byte 0x67
+ mov %rax, $r3
+ imull $n0, %eax
+ vpaddq $TEMP2, $ACC4, $ACC4
+ vpmuludq $Y2, $TEMP0, $TEMP0
+ .byte 0xc4,0x41,0x7e,0x6f,0x9d,0x58,0x00,0x00,0x00 # vmovdqu 32*7-8-128($np), $TEMP2
+ and \$0x1fffffff, %eax
+ vpaddq $TEMP0, $ACC5, $ACC5
+ vpmuludq $Y2, $TEMP1, $TEMP1
+ vmovdqu 32*8-8-128($np), $TEMP0
+ vpaddq $TEMP1, $ACC6, $ACC6
+ vpmuludq $Y2, $TEMP2, $TEMP2
+ vmovdqu 32*9-8-128($np), $ACC9
+ vmovd %eax, $ACC0 # borrow ACC0 for Y2
+ imulq -128($np), %rax
+ vpaddq $TEMP2, $ACC7, $ACC7
+ vpmuludq $Y2, $TEMP0, $TEMP0
+ vmovdqu 32*1-16-128($np), $TEMP1
+ vpbroadcastq $ACC0, $ACC0
+ vpaddq $TEMP0, $ACC8, $ACC8
+ vpmuludq $Y2, $ACC9, $ACC9
+ vmovdqu 32*2-16-128($np), $TEMP2
+ add %rax, $r3
+
+___
+($ACC0,$Y2)=($Y2,$ACC0);
+$code.=<<___;
+ vmovdqu 32*1-24-128($np), $ACC0
+ vpmuludq $Y1, $TEMP1, $TEMP1
+ vmovdqu 32*3-16-128($np), $TEMP0
+ vpaddq $TEMP1, $ACC1, $ACC1
+ vpmuludq $Y2, $ACC0, $ACC0
+ vpmuludq $Y1, $TEMP2, $TEMP2
+ .byte 0xc4,0x41,0x7e,0x6f,0xb5,0xf0,0xff,0xff,0xff # vmovdqu 32*4-16-128($np), $TEMP1
+ vpaddq $ACC1, $ACC0, $ACC0
+ vpaddq $TEMP2, $ACC2, $ACC2
+ vpmuludq $Y1, $TEMP0, $TEMP0
+ vmovdqu 32*5-16-128($np), $TEMP2
+ .byte 0x67
+ vmovq $ACC0, %rax
+ vmovdqu $ACC0, (%rsp) # transfer $r0-$r3
+ vpaddq $TEMP0, $ACC3, $ACC3
+ vpmuludq $Y1, $TEMP1, $TEMP1
+ vmovdqu 32*6-16-128($np), $TEMP0
+ vpaddq $TEMP1, $ACC4, $ACC4
+ vpmuludq $Y1, $TEMP2, $TEMP2
+ vmovdqu 32*7-16-128($np), $TEMP1
+ vpaddq $TEMP2, $ACC5, $ACC5
+ vpmuludq $Y1, $TEMP0, $TEMP0
+ vmovdqu 32*8-16-128($np), $TEMP2
+ vpaddq $TEMP0, $ACC6, $ACC6
+ vpmuludq $Y1, $TEMP1, $TEMP1
+ shr \$29, $r3
+ vmovdqu 32*9-16-128($np), $TEMP0
+ add $r3, %rax
+ vpaddq $TEMP1, $ACC7, $ACC7
+ vpmuludq $Y1, $TEMP2, $TEMP2
+ #vmovdqu 32*2-24-128($np), $TEMP1 # moved below
+ mov %rax, $r0
+ imull $n0, %eax
+ vpaddq $TEMP2, $ACC8, $ACC8
+ vpmuludq $Y1, $TEMP0, $TEMP0
+ and \$0x1fffffff, %eax
+ vmovd %eax, $Y1
+ vmovdqu 32*3-24-128($np), $TEMP2
+ .byte 0x67
+ vpaddq $TEMP0, $ACC9, $ACC9
+ vpbroadcastq $Y1, $Y1
+
+ vpmuludq 32*2-24-128($np), $Y2, $TEMP1 # see above
+ vmovdqu 32*4-24-128($np), $TEMP0
+ mov %rax, %rdx
+ imulq -128($np), %rax
+ mov 8(%rsp), $r1
+ vpaddq $TEMP1, $ACC2, $ACC1
+ vpmuludq $Y2, $TEMP2, $TEMP2
+ vmovdqu 32*5-24-128($np), $TEMP1
+ add %rax, $r0
+ mov %rdx, %rax
+ imulq 8-128($np), %rax
+ .byte 0x67
+ shr \$29, $r0
+ mov 16(%rsp), $r2
+ vpaddq $TEMP2, $ACC3, $ACC2
+ vpmuludq $Y2, $TEMP0, $TEMP0
+ vmovdqu 32*6-24-128($np), $TEMP2
+ add %rax, $r1
+ mov %rdx, %rax
+ imulq 16-128($np), %rax
+ vpaddq $TEMP0, $ACC4, $ACC3
+ vpmuludq $Y2, $TEMP1, $TEMP1
+ vmovdqu 32*7-24-128($np), $TEMP0
+ imulq 24-128($np), %rdx # future $r3
+ add %rax, $r2
+ lea ($r0,$r1), %rax
+ vpaddq $TEMP1, $ACC5, $ACC4
+ vpmuludq $Y2, $TEMP2, $TEMP2
+ vmovdqu 32*8-24-128($np), $TEMP1
+ mov %rax, $r1
+ imull $n0, %eax
+ vpmuludq $Y2, $TEMP0, $TEMP0
+ vpaddq $TEMP2, $ACC6, $ACC5
+ vmovdqu 32*9-24-128($np), $TEMP2
+ and \$0x1fffffff, %eax
+ vpaddq $TEMP0, $ACC7, $ACC6
+ vpmuludq $Y2, $TEMP1, $TEMP1
+ add 24(%rsp), %rdx
+ vpaddq $TEMP1, $ACC8, $ACC7
+ vpmuludq $Y2, $TEMP2, $TEMP2
+ vpaddq $TEMP2, $ACC9, $ACC8
+ vmovq $r3, $ACC9
+ mov %rdx, $r3
+
+ dec $i
+ jnz .LOOP_REDUCE_1024
+___
+($ACC0,$Y2)=($Y2,$ACC0);
+$code.=<<___;
+ lea 448(%rsp), $tp1 # size optimization
+ vpaddq $ACC9, $Y2, $ACC0
+ vpxor $ZERO, $ZERO, $ZERO
+
+ vpaddq 32*9-192($tp0), $ACC0, $ACC0
+ vpaddq 32*10-448($tp1), $ACC1, $ACC1
+ vpaddq 32*11-448($tp1), $ACC2, $ACC2
+ vpaddq 32*12-448($tp1), $ACC3, $ACC3
+ vpaddq 32*13-448($tp1), $ACC4, $ACC4
+ vpaddq 32*14-448($tp1), $ACC5, $ACC5
+ vpaddq 32*15-448($tp1), $ACC6, $ACC6
+ vpaddq 32*16-448($tp1), $ACC7, $ACC7
+ vpaddq 32*17-448($tp1), $ACC8, $ACC8
+
+ vpsrlq \$29, $ACC0, $TEMP1
+ vpand $AND_MASK, $ACC0, $ACC0
+ vpsrlq \$29, $ACC1, $TEMP2
+ vpand $AND_MASK, $ACC1, $ACC1
+ vpsrlq \$29, $ACC2, $TEMP3
+ vpermq \$0x93, $TEMP1, $TEMP1
+ vpand $AND_MASK, $ACC2, $ACC2
+ vpsrlq \$29, $ACC3, $TEMP4
+ vpermq \$0x93, $TEMP2, $TEMP2
+ vpand $AND_MASK, $ACC3, $ACC3
+ vpermq \$0x93, $TEMP3, $TEMP3
+
+ vpblendd \$3, $ZERO, $TEMP1, $TEMP0
+ vpermq \$0x93, $TEMP4, $TEMP4
+ vpblendd \$3, $TEMP1, $TEMP2, $TEMP1
+ vpaddq $TEMP0, $ACC0, $ACC0
+ vpblendd \$3, $TEMP2, $TEMP3, $TEMP2
+ vpaddq $TEMP1, $ACC1, $ACC1
+ vpblendd \$3, $TEMP3, $TEMP4, $TEMP3
+ vpaddq $TEMP2, $ACC2, $ACC2
+ vpblendd \$3, $TEMP4, $ZERO, $TEMP4
+ vpaddq $TEMP3, $ACC3, $ACC3
+ vpaddq $TEMP4, $ACC4, $ACC4
+
+ vpsrlq \$29, $ACC0, $TEMP1
+ vpand $AND_MASK, $ACC0, $ACC0
+ vpsrlq \$29, $ACC1, $TEMP2
+ vpand $AND_MASK, $ACC1, $ACC1
+ vpsrlq \$29, $ACC2, $TEMP3
+ vpermq \$0x93, $TEMP1, $TEMP1
+ vpand $AND_MASK, $ACC2, $ACC2
+ vpsrlq \$29, $ACC3, $TEMP4
+ vpermq \$0x93, $TEMP2, $TEMP2
+ vpand $AND_MASK, $ACC3, $ACC3
+ vpermq \$0x93, $TEMP3, $TEMP3
+
+ vpblendd \$3, $ZERO, $TEMP1, $TEMP0
+ vpermq \$0x93, $TEMP4, $TEMP4
+ vpblendd \$3, $TEMP1, $TEMP2, $TEMP1
+ vpaddq $TEMP0, $ACC0, $ACC0
+ vpblendd \$3, $TEMP2, $TEMP3, $TEMP2
+ vpaddq $TEMP1, $ACC1, $ACC1
+ vmovdqu $ACC0, 32*0-128($rp)
+ vpblendd \$3, $TEMP3, $TEMP4, $TEMP3
+ vpaddq $TEMP2, $ACC2, $ACC2
+ vmovdqu $ACC1, 32*1-128($rp)
+ vpblendd \$3, $TEMP4, $ZERO, $TEMP4
+ vpaddq $TEMP3, $ACC3, $ACC3
+ vmovdqu $ACC2, 32*2-128($rp)
+ vpaddq $TEMP4, $ACC4, $ACC4
+ vmovdqu $ACC3, 32*3-128($rp)
+___
+$TEMP5=$ACC0;
+$code.=<<___;
+ vpsrlq \$29, $ACC4, $TEMP1
+ vpand $AND_MASK, $ACC4, $ACC4
+ vpsrlq \$29, $ACC5, $TEMP2
+ vpand $AND_MASK, $ACC5, $ACC5
+ vpsrlq \$29, $ACC6, $TEMP3
+ vpermq \$0x93, $TEMP1, $TEMP1
+ vpand $AND_MASK, $ACC6, $ACC6
+ vpsrlq \$29, $ACC7, $TEMP4
+ vpermq \$0x93, $TEMP2, $TEMP2
+ vpand $AND_MASK, $ACC7, $ACC7
+ vpsrlq \$29, $ACC8, $TEMP5
+ vpermq \$0x93, $TEMP3, $TEMP3
+ vpand $AND_MASK, $ACC8, $ACC8
+ vpermq \$0x93, $TEMP4, $TEMP4
+
+ vpblendd \$3, $ZERO, $TEMP1, $TEMP0
+ vpermq \$0x93, $TEMP5, $TEMP5
+ vpblendd \$3, $TEMP1, $TEMP2, $TEMP1
+ vpaddq $TEMP0, $ACC4, $ACC4
+ vpblendd \$3, $TEMP2, $TEMP3, $TEMP2
+ vpaddq $TEMP1, $ACC5, $ACC5
+ vpblendd \$3, $TEMP3, $TEMP4, $TEMP3
+ vpaddq $TEMP2, $ACC6, $ACC6
+ vpblendd \$3, $TEMP4, $TEMP5, $TEMP4
+ vpaddq $TEMP3, $ACC7, $ACC7
+ vpaddq $TEMP4, $ACC8, $ACC8
+
+ vpsrlq \$29, $ACC4, $TEMP1
+ vpand $AND_MASK, $ACC4, $ACC4
+ vpsrlq \$29, $ACC5, $TEMP2
+ vpand $AND_MASK, $ACC5, $ACC5
+ vpsrlq \$29, $ACC6, $TEMP3
+ vpermq \$0x93, $TEMP1, $TEMP1
+ vpand $AND_MASK, $ACC6, $ACC6
+ vpsrlq \$29, $ACC7, $TEMP4
+ vpermq \$0x93, $TEMP2, $TEMP2
+ vpand $AND_MASK, $ACC7, $ACC7
+ vpsrlq \$29, $ACC8, $TEMP5
+ vpermq \$0x93, $TEMP3, $TEMP3
+ vpand $AND_MASK, $ACC8, $ACC8
+ vpermq \$0x93, $TEMP4, $TEMP4
+
+ vpblendd \$3, $ZERO, $TEMP1, $TEMP0
+ vpermq \$0x93, $TEMP5, $TEMP5
+ vpblendd \$3, $TEMP1, $TEMP2, $TEMP1
+ vpaddq $TEMP0, $ACC4, $ACC4
+ vpblendd \$3, $TEMP2, $TEMP3, $TEMP2
+ vpaddq $TEMP1, $ACC5, $ACC5
+ vmovdqu $ACC4, 32*4-128($rp)
+ vpblendd \$3, $TEMP3, $TEMP4, $TEMP3
+ vpaddq $TEMP2, $ACC6, $ACC6
+ vmovdqu $ACC5, 32*5-128($rp)
+ vpblendd \$3, $TEMP4, $TEMP5, $TEMP4
+ vpaddq $TEMP3, $ACC7, $ACC7
+ vmovdqu $ACC6, 32*6-128($rp)
+ vpaddq $TEMP4, $ACC8, $ACC8
+ vmovdqu $ACC7, 32*7-128($rp)
+ vmovdqu $ACC8, 32*8-128($rp)
+
+ mov $rp, $ap
+ dec $rep
+ jne .LOOP_GRANDE_SQR_1024
+
+ vzeroall
+ mov %rbp, %rax
+.cfi_def_cfa_register %rax
+___
+$code.=<<___ if ($win64);
+.Lsqr_1024_in_tail:
+ movaps -0xd8(%rax),%xmm6
+ movaps -0xc8(%rax),%xmm7
+ movaps -0xb8(%rax),%xmm8
+ movaps -0xa8(%rax),%xmm9
+ movaps -0x98(%rax),%xmm10
+ movaps -0x88(%rax),%xmm11
+ movaps -0x78(%rax),%xmm12
+ movaps -0x68(%rax),%xmm13
+ movaps -0x58(%rax),%xmm14
+ movaps -0x48(%rax),%xmm15
+___
+$code.=<<___;
+ mov -48(%rax),%r15
+.cfi_restore %r15
+ mov -40(%rax),%r14
+.cfi_restore %r14
+ mov -32(%rax),%r13
+.cfi_restore %r13
+ mov -24(%rax),%r12
+.cfi_restore %r12
+ mov -16(%rax),%rbp
+.cfi_restore %rbp
+ mov -8(%rax),%rbx
+.cfi_restore %rbx
+ lea (%rax),%rsp # restore %rsp
+.cfi_def_cfa_register %rsp
+.Lsqr_1024_epilogue:
+ ret
+.cfi_endproc
+.size rsaz_1024_sqr_avx2,.-rsaz_1024_sqr_avx2
+___
+}
+
+{ # void AMM_WW(
+my $rp="%rdi"; # BN_ULONG *rp,
+my $ap="%rsi"; # const BN_ULONG *ap,
+my $bp="%rdx"; # const BN_ULONG *bp,
+my $np="%rcx"; # const BN_ULONG *np,
+my $n0="%r8d"; # unsigned int n0);
+
+# The registers that hold the accumulated redundant result
+# The AMM works on 1024 bit operands, and redundant word size is 29
+# Therefore: ceil(1024/29)/4 = 9
+my $ACC0="%ymm0";
+my $ACC1="%ymm1";
+my $ACC2="%ymm2";
+my $ACC3="%ymm3";
+my $ACC4="%ymm4";
+my $ACC5="%ymm5";
+my $ACC6="%ymm6";
+my $ACC7="%ymm7";
+my $ACC8="%ymm8";
+my $ACC9="%ymm9";
+
+# Registers that hold the broadcasted words of multiplier, currently used
+my $Bi="%ymm10";
+my $Yi="%ymm11";
+
+# Helper registers
+my $TEMP0=$ACC0;
+my $TEMP1="%ymm12";
+my $TEMP2="%ymm13";
+my $ZERO="%ymm14";
+my $AND_MASK="%ymm15";
+
+# alu registers that hold the first words of the ACC
+my $r0="%r9";
+my $r1="%r10";
+my $r2="%r11";
+my $r3="%r12";
+
+my $i="%r14d";
+my $tmp="%r15";
+
+$bp="%r13"; # reassigned argument
+
+$code.=<<___;
+.globl rsaz_1024_mul_avx2
+.type rsaz_1024_mul_avx2,\@function,5
+.align 64
+rsaz_1024_mul_avx2:
+.cfi_startproc
+ lea (%rsp), %rax
+.cfi_def_cfa_register %rax
+ push %rbx
+.cfi_push %rbx
+ push %rbp
+.cfi_push %rbp
+ push %r12
+.cfi_push %r12
+ push %r13
+.cfi_push %r13
+ push %r14
+.cfi_push %r14
+ push %r15
+.cfi_push %r15
+___
+$code.=<<___ if ($win64);
+ vzeroupper
+ lea -0xa8(%rsp),%rsp
+ vmovaps %xmm6,-0xd8(%rax)
+ vmovaps %xmm7,-0xc8(%rax)
+ vmovaps %xmm8,-0xb8(%rax)
+ vmovaps %xmm9,-0xa8(%rax)
+ vmovaps %xmm10,-0x98(%rax)
+ vmovaps %xmm11,-0x88(%rax)
+ vmovaps %xmm12,-0x78(%rax)
+ vmovaps %xmm13,-0x68(%rax)
+ vmovaps %xmm14,-0x58(%rax)
+ vmovaps %xmm15,-0x48(%rax)
+.Lmul_1024_body:
+___
+$code.=<<___;
+ mov %rax,%rbp
+.cfi_def_cfa_register %rbp
+ vzeroall
+ mov %rdx, $bp # reassigned argument
+ sub \$64,%rsp
+
+ # unaligned 256-bit load that crosses page boundary can
+ # cause severe performance degradation here, so if $ap does
+ # cross page boundary, swap it with $bp [meaning that caller
+ # is advised to lay down $ap and $bp next to each other, so
+ # that only one can cross page boundary].
+ .byte 0x67,0x67
+ mov $ap, $tmp
+ and \$4095, $tmp
+ add \$32*10, $tmp
+ shr \$12, $tmp
+ mov $ap, $tmp
+ cmovnz $bp, $ap
+ cmovnz $tmp, $bp
+
+ mov $np, $tmp
+ sub \$-128,$ap # size optimization
+ sub \$-128,$np
+ sub \$-128,$rp
+
+ and \$4095, $tmp # see if $np crosses page
+ add \$32*10, $tmp
+ .byte 0x67,0x67
+ shr \$12, $tmp
+ jz .Lmul_1024_no_n_copy
+
+ # unaligned 256-bit load that crosses page boundary can
+ # cause severe performance degradation here, so if $np does
+ # cross page boundary, copy it to stack and make sure stack
+ # frame doesn't...
+ sub \$32*10,%rsp
+ vmovdqu 32*0-128($np), $ACC0
+ and \$-512, %rsp
+ vmovdqu 32*1-128($np), $ACC1
+ vmovdqu 32*2-128($np), $ACC2
+ vmovdqu 32*3-128($np), $ACC3
+ vmovdqu 32*4-128($np), $ACC4
+ vmovdqu 32*5-128($np), $ACC5
+ vmovdqu 32*6-128($np), $ACC6
+ vmovdqu 32*7-128($np), $ACC7
+ vmovdqu 32*8-128($np), $ACC8
+ lea 64+128(%rsp),$np
+ vmovdqu $ACC0, 32*0-128($np)
+ vpxor $ACC0, $ACC0, $ACC0
+ vmovdqu $ACC1, 32*1-128($np)
+ vpxor $ACC1, $ACC1, $ACC1
+ vmovdqu $ACC2, 32*2-128($np)
+ vpxor $ACC2, $ACC2, $ACC2
+ vmovdqu $ACC3, 32*3-128($np)
+ vpxor $ACC3, $ACC3, $ACC3
+ vmovdqu $ACC4, 32*4-128($np)
+ vpxor $ACC4, $ACC4, $ACC4
+ vmovdqu $ACC5, 32*5-128($np)
+ vpxor $ACC5, $ACC5, $ACC5
+ vmovdqu $ACC6, 32*6-128($np)
+ vpxor $ACC6, $ACC6, $ACC6
+ vmovdqu $ACC7, 32*7-128($np)
+ vpxor $ACC7, $ACC7, $ACC7
+ vmovdqu $ACC8, 32*8-128($np)
+ vmovdqa $ACC0, $ACC8
+ vmovdqu $ACC9, 32*9-128($np) # $ACC9 is zero after vzeroall
+.Lmul_1024_no_n_copy:
+ and \$-64,%rsp
+
+ mov ($bp), %rbx
+ vpbroadcastq ($bp), $Bi
+ vmovdqu $ACC0, (%rsp) # clear top of stack
+ xor $r0, $r0
+ .byte 0x67
+ xor $r1, $r1
+ xor $r2, $r2
+ xor $r3, $r3
+
+ vmovdqu .Land_mask(%rip), $AND_MASK
+ mov \$9, $i
+ vmovdqu $ACC9, 32*9-128($rp) # $ACC9 is zero after vzeroall
+ jmp .Loop_mul_1024
+
+.align 32
+.Loop_mul_1024:
+ vpsrlq \$29, $ACC3, $ACC9 # correct $ACC3(*)
+ mov %rbx, %rax
+ imulq -128($ap), %rax
+ add $r0, %rax
+ mov %rbx, $r1
+ imulq 8-128($ap), $r1
+ add 8(%rsp), $r1
+
+ mov %rax, $r0
+ imull $n0, %eax
+ and \$0x1fffffff, %eax
+
+ mov %rbx, $r2
+ imulq 16-128($ap), $r2
+ add 16(%rsp), $r2
+
+ mov %rbx, $r3
+ imulq 24-128($ap), $r3
+ add 24(%rsp), $r3
+ vpmuludq 32*1-128($ap),$Bi,$TEMP0
+ vmovd %eax, $Yi
+ vpaddq $TEMP0,$ACC1,$ACC1
+ vpmuludq 32*2-128($ap),$Bi,$TEMP1
+ vpbroadcastq $Yi, $Yi
+ vpaddq $TEMP1,$ACC2,$ACC2
+ vpmuludq 32*3-128($ap),$Bi,$TEMP2
+ vpand $AND_MASK, $ACC3, $ACC3 # correct $ACC3
+ vpaddq $TEMP2,$ACC3,$ACC3
+ vpmuludq 32*4-128($ap),$Bi,$TEMP0
+ vpaddq $TEMP0,$ACC4,$ACC4
+ vpmuludq 32*5-128($ap),$Bi,$TEMP1
+ vpaddq $TEMP1,$ACC5,$ACC5
+ vpmuludq 32*6-128($ap),$Bi,$TEMP2
+ vpaddq $TEMP2,$ACC6,$ACC6
+ vpmuludq 32*7-128($ap),$Bi,$TEMP0
+ vpermq \$0x93, $ACC9, $ACC9 # correct $ACC3
+ vpaddq $TEMP0,$ACC7,$ACC7
+ vpmuludq 32*8-128($ap),$Bi,$TEMP1
+ vpbroadcastq 8($bp), $Bi
+ vpaddq $TEMP1,$ACC8,$ACC8
+
+ mov %rax,%rdx
+ imulq -128($np),%rax
+ add %rax,$r0
+ mov %rdx,%rax
+ imulq 8-128($np),%rax
+ add %rax,$r1
+ mov %rdx,%rax
+ imulq 16-128($np),%rax
+ add %rax,$r2
+ shr \$29, $r0
+ imulq 24-128($np),%rdx
+ add %rdx,$r3
+ add $r0, $r1
+
+ vpmuludq 32*1-128($np),$Yi,$TEMP2
+ vmovq $Bi, %rbx
+ vpaddq $TEMP2,$ACC1,$ACC1
+ vpmuludq 32*2-128($np),$Yi,$TEMP0
+ vpaddq $TEMP0,$ACC2,$ACC2
+ vpmuludq 32*3-128($np),$Yi,$TEMP1
+ vpaddq $TEMP1,$ACC3,$ACC3
+ vpmuludq 32*4-128($np),$Yi,$TEMP2
+ vpaddq $TEMP2,$ACC4,$ACC4
+ vpmuludq 32*5-128($np),$Yi,$TEMP0
+ vpaddq $TEMP0,$ACC5,$ACC5
+ vpmuludq 32*6-128($np),$Yi,$TEMP1
+ vpaddq $TEMP1,$ACC6,$ACC6
+ vpmuludq 32*7-128($np),$Yi,$TEMP2
+ vpblendd \$3, $ZERO, $ACC9, $TEMP1 # correct $ACC3
+ vpaddq $TEMP2,$ACC7,$ACC7
+ vpmuludq 32*8-128($np),$Yi,$TEMP0
+ vpaddq $TEMP1, $ACC3, $ACC3 # correct $ACC3
+ vpaddq $TEMP0,$ACC8,$ACC8
+
+ mov %rbx, %rax
+ imulq -128($ap),%rax
+ add %rax,$r1
+ vmovdqu -8+32*1-128($ap),$TEMP1
+ mov %rbx, %rax
+ imulq 8-128($ap),%rax
+ add %rax,$r2
+ vmovdqu -8+32*2-128($ap),$TEMP2
+
+ mov $r1, %rax
+ vpblendd \$0xfc, $ZERO, $ACC9, $ACC9 # correct $ACC3
+ imull $n0, %eax
+ vpaddq $ACC9,$ACC4,$ACC4 # correct $ACC3
+ and \$0x1fffffff, %eax
+
+ imulq 16-128($ap),%rbx
+ add %rbx,$r3
+ vpmuludq $Bi,$TEMP1,$TEMP1
+ vmovd %eax, $Yi
+ vmovdqu -8+32*3-128($ap),$TEMP0
+ vpaddq $TEMP1,$ACC1,$ACC1
+ vpmuludq $Bi,$TEMP2,$TEMP2
+ vpbroadcastq $Yi, $Yi
+ vmovdqu -8+32*4-128($ap),$TEMP1
+ vpaddq $TEMP2,$ACC2,$ACC2
+ vpmuludq $Bi,$TEMP0,$TEMP0
+ vmovdqu -8+32*5-128($ap),$TEMP2
+ vpaddq $TEMP0,$ACC3,$ACC3
+ vpmuludq $Bi,$TEMP1,$TEMP1
+ vmovdqu -8+32*6-128($ap),$TEMP0
+ vpaddq $TEMP1,$ACC4,$ACC4
+ vpmuludq $Bi,$TEMP2,$TEMP2
+ vmovdqu -8+32*7-128($ap),$TEMP1
+ vpaddq $TEMP2,$ACC5,$ACC5
+ vpmuludq $Bi,$TEMP0,$TEMP0
+ vmovdqu -8+32*8-128($ap),$TEMP2
+ vpaddq $TEMP0,$ACC6,$ACC6
+ vpmuludq $Bi,$TEMP1,$TEMP1
+ vmovdqu -8+32*9-128($ap),$ACC9
+ vpaddq $TEMP1,$ACC7,$ACC7
+ vpmuludq $Bi,$TEMP2,$TEMP2
+ vpaddq $TEMP2,$ACC8,$ACC8
+ vpmuludq $Bi,$ACC9,$ACC9
+ vpbroadcastq 16($bp), $Bi
+
+ mov %rax,%rdx
+ imulq -128($np),%rax
+ add %rax,$r1
+ vmovdqu -8+32*1-128($np),$TEMP0
+ mov %rdx,%rax
+ imulq 8-128($np),%rax
+ add %rax,$r2
+ vmovdqu -8+32*2-128($np),$TEMP1
+ shr \$29, $r1
+ imulq 16-128($np),%rdx
+ add %rdx,$r3
+ add $r1, $r2
+
+ vpmuludq $Yi,$TEMP0,$TEMP0
+ vmovq $Bi, %rbx
+ vmovdqu -8+32*3-128($np),$TEMP2
+ vpaddq $TEMP0,$ACC1,$ACC1
+ vpmuludq $Yi,$TEMP1,$TEMP1
+ vmovdqu -8+32*4-128($np),$TEMP0
+ vpaddq $TEMP1,$ACC2,$ACC2
+ vpmuludq $Yi,$TEMP2,$TEMP2
+ vmovdqu -8+32*5-128($np),$TEMP1
+ vpaddq $TEMP2,$ACC3,$ACC3
+ vpmuludq $Yi,$TEMP0,$TEMP0
+ vmovdqu -8+32*6-128($np),$TEMP2
+ vpaddq $TEMP0,$ACC4,$ACC4
+ vpmuludq $Yi,$TEMP1,$TEMP1
+ vmovdqu -8+32*7-128($np),$TEMP0
+ vpaddq $TEMP1,$ACC5,$ACC5
+ vpmuludq $Yi,$TEMP2,$TEMP2
+ vmovdqu -8+32*8-128($np),$TEMP1
+ vpaddq $TEMP2,$ACC6,$ACC6
+ vpmuludq $Yi,$TEMP0,$TEMP0
+ vmovdqu -8+32*9-128($np),$TEMP2
+ vpaddq $TEMP0,$ACC7,$ACC7
+ vpmuludq $Yi,$TEMP1,$TEMP1
+ vpaddq $TEMP1,$ACC8,$ACC8
+ vpmuludq $Yi,$TEMP2,$TEMP2
+ vpaddq $TEMP2,$ACC9,$ACC9
+
+ vmovdqu -16+32*1-128($ap),$TEMP0
+ mov %rbx,%rax
+ imulq -128($ap),%rax
+ add $r2,%rax
+
+ vmovdqu -16+32*2-128($ap),$TEMP1
+ mov %rax,$r2
+ imull $n0, %eax
+ and \$0x1fffffff, %eax
+
+ imulq 8-128($ap),%rbx
+ add %rbx,$r3
+ vpmuludq $Bi,$TEMP0,$TEMP0
+ vmovd %eax, $Yi
+ vmovdqu -16+32*3-128($ap),$TEMP2
+ vpaddq $TEMP0,$ACC1,$ACC1
+ vpmuludq $Bi,$TEMP1,$TEMP1
+ vpbroadcastq $Yi, $Yi
+ vmovdqu -16+32*4-128($ap),$TEMP0
+ vpaddq $TEMP1,$ACC2,$ACC2
+ vpmuludq $Bi,$TEMP2,$TEMP2
+ vmovdqu -16+32*5-128($ap),$TEMP1
+ vpaddq $TEMP2,$ACC3,$ACC3
+ vpmuludq $Bi,$TEMP0,$TEMP0
+ vmovdqu -16+32*6-128($ap),$TEMP2
+ vpaddq $TEMP0,$ACC4,$ACC4
+ vpmuludq $Bi,$TEMP1,$TEMP1
+ vmovdqu -16+32*7-128($ap),$TEMP0
+ vpaddq $TEMP1,$ACC5,$ACC5
+ vpmuludq $Bi,$TEMP2,$TEMP2
+ vmovdqu -16+32*8-128($ap),$TEMP1
+ vpaddq $TEMP2,$ACC6,$ACC6
+ vpmuludq $Bi,$TEMP0,$TEMP0
+ vmovdqu -16+32*9-128($ap),$TEMP2
+ vpaddq $TEMP0,$ACC7,$ACC7
+ vpmuludq $Bi,$TEMP1,$TEMP1
+ vpaddq $TEMP1,$ACC8,$ACC8
+ vpmuludq $Bi,$TEMP2,$TEMP2
+ vpbroadcastq 24($bp), $Bi
+ vpaddq $TEMP2,$ACC9,$ACC9
+
+ vmovdqu -16+32*1-128($np),$TEMP0
+ mov %rax,%rdx
+ imulq -128($np),%rax
+ add %rax,$r2
+ vmovdqu -16+32*2-128($np),$TEMP1
+ imulq 8-128($np),%rdx
+ add %rdx,$r3
+ shr \$29, $r2
+
+ vpmuludq $Yi,$TEMP0,$TEMP0
+ vmovq $Bi, %rbx
+ vmovdqu -16+32*3-128($np),$TEMP2
+ vpaddq $TEMP0,$ACC1,$ACC1
+ vpmuludq $Yi,$TEMP1,$TEMP1
+ vmovdqu -16+32*4-128($np),$TEMP0
+ vpaddq $TEMP1,$ACC2,$ACC2
+ vpmuludq $Yi,$TEMP2,$TEMP2
+ vmovdqu -16+32*5-128($np),$TEMP1
+ vpaddq $TEMP2,$ACC3,$ACC3
+ vpmuludq $Yi,$TEMP0,$TEMP0
+ vmovdqu -16+32*6-128($np),$TEMP2
+ vpaddq $TEMP0,$ACC4,$ACC4
+ vpmuludq $Yi,$TEMP1,$TEMP1
+ vmovdqu -16+32*7-128($np),$TEMP0
+ vpaddq $TEMP1,$ACC5,$ACC5
+ vpmuludq $Yi,$TEMP2,$TEMP2
+ vmovdqu -16+32*8-128($np),$TEMP1
+ vpaddq $TEMP2,$ACC6,$ACC6
+ vpmuludq $Yi,$TEMP0,$TEMP0
+ vmovdqu -16+32*9-128($np),$TEMP2
+ vpaddq $TEMP0,$ACC7,$ACC7
+ vpmuludq $Yi,$TEMP1,$TEMP1
+ vmovdqu -24+32*1-128($ap),$TEMP0
+ vpaddq $TEMP1,$ACC8,$ACC8
+ vpmuludq $Yi,$TEMP2,$TEMP2
+ vmovdqu -24+32*2-128($ap),$TEMP1
+ vpaddq $TEMP2,$ACC9,$ACC9
+
+ add $r2, $r3
+ imulq -128($ap),%rbx
+ add %rbx,$r3
+
+ mov $r3, %rax
+ imull $n0, %eax
+ and \$0x1fffffff, %eax
+
+ vpmuludq $Bi,$TEMP0,$TEMP0
+ vmovd %eax, $Yi
+ vmovdqu -24+32*3-128($ap),$TEMP2
+ vpaddq $TEMP0,$ACC1,$ACC1
+ vpmuludq $Bi,$TEMP1,$TEMP1
+ vpbroadcastq $Yi, $Yi
+ vmovdqu -24+32*4-128($ap),$TEMP0
+ vpaddq $TEMP1,$ACC2,$ACC2
+ vpmuludq $Bi,$TEMP2,$TEMP2
+ vmovdqu -24+32*5-128($ap),$TEMP1
+ vpaddq $TEMP2,$ACC3,$ACC3
+ vpmuludq $Bi,$TEMP0,$TEMP0
+ vmovdqu -24+32*6-128($ap),$TEMP2
+ vpaddq $TEMP0,$ACC4,$ACC4
+ vpmuludq $Bi,$TEMP1,$TEMP1
+ vmovdqu -24+32*7-128($ap),$TEMP0
+ vpaddq $TEMP1,$ACC5,$ACC5
+ vpmuludq $Bi,$TEMP2,$TEMP2
+ vmovdqu -24+32*8-128($ap),$TEMP1
+ vpaddq $TEMP2,$ACC6,$ACC6
+ vpmuludq $Bi,$TEMP0,$TEMP0
+ vmovdqu -24+32*9-128($ap),$TEMP2
+ vpaddq $TEMP0,$ACC7,$ACC7
+ vpmuludq $Bi,$TEMP1,$TEMP1
+ vpaddq $TEMP1,$ACC8,$ACC8
+ vpmuludq $Bi,$TEMP2,$TEMP2
+ vpbroadcastq 32($bp), $Bi
+ vpaddq $TEMP2,$ACC9,$ACC9
+ add \$32, $bp # $bp++
+
+ vmovdqu -24+32*1-128($np),$TEMP0
+ imulq -128($np),%rax
+ add %rax,$r3
+ shr \$29, $r3
+
+ vmovdqu -24+32*2-128($np),$TEMP1
+ vpmuludq $Yi,$TEMP0,$TEMP0
+ vmovq $Bi, %rbx
+ vmovdqu -24+32*3-128($np),$TEMP2
+ vpaddq $TEMP0,$ACC1,$ACC0 # $ACC0==$TEMP0
+ vpmuludq $Yi,$TEMP1,$TEMP1
+ vmovdqu $ACC0, (%rsp) # transfer $r0-$r3
+ vpaddq $TEMP1,$ACC2,$ACC1
+ vmovdqu -24+32*4-128($np),$TEMP0
+ vpmuludq $Yi,$TEMP2,$TEMP2
+ vmovdqu -24+32*5-128($np),$TEMP1
+ vpaddq $TEMP2,$ACC3,$ACC2
+ vpmuludq $Yi,$TEMP0,$TEMP0
+ vmovdqu -24+32*6-128($np),$TEMP2
+ vpaddq $TEMP0,$ACC4,$ACC3
+ vpmuludq $Yi,$TEMP1,$TEMP1
+ vmovdqu -24+32*7-128($np),$TEMP0
+ vpaddq $TEMP1,$ACC5,$ACC4
+ vpmuludq $Yi,$TEMP2,$TEMP2
+ vmovdqu -24+32*8-128($np),$TEMP1
+ vpaddq $TEMP2,$ACC6,$ACC5
+ vpmuludq $Yi,$TEMP0,$TEMP0
+ vmovdqu -24+32*9-128($np),$TEMP2
+ mov $r3, $r0
+ vpaddq $TEMP0,$ACC7,$ACC6
+ vpmuludq $Yi,$TEMP1,$TEMP1
+ add (%rsp), $r0
+ vpaddq $TEMP1,$ACC8,$ACC7
+ vpmuludq $Yi,$TEMP2,$TEMP2
+ vmovq $r3, $TEMP1
+ vpaddq $TEMP2,$ACC9,$ACC8
+
+ dec $i
+ jnz .Loop_mul_1024
+___
+
+# (*) Original implementation was correcting ACC1-ACC3 for overflow
+# after 7 loop runs, or after 28 iterations, or 56 additions.
+# But as we underutilize resources, it's possible to correct in
+# each iteration with marginal performance loss. But then, as
+# we do it in each iteration, we can correct less digits, and
+# avoid performance penalties completely.
+
+$TEMP0 = $ACC9;
+$TEMP3 = $Bi;
+$TEMP4 = $Yi;
+$code.=<<___;
+ vpaddq (%rsp), $TEMP1, $ACC0
+
+ vpsrlq \$29, $ACC0, $TEMP1
+ vpand $AND_MASK, $ACC0, $ACC0
+ vpsrlq \$29, $ACC1, $TEMP2
+ vpand $AND_MASK, $ACC1, $ACC1
+ vpsrlq \$29, $ACC2, $TEMP3
+ vpermq \$0x93, $TEMP1, $TEMP1
+ vpand $AND_MASK, $ACC2, $ACC2
+ vpsrlq \$29, $ACC3, $TEMP4
+ vpermq \$0x93, $TEMP2, $TEMP2
+ vpand $AND_MASK, $ACC3, $ACC3
+
+ vpblendd \$3, $ZERO, $TEMP1, $TEMP0
+ vpermq \$0x93, $TEMP3, $TEMP3
+ vpblendd \$3, $TEMP1, $TEMP2, $TEMP1
+ vpermq \$0x93, $TEMP4, $TEMP4
+ vpaddq $TEMP0, $ACC0, $ACC0
+ vpblendd \$3, $TEMP2, $TEMP3, $TEMP2
+ vpaddq $TEMP1, $ACC1, $ACC1
+ vpblendd \$3, $TEMP3, $TEMP4, $TEMP3
+ vpaddq $TEMP2, $ACC2, $ACC2
+ vpblendd \$3, $TEMP4, $ZERO, $TEMP4
+ vpaddq $TEMP3, $ACC3, $ACC3
+ vpaddq $TEMP4, $ACC4, $ACC4
+
+ vpsrlq \$29, $ACC0, $TEMP1
+ vpand $AND_MASK, $ACC0, $ACC0
+ vpsrlq \$29, $ACC1, $TEMP2
+ vpand $AND_MASK, $ACC1, $ACC1
+ vpsrlq \$29, $ACC2, $TEMP3
+ vpermq \$0x93, $TEMP1, $TEMP1
+ vpand $AND_MASK, $ACC2, $ACC2
+ vpsrlq \$29, $ACC3, $TEMP4
+ vpermq \$0x93, $TEMP2, $TEMP2
+ vpand $AND_MASK, $ACC3, $ACC3
+ vpermq \$0x93, $TEMP3, $TEMP3
+
+ vpblendd \$3, $ZERO, $TEMP1, $TEMP0
+ vpermq \$0x93, $TEMP4, $TEMP4
+ vpblendd \$3, $TEMP1, $TEMP2, $TEMP1
+ vpaddq $TEMP0, $ACC0, $ACC0
+ vpblendd \$3, $TEMP2, $TEMP3, $TEMP2
+ vpaddq $TEMP1, $ACC1, $ACC1
+ vpblendd \$3, $TEMP3, $TEMP4, $TEMP3
+ vpaddq $TEMP2, $ACC2, $ACC2
+ vpblendd \$3, $TEMP4, $ZERO, $TEMP4
+ vpaddq $TEMP3, $ACC3, $ACC3
+ vpaddq $TEMP4, $ACC4, $ACC4
+
+ vmovdqu $ACC0, 0-128($rp)
+ vmovdqu $ACC1, 32-128($rp)
+ vmovdqu $ACC2, 64-128($rp)
+ vmovdqu $ACC3, 96-128($rp)
+___
+
+$TEMP5=$ACC0;
+$code.=<<___;
+ vpsrlq \$29, $ACC4, $TEMP1
+ vpand $AND_MASK, $ACC4, $ACC4
+ vpsrlq \$29, $ACC5, $TEMP2
+ vpand $AND_MASK, $ACC5, $ACC5
+ vpsrlq \$29, $ACC6, $TEMP3
+ vpermq \$0x93, $TEMP1, $TEMP1
+ vpand $AND_MASK, $ACC6, $ACC6
+ vpsrlq \$29, $ACC7, $TEMP4
+ vpermq \$0x93, $TEMP2, $TEMP2
+ vpand $AND_MASK, $ACC7, $ACC7
+ vpsrlq \$29, $ACC8, $TEMP5
+ vpermq \$0x93, $TEMP3, $TEMP3
+ vpand $AND_MASK, $ACC8, $ACC8
+ vpermq \$0x93, $TEMP4, $TEMP4
+
+ vpblendd \$3, $ZERO, $TEMP1, $TEMP0
+ vpermq \$0x93, $TEMP5, $TEMP5
+ vpblendd \$3, $TEMP1, $TEMP2, $TEMP1
+ vpaddq $TEMP0, $ACC4, $ACC4
+ vpblendd \$3, $TEMP2, $TEMP3, $TEMP2
+ vpaddq $TEMP1, $ACC5, $ACC5
+ vpblendd \$3, $TEMP3, $TEMP4, $TEMP3
+ vpaddq $TEMP2, $ACC6, $ACC6
+ vpblendd \$3, $TEMP4, $TEMP5, $TEMP4
+ vpaddq $TEMP3, $ACC7, $ACC7
+ vpaddq $TEMP4, $ACC8, $ACC8
+
+ vpsrlq \$29, $ACC4, $TEMP1
+ vpand $AND_MASK, $ACC4, $ACC4
+ vpsrlq \$29, $ACC5, $TEMP2
+ vpand $AND_MASK, $ACC5, $ACC5
+ vpsrlq \$29, $ACC6, $TEMP3
+ vpermq \$0x93, $TEMP1, $TEMP1
+ vpand $AND_MASK, $ACC6, $ACC6
+ vpsrlq \$29, $ACC7, $TEMP4
+ vpermq \$0x93, $TEMP2, $TEMP2
+ vpand $AND_MASK, $ACC7, $ACC7
+ vpsrlq \$29, $ACC8, $TEMP5
+ vpermq \$0x93, $TEMP3, $TEMP3
+ vpand $AND_MASK, $ACC8, $ACC8
+ vpermq \$0x93, $TEMP4, $TEMP4
+
+ vpblendd \$3, $ZERO, $TEMP1, $TEMP0
+ vpermq \$0x93, $TEMP5, $TEMP5
+ vpblendd \$3, $TEMP1, $TEMP2, $TEMP1
+ vpaddq $TEMP0, $ACC4, $ACC4
+ vpblendd \$3, $TEMP2, $TEMP3, $TEMP2
+ vpaddq $TEMP1, $ACC5, $ACC5
+ vpblendd \$3, $TEMP3, $TEMP4, $TEMP3
+ vpaddq $TEMP2, $ACC6, $ACC6
+ vpblendd \$3, $TEMP4, $TEMP5, $TEMP4
+ vpaddq $TEMP3, $ACC7, $ACC7
+ vpaddq $TEMP4, $ACC8, $ACC8
+
+ vmovdqu $ACC4, 128-128($rp)
+ vmovdqu $ACC5, 160-128($rp)
+ vmovdqu $ACC6, 192-128($rp)
+ vmovdqu $ACC7, 224-128($rp)
+ vmovdqu $ACC8, 256-128($rp)
+ vzeroupper
+
+ mov %rbp, %rax
+.cfi_def_cfa_register %rax
+___
+$code.=<<___ if ($win64);
+.Lmul_1024_in_tail:
+ movaps -0xd8(%rax),%xmm6
+ movaps -0xc8(%rax),%xmm7
+ movaps -0xb8(%rax),%xmm8
+ movaps -0xa8(%rax),%xmm9
+ movaps -0x98(%rax),%xmm10
+ movaps -0x88(%rax),%xmm11
+ movaps -0x78(%rax),%xmm12
+ movaps -0x68(%rax),%xmm13
+ movaps -0x58(%rax),%xmm14
+ movaps -0x48(%rax),%xmm15
+___
+$code.=<<___;
+ mov -48(%rax),%r15
+.cfi_restore %r15
+ mov -40(%rax),%r14
+.cfi_restore %r14
+ mov -32(%rax),%r13
+.cfi_restore %r13
+ mov -24(%rax),%r12
+.cfi_restore %r12
+ mov -16(%rax),%rbp
+.cfi_restore %rbp
+ mov -8(%rax),%rbx
+.cfi_restore %rbx
+ lea (%rax),%rsp # restore %rsp
+.cfi_def_cfa_register %rsp
+.Lmul_1024_epilogue:
+ ret
+.cfi_endproc
+.size rsaz_1024_mul_avx2,.-rsaz_1024_mul_avx2
+___
+}
+{
+my ($out,$inp) = $win64 ? ("%rcx","%rdx") : ("%rdi","%rsi");
+my @T = map("%r$_",(8..11));
+
+$code.=<<___;
+.globl rsaz_1024_red2norm_avx2
+.type rsaz_1024_red2norm_avx2,\@abi-omnipotent
+.align 32
+rsaz_1024_red2norm_avx2:
+.cfi_startproc
+ sub \$-128,$inp # size optimization
+ xor %rax,%rax
+___
+
+for ($j=0,$i=0; $i<16; $i++) {
+ my $k=0;
+ while (29*$j<64*($i+1)) { # load data till boundary
+ $code.=" mov `8*$j-128`($inp), @T[0]\n";
+ $j++; $k++; push(@T,shift(@T));
+ }
+ $l=$k;
+ while ($k>1) { # shift loaded data but last value
+ $code.=" shl \$`29*($j-$k)`,@T[-$k]\n";
+ $k--;
+ }
+ $code.=<<___; # shift last value
+ mov @T[-1], @T[0]
+ shl \$`29*($j-1)`, @T[-1]
+ shr \$`-29*($j-1)`, @T[0]
+___
+ while ($l) { # accumulate all values
+ $code.=" add @T[-$l], %rax\n";
+ $l--;
+ }
+ $code.=<<___;
+ adc \$0, @T[0] # consume eventual carry
+ mov %rax, 8*$i($out)
+ mov @T[0], %rax
+___
+ push(@T,shift(@T));
+}
+$code.=<<___;
+ ret
+.cfi_endproc
+.size rsaz_1024_red2norm_avx2,.-rsaz_1024_red2norm_avx2
+
+.globl rsaz_1024_norm2red_avx2
+.type rsaz_1024_norm2red_avx2,\@abi-omnipotent
+.align 32
+rsaz_1024_norm2red_avx2:
+.cfi_startproc
+ sub \$-128,$out # size optimization
+ mov ($inp),@T[0]
+ mov \$0x1fffffff,%eax
+___
+for ($j=0,$i=0; $i<16; $i++) {
+ $code.=" mov `8*($i+1)`($inp),@T[1]\n" if ($i<15);
+ $code.=" xor @T[1],@T[1]\n" if ($i==15);
+ my $k=1;
+ while (29*($j+1)<64*($i+1)) {
+ $code.=<<___;
+ mov @T[0],@T[-$k]
+ shr \$`29*$j`,@T[-$k]
+ and %rax,@T[-$k] # &0x1fffffff
+ mov @T[-$k],`8*$j-128`($out)
+___
+ $j++; $k++;
+ }
+ $code.=<<___;
+ shrd \$`29*$j`,@T[1],@T[0]
+ and %rax,@T[0]
+ mov @T[0],`8*$j-128`($out)
+___
+ $j++;
+ push(@T,shift(@T));
+}
+$code.=<<___;
+ mov @T[0],`8*$j-128`($out) # zero
+ mov @T[0],`8*($j+1)-128`($out)
+ mov @T[0],`8*($j+2)-128`($out)
+ mov @T[0],`8*($j+3)-128`($out)
+ ret
+.cfi_endproc
+.size rsaz_1024_norm2red_avx2,.-rsaz_1024_norm2red_avx2
+___
+}
+{
+my ($out,$inp,$power) = $win64 ? ("%rcx","%rdx","%r8d") : ("%rdi","%rsi","%edx");
+
+$code.=<<___;
+.globl rsaz_1024_scatter5_avx2
+.type rsaz_1024_scatter5_avx2,\@abi-omnipotent
+.align 32
+rsaz_1024_scatter5_avx2:
+.cfi_startproc
+ vzeroupper
+ vmovdqu .Lscatter_permd(%rip),%ymm5
+ shl \$4,$power
+ lea ($out,$power),$out
+ mov \$9,%eax
+ jmp .Loop_scatter_1024
+
+.align 32
+.Loop_scatter_1024:
+ vmovdqu ($inp),%ymm0
+ lea 32($inp),$inp
+ vpermd %ymm0,%ymm5,%ymm0
+ vmovdqu %xmm0,($out)
+ lea 16*32($out),$out
+ dec %eax
+ jnz .Loop_scatter_1024
+
+ vzeroupper
+ ret
+.cfi_endproc
+.size rsaz_1024_scatter5_avx2,.-rsaz_1024_scatter5_avx2
+
+.globl rsaz_1024_gather5_avx2
+.type rsaz_1024_gather5_avx2,\@abi-omnipotent
+.align 32
+rsaz_1024_gather5_avx2:
+.cfi_startproc
+ vzeroupper
+ mov %rsp,%r11
+.cfi_def_cfa_register %r11
+___
+$code.=<<___ if ($win64);
+ lea -0x88(%rsp),%rax
+.LSEH_begin_rsaz_1024_gather5:
+ # I can't trust assembler to use specific encoding:-(
+ .byte 0x48,0x8d,0x60,0xe0 # lea -0x20(%rax),%rsp
+ .byte 0xc5,0xf8,0x29,0x70,0xe0 # vmovaps %xmm6,-0x20(%rax)
+ .byte 0xc5,0xf8,0x29,0x78,0xf0 # vmovaps %xmm7,-0x10(%rax)
+ .byte 0xc5,0x78,0x29,0x40,0x00 # vmovaps %xmm8,0(%rax)
+ .byte 0xc5,0x78,0x29,0x48,0x10 # vmovaps %xmm9,0x10(%rax)
+ .byte 0xc5,0x78,0x29,0x50,0x20 # vmovaps %xmm10,0x20(%rax)
+ .byte 0xc5,0x78,0x29,0x58,0x30 # vmovaps %xmm11,0x30(%rax)
+ .byte 0xc5,0x78,0x29,0x60,0x40 # vmovaps %xmm12,0x40(%rax)
+ .byte 0xc5,0x78,0x29,0x68,0x50 # vmovaps %xmm13,0x50(%rax)
+ .byte 0xc5,0x78,0x29,0x70,0x60 # vmovaps %xmm14,0x60(%rax)
+ .byte 0xc5,0x78,0x29,0x78,0x70 # vmovaps %xmm15,0x70(%rax)
+___
+$code.=<<___;
+ lea -0x100(%rsp),%rsp
+ and \$-32, %rsp
+ lea .Linc(%rip), %r10
+ lea -128(%rsp),%rax # control u-op density
+
+ vmovd $power, %xmm4
+ vmovdqa (%r10),%ymm0
+ vmovdqa 32(%r10),%ymm1
+ vmovdqa 64(%r10),%ymm5
+ vpbroadcastd %xmm4,%ymm4
+
+ vpaddd %ymm5, %ymm0, %ymm2
+ vpcmpeqd %ymm4, %ymm0, %ymm0
+ vpaddd %ymm5, %ymm1, %ymm3
+ vpcmpeqd %ymm4, %ymm1, %ymm1
+ vmovdqa %ymm0, 32*0+128(%rax)
+ vpaddd %ymm5, %ymm2, %ymm0
+ vpcmpeqd %ymm4, %ymm2, %ymm2
+ vmovdqa %ymm1, 32*1+128(%rax)
+ vpaddd %ymm5, %ymm3, %ymm1
+ vpcmpeqd %ymm4, %ymm3, %ymm3
+ vmovdqa %ymm2, 32*2+128(%rax)
+ vpaddd %ymm5, %ymm0, %ymm2
+ vpcmpeqd %ymm4, %ymm0, %ymm0
+ vmovdqa %ymm3, 32*3+128(%rax)
+ vpaddd %ymm5, %ymm1, %ymm3
+ vpcmpeqd %ymm4, %ymm1, %ymm1
+ vmovdqa %ymm0, 32*4+128(%rax)
+ vpaddd %ymm5, %ymm2, %ymm8
+ vpcmpeqd %ymm4, %ymm2, %ymm2
+ vmovdqa %ymm1, 32*5+128(%rax)
+ vpaddd %ymm5, %ymm3, %ymm9
+ vpcmpeqd %ymm4, %ymm3, %ymm3
+ vmovdqa %ymm2, 32*6+128(%rax)
+ vpaddd %ymm5, %ymm8, %ymm10
+ vpcmpeqd %ymm4, %ymm8, %ymm8
+ vmovdqa %ymm3, 32*7+128(%rax)
+ vpaddd %ymm5, %ymm9, %ymm11
+ vpcmpeqd %ymm4, %ymm9, %ymm9
+ vpaddd %ymm5, %ymm10, %ymm12
+ vpcmpeqd %ymm4, %ymm10, %ymm10
+ vpaddd %ymm5, %ymm11, %ymm13
+ vpcmpeqd %ymm4, %ymm11, %ymm11
+ vpaddd %ymm5, %ymm12, %ymm14
+ vpcmpeqd %ymm4, %ymm12, %ymm12
+ vpaddd %ymm5, %ymm13, %ymm15
+ vpcmpeqd %ymm4, %ymm13, %ymm13
+ vpcmpeqd %ymm4, %ymm14, %ymm14
+ vpcmpeqd %ymm4, %ymm15, %ymm15
+
+ vmovdqa -32(%r10),%ymm7 # .Lgather_permd
+ lea 128($inp), $inp
+ mov \$9,$power
+
+.Loop_gather_1024:
+ vmovdqa 32*0-128($inp), %ymm0
+ vmovdqa 32*1-128($inp), %ymm1
+ vmovdqa 32*2-128($inp), %ymm2
+ vmovdqa 32*3-128($inp), %ymm3
+ vpand 32*0+128(%rax), %ymm0, %ymm0
+ vpand 32*1+128(%rax), %ymm1, %ymm1
+ vpand 32*2+128(%rax), %ymm2, %ymm2
+ vpor %ymm0, %ymm1, %ymm4
+ vpand 32*3+128(%rax), %ymm3, %ymm3
+ vmovdqa 32*4-128($inp), %ymm0
+ vmovdqa 32*5-128($inp), %ymm1
+ vpor %ymm2, %ymm3, %ymm5
+ vmovdqa 32*6-128($inp), %ymm2
+ vmovdqa 32*7-128($inp), %ymm3
+ vpand 32*4+128(%rax), %ymm0, %ymm0
+ vpand 32*5+128(%rax), %ymm1, %ymm1
+ vpand 32*6+128(%rax), %ymm2, %ymm2
+ vpor %ymm0, %ymm4, %ymm4
+ vpand 32*7+128(%rax), %ymm3, %ymm3
+ vpand 32*8-128($inp), %ymm8, %ymm0
+ vpor %ymm1, %ymm5, %ymm5
+ vpand 32*9-128($inp), %ymm9, %ymm1
+ vpor %ymm2, %ymm4, %ymm4
+ vpand 32*10-128($inp),%ymm10, %ymm2
+ vpor %ymm3, %ymm5, %ymm5
+ vpand 32*11-128($inp),%ymm11, %ymm3
+ vpor %ymm0, %ymm4, %ymm4
+ vpand 32*12-128($inp),%ymm12, %ymm0
+ vpor %ymm1, %ymm5, %ymm5
+ vpand 32*13-128($inp),%ymm13, %ymm1
+ vpor %ymm2, %ymm4, %ymm4
+ vpand 32*14-128($inp),%ymm14, %ymm2
+ vpor %ymm3, %ymm5, %ymm5
+ vpand 32*15-128($inp),%ymm15, %ymm3
+ lea 32*16($inp), $inp
+ vpor %ymm0, %ymm4, %ymm4
+ vpor %ymm1, %ymm5, %ymm5
+ vpor %ymm2, %ymm4, %ymm4
+ vpor %ymm3, %ymm5, %ymm5
+
+ vpor %ymm5, %ymm4, %ymm4
+ vextracti128 \$1, %ymm4, %xmm5 # upper half is cleared
+ vpor %xmm4, %xmm5, %xmm5
+ vpermd %ymm5,%ymm7,%ymm5
+ vmovdqu %ymm5,($out)
+ lea 32($out),$out
+ dec $power
+ jnz .Loop_gather_1024
+
+ vpxor %ymm0,%ymm0,%ymm0
+ vmovdqu %ymm0,($out)
+ vzeroupper
+___
+$code.=<<___ if ($win64);
+ movaps -0xa8(%r11),%xmm6
+ movaps -0x98(%r11),%xmm7
+ movaps -0x88(%r11),%xmm8
+ movaps -0x78(%r11),%xmm9
+ movaps -0x68(%r11),%xmm10
+ movaps -0x58(%r11),%xmm11
+ movaps -0x48(%r11),%xmm12
+ movaps -0x38(%r11),%xmm13
+ movaps -0x28(%r11),%xmm14
+ movaps -0x18(%r11),%xmm15
+___
+$code.=<<___;
+ lea (%r11),%rsp
+.cfi_def_cfa_register %rsp
+ ret
+.cfi_endproc
+.LSEH_end_rsaz_1024_gather5:
+.size rsaz_1024_gather5_avx2,.-rsaz_1024_gather5_avx2
+___
+}
+
+$code.=<<___;
+.extern OPENSSL_ia32cap_P
+.globl rsaz_avx2_eligible
+.type rsaz_avx2_eligible,\@abi-omnipotent
+.align 32
+rsaz_avx2_eligible:
+ mov OPENSSL_ia32cap_P+8(%rip),%eax
+___
+$code.=<<___ if ($addx);
+ mov \$`1<<8|1<<19`,%ecx
+ mov \$0,%edx
+ and %eax,%ecx
+ cmp \$`1<<8|1<<19`,%ecx # check for BMI2+AD*X
+ cmove %edx,%eax
+___
+$code.=<<___;
+ and \$`1<<5`,%eax
+ shr \$5,%eax
+ ret
+.size rsaz_avx2_eligible,.-rsaz_avx2_eligible
+
+.align 64
+.Land_mask:
+ .quad 0x1fffffff,0x1fffffff,0x1fffffff,0x1fffffff
+.Lscatter_permd:
+ .long 0,2,4,6,7,7,7,7
+.Lgather_permd:
+ .long 0,7,1,7,2,7,3,7
+.Linc:
+ .long 0,0,0,0, 1,1,1,1
+ .long 2,2,2,2, 3,3,3,3
+ .long 4,4,4,4, 4,4,4,4
+.align 64
+___
+
+if ($win64) {
+$rec="%rcx";
+$frame="%rdx";
+$context="%r8";
+$disp="%r9";
+
+$code.=<<___
+.extern __imp_RtlVirtualUnwind
+.type rsaz_se_handler,\@abi-omnipotent
+.align 16
+rsaz_se_handler:
+ push %rsi
+ push %rdi
+ push %rbx
+ push %rbp
+ push %r12
+ push %r13
+ push %r14
+ push %r15
+ pushfq
+ sub \$64,%rsp
+
+ mov 120($context),%rax # pull context->Rax
+ mov 248($context),%rbx # pull context->Rip
+
+ mov 8($disp),%rsi # disp->ImageBase
+ mov 56($disp),%r11 # disp->HandlerData
+
+ mov 0(%r11),%r10d # HandlerData[0]
+ lea (%rsi,%r10),%r10 # prologue label
+ cmp %r10,%rbx # context->Rip<prologue label
+ jb .Lcommon_seh_tail
+
+ mov 4(%r11),%r10d # HandlerData[1]
+ lea (%rsi,%r10),%r10 # epilogue label
+ cmp %r10,%rbx # context->Rip>=epilogue label
+ jae .Lcommon_seh_tail
+
+ mov 160($context),%rbp # pull context->Rbp
+
+ mov 8(%r11),%r10d # HandlerData[2]
+ lea (%rsi,%r10),%r10 # "in tail" label
+ cmp %r10,%rbx # context->Rip>="in tail" label
+ cmovc %rbp,%rax
+
+ mov -48(%rax),%r15
+ mov -40(%rax),%r14
+ mov -32(%rax),%r13
+ mov -24(%rax),%r12
+ mov -16(%rax),%rbp
+ mov -8(%rax),%rbx
+ mov %r15,240($context)
+ mov %r14,232($context)
+ mov %r13,224($context)
+ mov %r12,216($context)
+ mov %rbp,160($context)
+ mov %rbx,144($context)
+
+ lea -0xd8(%rax),%rsi # %xmm save area
+ lea 512($context),%rdi # & context.Xmm6
+ mov \$20,%ecx # 10*sizeof(%xmm0)/sizeof(%rax)
+ .long 0xa548f3fc # cld; rep movsq
+
+.Lcommon_seh_tail:
+ mov 8(%rax),%rdi
+ mov 16(%rax),%rsi
+ mov %rax,152($context) # restore context->Rsp
+ mov %rsi,168($context) # restore context->Rsi
+ mov %rdi,176($context) # restore context->Rdi
+
+ mov 40($disp),%rdi # disp->ContextRecord
+ mov $context,%rsi # context
+ mov \$154,%ecx # sizeof(CONTEXT)
+ .long 0xa548f3fc # cld; rep movsq
+
+ mov $disp,%rsi
+ xor %rcx,%rcx # arg1, UNW_FLAG_NHANDLER
+ mov 8(%rsi),%rdx # arg2, disp->ImageBase
+ mov 0(%rsi),%r8 # arg3, disp->ControlPc
+ mov 16(%rsi),%r9 # arg4, disp->FunctionEntry
+ mov 40(%rsi),%r10 # disp->ContextRecord
+ lea 56(%rsi),%r11 # &disp->HandlerData
+ lea 24(%rsi),%r12 # &disp->EstablisherFrame
+ mov %r10,32(%rsp) # arg5
+ mov %r11,40(%rsp) # arg6
+ mov %r12,48(%rsp) # arg7
+ mov %rcx,56(%rsp) # arg8, (NULL)
+ call *__imp_RtlVirtualUnwind(%rip)
+
+ mov \$1,%eax # ExceptionContinueSearch
+ add \$64,%rsp
+ popfq
+ pop %r15
+ pop %r14
+ pop %r13
+ pop %r12
+ pop %rbp
+ pop %rbx
+ pop %rdi
+ pop %rsi
+ ret
+.size rsaz_se_handler,.-rsaz_se_handler
+
+.section .pdata
+.align 4
+ .rva .LSEH_begin_rsaz_1024_sqr_avx2
+ .rva .LSEH_end_rsaz_1024_sqr_avx2
+ .rva .LSEH_info_rsaz_1024_sqr_avx2
+
+ .rva .LSEH_begin_rsaz_1024_mul_avx2
+ .rva .LSEH_end_rsaz_1024_mul_avx2
+ .rva .LSEH_info_rsaz_1024_mul_avx2
+
+ .rva .LSEH_begin_rsaz_1024_gather5
+ .rva .LSEH_end_rsaz_1024_gather5
+ .rva .LSEH_info_rsaz_1024_gather5
+.section .xdata
+.align 8
+.LSEH_info_rsaz_1024_sqr_avx2:
+ .byte 9,0,0,0
+ .rva rsaz_se_handler
+ .rva .Lsqr_1024_body,.Lsqr_1024_epilogue,.Lsqr_1024_in_tail
+ .long 0
+.LSEH_info_rsaz_1024_mul_avx2:
+ .byte 9,0,0,0
+ .rva rsaz_se_handler
+ .rva .Lmul_1024_body,.Lmul_1024_epilogue,.Lmul_1024_in_tail
+ .long 0
+.LSEH_info_rsaz_1024_gather5:
+ .byte 0x01,0x36,0x17,0x0b
+ .byte 0x36,0xf8,0x09,0x00 # vmovaps 0x90(rsp),xmm15
+ .byte 0x31,0xe8,0x08,0x00 # vmovaps 0x80(rsp),xmm14
+ .byte 0x2c,0xd8,0x07,0x00 # vmovaps 0x70(rsp),xmm13
+ .byte 0x27,0xc8,0x06,0x00 # vmovaps 0x60(rsp),xmm12
+ .byte 0x22,0xb8,0x05,0x00 # vmovaps 0x50(rsp),xmm11
+ .byte 0x1d,0xa8,0x04,0x00 # vmovaps 0x40(rsp),xmm10
+ .byte 0x18,0x98,0x03,0x00 # vmovaps 0x30(rsp),xmm9
+ .byte 0x13,0x88,0x02,0x00 # vmovaps 0x20(rsp),xmm8
+ .byte 0x0e,0x78,0x01,0x00 # vmovaps 0x10(rsp),xmm7
+ .byte 0x09,0x68,0x00,0x00 # vmovaps 0x00(rsp),xmm6
+ .byte 0x04,0x01,0x15,0x00 # sub rsp,0xa8
+ .byte 0x00,0xb3,0x00,0x00 # set_frame r11
+___
+}
+
+foreach (split("\n",$code)) {
+ s/\`([^\`]*)\`/eval($1)/ge;
+
+ s/\b(sh[rl]d?\s+\$)(-?[0-9]+)/$1.$2%64/ge or
+
+ s/\b(vmov[dq])\b(.+)%ymm([0-9]+)/$1$2%xmm$3/go or
+ s/\b(vmovdqu)\b(.+)%x%ymm([0-9]+)/$1$2%xmm$3/go or
+ s/\b(vpinsr[qd])\b(.+)%ymm([0-9]+)/$1$2%xmm$3/go or
+ s/\b(vpextr[qd])\b(.+)%ymm([0-9]+)/$1$2%xmm$3/go or
+ s/\b(vpbroadcast[qd]\s+)%ymm([0-9]+)/$1%xmm$2/go;
+ print $_,"\n";
+}
+
+}}} else {{{
+print <<___; # assembler is too old
+.text
+
+.globl rsaz_avx2_eligible
+.type rsaz_avx2_eligible,\@abi-omnipotent
+rsaz_avx2_eligible:
+ xor %eax,%eax
+ ret
+.size rsaz_avx2_eligible,.-rsaz_avx2_eligible
+
+.globl rsaz_1024_sqr_avx2
+.globl rsaz_1024_mul_avx2
+.globl rsaz_1024_norm2red_avx2
+.globl rsaz_1024_red2norm_avx2
+.globl rsaz_1024_scatter5_avx2
+.globl rsaz_1024_gather5_avx2
+.type rsaz_1024_sqr_avx2,\@abi-omnipotent
+rsaz_1024_sqr_avx2:
+rsaz_1024_mul_avx2:
+rsaz_1024_norm2red_avx2:
+rsaz_1024_red2norm_avx2:
+rsaz_1024_scatter5_avx2:
+rsaz_1024_gather5_avx2:
+ .byte 0x0f,0x0b # ud2
+ ret
+.size rsaz_1024_sqr_avx2,.-rsaz_1024_sqr_avx2
+___
+}}}
+
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/rsaz-x86_64.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/rsaz-x86_64.pl
new file mode 100755
index 0000000..cf08ce9
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/rsaz-x86_64.pl
@@ -0,0 +1,2431 @@
+#! /usr/bin/env perl
+# Copyright 2013-2020 The OpenSSL Project Authors. All Rights Reserved.
+# Copyright (c) 2012, Intel Corporation. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+#
+# Originally written by Shay Gueron (1, 2), and Vlad Krasnov (1)
+# (1) Intel Corporation, Israel Development Center, Haifa, Israel
+# (2) University of Haifa, Israel
+#
+# References:
+# [1] S. Gueron, "Efficient Software Implementations of Modular
+# Exponentiation", http://eprint.iacr.org/2011/239
+# [2] S. Gueron, V. Krasnov. "Speeding up Big-Numbers Squaring".
+# IEEE Proceedings of 9th International Conference on Information
+# Technology: New Generations (ITNG 2012), 821-823 (2012).
+# [3] S. Gueron, Efficient Software Implementations of Modular Exponentiation
+# Journal of Cryptographic Engineering 2:31-43 (2012).
+# [4] S. Gueron, V. Krasnov: "[PATCH] Efficient and side channel analysis
+# resistant 512-bit and 1024-bit modular exponentiation for optimizing
+# RSA1024 and RSA2048 on x86_64 platforms",
+# http://rt.openssl.org/Ticket/Display.html?id=2582&user=guest&pass=guest
+#
+# While original submission covers 512- and 1024-bit exponentiation,
+# this module is limited to 512-bit version only (and as such
+# accelerates RSA1024 sign). This is because improvement for longer
+# keys is not high enough to justify the effort, highest measured
+# was ~5% on Westmere. [This is relative to OpenSSL 1.0.2, upcoming
+# for the moment of this writing!] Nor does this module implement
+# "monolithic" complete exponentiation jumbo-subroutine, but adheres
+# to more modular mixture of C and assembly. And it's optimized even
+# for processors other than Intel Core family (see table below for
+# improvement coefficients).
+# <appro@openssl.org>
+#
+# RSA1024 sign/sec this/original |this/rsax(*) this/fips(*)
+# ----------------+---------------------------
+# Opteron +13% |+5% +20%
+# Bulldozer -0% |-1% +10%
+# P4 +11% |+7% +8%
+# Westmere +5% |+14% +17%
+# Sandy Bridge +2% |+12% +29%
+# Ivy Bridge +1% |+11% +35%
+# Haswell(**) -0% |+12% +39%
+# Atom +13% |+11% +4%
+# VIA Nano +70% |+9% +25%
+#
+# (*) rsax engine and fips numbers are presented for reference
+# purposes;
+# (**) MULX was attempted, but found to give only marginal improvement;
+
+$flavour = shift;
+$output = shift;
+if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
+
+$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or
+( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or
+die "can't locate x86_64-xlate.pl";
+
+open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\"";
+*STDOUT=*OUT;
+
+if (`$ENV{CC} -Wa,-v -c -o /dev/null -x assembler /dev/null 2>&1`
+ =~ /GNU assembler version ([2-9]\.[0-9]+)/) {
+ $addx = ($1>=2.23);
+}
+
+if (!$addx && $win64 && ($flavour =~ /nasm/ || $ENV{ASM} =~ /nasm/) &&
+ `nasm -v 2>&1` =~ /NASM version ([2-9]\.[0-9]+)/) {
+ $addx = ($1>=2.10);
+}
+
+if (!$addx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
+ `ml64 2>&1` =~ /Version ([0-9]+)\./) {
+ $addx = ($1>=12);
+}
+
+if (!$addx && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|.*based on LLVM) ([0-9]+)\.([0-9]+)/) {
+ my $ver = $2 + $3/100.0; # 3.1->3.01, 3.10->3.10
+ $addx = ($ver>=3.03);
+}
+
+($out, $inp, $mod) = ("%rdi", "%rsi", "%rbp"); # common internal API
+{
+my ($out,$inp,$mod,$n0,$times) = ("%rdi","%rsi","%rdx","%rcx","%r8d");
+
+$code.=<<___;
+.text
+
+.extern OPENSSL_ia32cap_P
+
+.globl rsaz_512_sqr
+.type rsaz_512_sqr,\@function,5
+.align 32
+rsaz_512_sqr: # 25-29% faster than rsaz_512_mul
+.cfi_startproc
+ push %rbx
+.cfi_push %rbx
+ push %rbp
+.cfi_push %rbp
+ push %r12
+.cfi_push %r12
+ push %r13
+.cfi_push %r13
+ push %r14
+.cfi_push %r14
+ push %r15
+.cfi_push %r15
+
+ subq \$128+24, %rsp
+.cfi_adjust_cfa_offset 128+24
+.Lsqr_body:
+ movq $mod, %xmm1 # common off-load
+ movq ($inp), %rdx
+ movq 8($inp), %rax
+ movq $n0, 128(%rsp)
+___
+$code.=<<___ if ($addx);
+ movl \$0x80100,%r11d
+ andl OPENSSL_ia32cap_P+8(%rip),%r11d
+ cmpl \$0x80100,%r11d # check for MULX and ADO/CX
+ je .Loop_sqrx
+___
+$code.=<<___;
+ jmp .Loop_sqr
+
+.align 32
+.Loop_sqr:
+ movl $times,128+8(%rsp)
+#first iteration
+ movq %rdx, %rbx # 0($inp)
+ mov %rax, %rbp # 8($inp)
+ mulq %rdx
+ movq %rax, %r8
+ movq 16($inp), %rax
+ movq %rdx, %r9
+
+ mulq %rbx
+ addq %rax, %r9
+ movq 24($inp), %rax
+ movq %rdx, %r10
+ adcq \$0, %r10
+
+ mulq %rbx
+ addq %rax, %r10
+ movq 32($inp), %rax
+ movq %rdx, %r11
+ adcq \$0, %r11
+
+ mulq %rbx
+ addq %rax, %r11
+ movq 40($inp), %rax
+ movq %rdx, %r12
+ adcq \$0, %r12
+
+ mulq %rbx
+ addq %rax, %r12
+ movq 48($inp), %rax
+ movq %rdx, %r13
+ adcq \$0, %r13
+
+ mulq %rbx
+ addq %rax, %r13
+ movq 56($inp), %rax
+ movq %rdx, %r14
+ adcq \$0, %r14
+
+ mulq %rbx
+ addq %rax, %r14
+ movq %rbx, %rax
+ adcq \$0, %rdx
+
+ xorq %rcx,%rcx # rcx:r8 = r8 << 1
+ addq %r8, %r8
+ movq %rdx, %r15
+ adcq \$0, %rcx
+
+ mulq %rax
+ addq %r8, %rdx
+ adcq \$0, %rcx
+
+ movq %rax, (%rsp)
+ movq %rdx, 8(%rsp)
+
+#second iteration
+ movq 16($inp), %rax
+ mulq %rbp
+ addq %rax, %r10
+ movq 24($inp), %rax
+ movq %rdx, %rbx
+ adcq \$0, %rbx
+
+ mulq %rbp
+ addq %rax, %r11
+ movq 32($inp), %rax
+ adcq \$0, %rdx
+ addq %rbx, %r11
+ movq %rdx, %rbx
+ adcq \$0, %rbx
+
+ mulq %rbp
+ addq %rax, %r12
+ movq 40($inp), %rax
+ adcq \$0, %rdx
+ addq %rbx, %r12
+ movq %rdx, %rbx
+ adcq \$0, %rbx
+
+ mulq %rbp
+ addq %rax, %r13
+ movq 48($inp), %rax
+ adcq \$0, %rdx
+ addq %rbx, %r13
+ movq %rdx, %rbx
+ adcq \$0, %rbx
+
+ mulq %rbp
+ addq %rax, %r14
+ movq 56($inp), %rax
+ adcq \$0, %rdx
+ addq %rbx, %r14
+ movq %rdx, %rbx
+ adcq \$0, %rbx
+
+ mulq %rbp
+ addq %rax, %r15
+ movq %rbp, %rax
+ adcq \$0, %rdx
+ addq %rbx, %r15
+ adcq \$0, %rdx
+
+ xorq %rbx, %rbx # rbx:r10:r9 = r10:r9 << 1
+ addq %r9, %r9
+ movq %rdx, %r8
+ adcq %r10, %r10
+ adcq \$0, %rbx
+
+ mulq %rax
+ # rcx <= 2 and rax <= 0xFFFF..F9, so carry must be zero here
+ addq %rcx, %rax
+ movq 16($inp), %rbp
+ addq %rax, %r9
+ movq 24($inp), %rax
+ adcq %rdx, %r10
+ adcq \$0, %rbx
+
+ movq %r9, 16(%rsp)
+ movq %r10, 24(%rsp)
+
+#third iteration
+ mulq %rbp
+ addq %rax, %r12
+ movq 32($inp), %rax
+ movq %rdx, %rcx
+ adcq \$0, %rcx
+
+ mulq %rbp
+ addq %rax, %r13
+ movq 40($inp), %rax
+ adcq \$0, %rdx
+ addq %rcx, %r13
+ movq %rdx, %rcx
+ adcq \$0, %rcx
+
+ mulq %rbp
+ addq %rax, %r14
+ movq 48($inp), %rax
+ adcq \$0, %rdx
+ addq %rcx, %r14
+ movq %rdx, %rcx
+ adcq \$0, %rcx
+
+ mulq %rbp
+ addq %rax, %r15
+ movq 56($inp), %rax
+ adcq \$0, %rdx
+ addq %rcx, %r15
+ movq %rdx, %rcx
+ adcq \$0, %rcx
+
+ mulq %rbp
+ addq %rax, %r8
+ movq %rbp, %rax
+ adcq \$0, %rdx
+ addq %rcx, %r8
+ adcq \$0, %rdx
+
+ xorq %rcx, %rcx # rcx:r12:r11 = r12:r11 << 1
+ addq %r11, %r11
+ movq %rdx, %r9
+ adcq %r12, %r12
+ adcq \$0, %rcx
+
+ mulq %rax
+ # rbx <= 2 and rax <= 0xFFFF..F9, so carry must be zero here
+ addq %rbx, %rax
+ movq 24($inp), %r10
+ addq %rax, %r11
+ movq 32($inp), %rax
+ adcq %rdx, %r12
+ adcq \$0, %rcx
+
+ movq %r11, 32(%rsp)
+ movq %r12, 40(%rsp)
+
+#fourth iteration
+ mov %rax, %r11 # 32($inp)
+ mulq %r10
+ addq %rax, %r14
+ movq 40($inp), %rax
+ movq %rdx, %rbx
+ adcq \$0, %rbx
+
+ mov %rax, %r12 # 40($inp)
+ mulq %r10
+ addq %rax, %r15
+ movq 48($inp), %rax
+ adcq \$0, %rdx
+ addq %rbx, %r15
+ movq %rdx, %rbx
+ adcq \$0, %rbx
+
+ mov %rax, %rbp # 48($inp)
+ mulq %r10
+ addq %rax, %r8
+ movq 56($inp), %rax
+ adcq \$0, %rdx
+ addq %rbx, %r8
+ movq %rdx, %rbx
+ adcq \$0, %rbx
+
+ mulq %r10
+ addq %rax, %r9
+ movq %r10, %rax
+ adcq \$0, %rdx
+ addq %rbx, %r9
+ adcq \$0, %rdx
+
+ xorq %rbx, %rbx # rbx:r13:r14 = r13:r14 << 1
+ addq %r13, %r13
+ movq %rdx, %r10
+ adcq %r14, %r14
+ adcq \$0, %rbx
+
+ mulq %rax
+ # rcx <= 2 and rax <= 0xFFFF..F9, so carry must be zero here
+ addq %rcx, %rax
+ addq %rax, %r13
+ movq %r12, %rax # 40($inp)
+ adcq %rdx, %r14
+ adcq \$0, %rbx
+
+ movq %r13, 48(%rsp)
+ movq %r14, 56(%rsp)
+
+#fifth iteration
+ mulq %r11
+ addq %rax, %r8
+ movq %rbp, %rax # 48($inp)
+ movq %rdx, %rcx
+ adcq \$0, %rcx
+
+ mulq %r11
+ addq %rax, %r9
+ movq 56($inp), %rax
+ adcq \$0, %rdx
+ addq %rcx, %r9
+ movq %rdx, %rcx
+ adcq \$0, %rcx
+
+ mov %rax, %r14 # 56($inp)
+ mulq %r11
+ addq %rax, %r10
+ movq %r11, %rax
+ adcq \$0, %rdx
+ addq %rcx, %r10
+ adcq \$0, %rdx
+
+ xorq %rcx, %rcx # rcx:r8:r15 = r8:r15 << 1
+ addq %r15, %r15
+ movq %rdx, %r11
+ adcq %r8, %r8
+ adcq \$0, %rcx
+
+ mulq %rax
+ # rbx <= 2 and rax <= 0xFFFF..F9, so carry must be zero here
+ addq %rbx, %rax
+ addq %rax, %r15
+ movq %rbp, %rax # 48($inp)
+ adcq %rdx, %r8
+ adcq \$0, %rcx
+
+ movq %r15, 64(%rsp)
+ movq %r8, 72(%rsp)
+
+#sixth iteration
+ mulq %r12
+ addq %rax, %r10
+ movq %r14, %rax # 56($inp)
+ movq %rdx, %rbx
+ adcq \$0, %rbx
+
+ mulq %r12
+ addq %rax, %r11
+ movq %r12, %rax
+ adcq \$0, %rdx
+ addq %rbx, %r11
+ adcq \$0, %rdx
+
+ xorq %rbx, %rbx # rbx:r10:r9 = r10:r9 << 1
+ addq %r9, %r9
+ movq %rdx, %r12
+ adcq %r10, %r10
+ adcq \$0, %rbx
+
+ mulq %rax
+ # rcx <= 2 and rax <= 0xFFFF..F9, so carry must be zero here
+ addq %rcx, %rax
+ addq %rax, %r9
+ movq %r14, %rax # 56($inp)
+ adcq %rdx, %r10
+ adcq \$0, %rbx
+
+ movq %r9, 80(%rsp)
+ movq %r10, 88(%rsp)
+
+#seventh iteration
+ mulq %rbp
+ addq %rax, %r12
+ movq %rbp, %rax
+ adcq \$0, %rdx
+
+ xorq %rcx, %rcx # rcx:r12:r11 = r12:r11 << 1
+ addq %r11, %r11
+ movq %rdx, %r13
+ adcq %r12, %r12
+ adcq \$0, %rcx
+
+ mulq %rax
+ # rbx <= 2 and rax <= 0xFFFF..F9, so carry must be zero here
+ addq %rbx, %rax
+ addq %rax, %r11
+ movq %r14, %rax # 56($inp)
+ adcq %rdx, %r12
+ adcq \$0, %rcx
+
+ movq %r11, 96(%rsp)
+ movq %r12, 104(%rsp)
+
+#eighth iteration
+ xorq %rbx, %rbx # rbx:r13 = r13 << 1
+ addq %r13, %r13
+ adcq \$0, %rbx
+
+ mulq %rax
+ # rcx <= 2 and rax <= 0xFFFF..F9, so carry must be zero here
+ addq %rcx, %rax
+ addq %r13, %rax
+ adcq %rbx, %rdx
+
+ movq (%rsp), %r8
+ movq 8(%rsp), %r9
+ movq 16(%rsp), %r10
+ movq 24(%rsp), %r11
+ movq 32(%rsp), %r12
+ movq 40(%rsp), %r13
+ movq 48(%rsp), %r14
+ movq 56(%rsp), %r15
+ movq %xmm1, %rbp
+
+ movq %rax, 112(%rsp)
+ movq %rdx, 120(%rsp)
+
+ call __rsaz_512_reduce
+
+ addq 64(%rsp), %r8
+ adcq 72(%rsp), %r9
+ adcq 80(%rsp), %r10
+ adcq 88(%rsp), %r11
+ adcq 96(%rsp), %r12
+ adcq 104(%rsp), %r13
+ adcq 112(%rsp), %r14
+ adcq 120(%rsp), %r15
+ sbbq %rcx, %rcx
+
+ call __rsaz_512_subtract
+
+ movq %r8, %rdx
+ movq %r9, %rax
+ movl 128+8(%rsp), $times
+ movq $out, $inp
+
+ decl $times
+ jnz .Loop_sqr
+___
+if ($addx) {
+$code.=<<___;
+ jmp .Lsqr_tail
+
+.align 32
+.Loop_sqrx:
+ movl $times,128+8(%rsp)
+ movq $out, %xmm0 # off-load
+#first iteration
+ mulx %rax, %r8, %r9
+ mov %rax, %rbx
+
+ mulx 16($inp), %rcx, %r10
+ xor %rbp, %rbp # cf=0, of=0
+
+ mulx 24($inp), %rax, %r11
+ adcx %rcx, %r9
+
+ .byte 0xc4,0x62,0xf3,0xf6,0xa6,0x20,0x00,0x00,0x00 # mulx 32($inp), %rcx, %r12
+ adcx %rax, %r10
+
+ .byte 0xc4,0x62,0xfb,0xf6,0xae,0x28,0x00,0x00,0x00 # mulx 40($inp), %rax, %r13
+ adcx %rcx, %r11
+
+ mulx 48($inp), %rcx, %r14
+ adcx %rax, %r12
+ adcx %rcx, %r13
+
+ mulx 56($inp), %rax, %r15
+ adcx %rax, %r14
+ adcx %rbp, %r15 # %rbp is 0
+
+ mulx %rdx, %rax, $out
+ mov %rbx, %rdx # 8($inp)
+ xor %rcx, %rcx
+ adox %r8, %r8
+ adcx $out, %r8
+ adox %rbp, %rcx
+ adcx %rbp, %rcx
+
+ mov %rax, (%rsp)
+ mov %r8, 8(%rsp)
+
+#second iteration
+ .byte 0xc4,0xe2,0xfb,0xf6,0x9e,0x10,0x00,0x00,0x00 # mulx 16($inp), %rax, %rbx
+ adox %rax, %r10
+ adcx %rbx, %r11
+
+ mulx 24($inp), $out, %r8
+ adox $out, %r11
+ .byte 0x66
+ adcx %r8, %r12
+
+ mulx 32($inp), %rax, %rbx
+ adox %rax, %r12
+ adcx %rbx, %r13
+
+ mulx 40($inp), $out, %r8
+ adox $out, %r13
+ adcx %r8, %r14
+
+ .byte 0xc4,0xe2,0xfb,0xf6,0x9e,0x30,0x00,0x00,0x00 # mulx 48($inp), %rax, %rbx
+ adox %rax, %r14
+ adcx %rbx, %r15
+
+ .byte 0xc4,0x62,0xc3,0xf6,0x86,0x38,0x00,0x00,0x00 # mulx 56($inp), $out, %r8
+ adox $out, %r15
+ adcx %rbp, %r8
+ mulx %rdx, %rax, $out
+ adox %rbp, %r8
+ .byte 0x48,0x8b,0x96,0x10,0x00,0x00,0x00 # mov 16($inp), %rdx
+
+ xor %rbx, %rbx
+ adox %r9, %r9
+ # rcx <= 2 and rax <= 0xFFFF..F9, so carry must be zero here
+ adcx %rcx, %rax
+ adox %r10, %r10
+ adcx %rax, %r9
+ adox %rbp, %rbx
+ adcx $out, %r10
+ adcx %rbp, %rbx
+
+ mov %r9, 16(%rsp)
+ .byte 0x4c,0x89,0x94,0x24,0x18,0x00,0x00,0x00 # mov %r10, 24(%rsp)
+
+#third iteration
+ mulx 24($inp), $out, %r9
+ adox $out, %r12
+ adcx %r9, %r13
+
+ mulx 32($inp), %rax, %rcx
+ adox %rax, %r13
+ adcx %rcx, %r14
+
+ .byte 0xc4,0x62,0xc3,0xf6,0x8e,0x28,0x00,0x00,0x00 # mulx 40($inp), $out, %r9
+ adox $out, %r14
+ adcx %r9, %r15
+
+ .byte 0xc4,0xe2,0xfb,0xf6,0x8e,0x30,0x00,0x00,0x00 # mulx 48($inp), %rax, %rcx
+ adox %rax, %r15
+ adcx %rcx, %r8
+
+ mulx 56($inp), $out, %r9
+ adox $out, %r8
+ adcx %rbp, %r9
+ mulx %rdx, %rax, $out
+ adox %rbp, %r9
+ mov 24($inp), %rdx
+
+ xor %rcx, %rcx
+ adox %r11, %r11
+ # rbx <= 2 and rax <= 0xFFFF..F9, so carry must be zero here
+ adcx %rbx, %rax
+ adox %r12, %r12
+ adcx %rax, %r11
+ adox %rbp, %rcx
+ adcx $out, %r12
+ adcx %rbp, %rcx
+
+ mov %r11, 32(%rsp)
+ mov %r12, 40(%rsp)
+
+#fourth iteration
+ mulx 32($inp), %rax, %rbx
+ adox %rax, %r14
+ adcx %rbx, %r15
+
+ mulx 40($inp), $out, %r10
+ adox $out, %r15
+ adcx %r10, %r8
+
+ mulx 48($inp), %rax, %rbx
+ adox %rax, %r8
+ adcx %rbx, %r9
+
+ mulx 56($inp), $out, %r10
+ adox $out, %r9
+ adcx %rbp, %r10
+ mulx %rdx, %rax, $out
+ adox %rbp, %r10
+ mov 32($inp), %rdx
+
+ xor %rbx, %rbx
+ adox %r13, %r13
+ # rcx <= 2 and rax <= 0xFFFF..F9, so carry must be zero here
+ adcx %rcx, %rax
+ adox %r14, %r14
+ adcx %rax, %r13
+ adox %rbp, %rbx
+ adcx $out, %r14
+ adcx %rbp, %rbx
+
+ mov %r13, 48(%rsp)
+ mov %r14, 56(%rsp)
+
+#fifth iteration
+ mulx 40($inp), $out, %r11
+ adox $out, %r8
+ adcx %r11, %r9
+
+ mulx 48($inp), %rax, %rcx
+ adox %rax, %r9
+ adcx %rcx, %r10
+
+ mulx 56($inp), $out, %r11
+ adox $out, %r10
+ adcx %rbp, %r11
+ mulx %rdx, %rax, $out
+ mov 40($inp), %rdx
+ adox %rbp, %r11
+
+ xor %rcx, %rcx
+ adox %r15, %r15
+ # rbx <= 2 and rax <= 0xFFFF..F9, so carry must be zero here
+ adcx %rbx, %rax
+ adox %r8, %r8
+ adcx %rax, %r15
+ adox %rbp, %rcx
+ adcx $out, %r8
+ adcx %rbp, %rcx
+
+ mov %r15, 64(%rsp)
+ mov %r8, 72(%rsp)
+
+#sixth iteration
+ .byte 0xc4,0xe2,0xfb,0xf6,0x9e,0x30,0x00,0x00,0x00 # mulx 48($inp), %rax, %rbx
+ adox %rax, %r10
+ adcx %rbx, %r11
+
+ .byte 0xc4,0x62,0xc3,0xf6,0xa6,0x38,0x00,0x00,0x00 # mulx 56($inp), $out, %r12
+ adox $out, %r11
+ adcx %rbp, %r12
+ mulx %rdx, %rax, $out
+ adox %rbp, %r12
+ mov 48($inp), %rdx
+
+ xor %rbx, %rbx
+ adox %r9, %r9
+ # rcx <= 2 and rax <= 0xFFFF..F9, so carry must be zero here
+ adcx %rcx, %rax
+ adox %r10, %r10
+ adcx %rax, %r9
+ adcx $out, %r10
+ adox %rbp, %rbx
+ adcx %rbp, %rbx
+
+ mov %r9, 80(%rsp)
+ mov %r10, 88(%rsp)
+
+#seventh iteration
+ .byte 0xc4,0x62,0xfb,0xf6,0xae,0x38,0x00,0x00,0x00 # mulx 56($inp), %rax, %r13
+ adox %rax, %r12
+ adox %rbp, %r13
+
+ mulx %rdx, %rax, $out
+ xor %rcx, %rcx
+ mov 56($inp), %rdx
+ adox %r11, %r11
+ # rbx <= 2 and rax <= 0xFFFF..F9, so carry must be zero here
+ adcx %rbx, %rax
+ adox %r12, %r12
+ adcx %rax, %r11
+ adox %rbp, %rcx
+ adcx $out, %r12
+ adcx %rbp, %rcx
+
+ .byte 0x4c,0x89,0x9c,0x24,0x60,0x00,0x00,0x00 # mov %r11, 96(%rsp)
+ .byte 0x4c,0x89,0xa4,0x24,0x68,0x00,0x00,0x00 # mov %r12, 104(%rsp)
+
+#eighth iteration
+ mulx %rdx, %rax, %rdx
+ xor %rbx, %rbx
+ adox %r13, %r13
+ # rcx <= 2 and rax <= 0xFFFF..F9, so carry must be zero here
+ adcx %rcx, %rax
+ adox %rbp, %rbx
+ adcx %r13, %rax
+ adcx %rdx, %rbx
+
+ movq %xmm0, $out
+ movq %xmm1, %rbp
+
+ movq 128(%rsp), %rdx # pull $n0
+ movq (%rsp), %r8
+ movq 8(%rsp), %r9
+ movq 16(%rsp), %r10
+ movq 24(%rsp), %r11
+ movq 32(%rsp), %r12
+ movq 40(%rsp), %r13
+ movq 48(%rsp), %r14
+ movq 56(%rsp), %r15
+
+ movq %rax, 112(%rsp)
+ movq %rbx, 120(%rsp)
+
+ call __rsaz_512_reducex
+
+ addq 64(%rsp), %r8
+ adcq 72(%rsp), %r9
+ adcq 80(%rsp), %r10
+ adcq 88(%rsp), %r11
+ adcq 96(%rsp), %r12
+ adcq 104(%rsp), %r13
+ adcq 112(%rsp), %r14
+ adcq 120(%rsp), %r15
+ sbbq %rcx, %rcx
+
+ call __rsaz_512_subtract
+
+ movq %r8, %rdx
+ movq %r9, %rax
+ movl 128+8(%rsp), $times
+ movq $out, $inp
+
+ decl $times
+ jnz .Loop_sqrx
+
+.Lsqr_tail:
+___
+}
+$code.=<<___;
+
+ leaq 128+24+48(%rsp), %rax
+.cfi_def_cfa %rax,8
+ movq -48(%rax), %r15
+.cfi_restore %r15
+ movq -40(%rax), %r14
+.cfi_restore %r14
+ movq -32(%rax), %r13
+.cfi_restore %r13
+ movq -24(%rax), %r12
+.cfi_restore %r12
+ movq -16(%rax), %rbp
+.cfi_restore %rbp
+ movq -8(%rax), %rbx
+.cfi_restore %rbx
+ leaq (%rax), %rsp
+.cfi_def_cfa_register %rsp
+.Lsqr_epilogue:
+ ret
+.cfi_endproc
+.size rsaz_512_sqr,.-rsaz_512_sqr
+___
+}
+{
+my ($out,$ap,$bp,$mod,$n0) = ("%rdi","%rsi","%rdx","%rcx","%r8");
+$code.=<<___;
+.globl rsaz_512_mul
+.type rsaz_512_mul,\@function,5
+.align 32
+rsaz_512_mul:
+.cfi_startproc
+ push %rbx
+.cfi_push %rbx
+ push %rbp
+.cfi_push %rbp
+ push %r12
+.cfi_push %r12
+ push %r13
+.cfi_push %r13
+ push %r14
+.cfi_push %r14
+ push %r15
+.cfi_push %r15
+
+ subq \$128+24, %rsp
+.cfi_adjust_cfa_offset 128+24
+.Lmul_body:
+ movq $out, %xmm0 # off-load arguments
+ movq $mod, %xmm1
+ movq $n0, 128(%rsp)
+___
+$code.=<<___ if ($addx);
+ movl \$0x80100,%r11d
+ andl OPENSSL_ia32cap_P+8(%rip),%r11d
+ cmpl \$0x80100,%r11d # check for MULX and ADO/CX
+ je .Lmulx
+___
+$code.=<<___;
+ movq ($bp), %rbx # pass b[0]
+ movq $bp, %rbp # pass argument
+ call __rsaz_512_mul
+
+ movq %xmm0, $out
+ movq %xmm1, %rbp
+
+ movq (%rsp), %r8
+ movq 8(%rsp), %r9
+ movq 16(%rsp), %r10
+ movq 24(%rsp), %r11
+ movq 32(%rsp), %r12
+ movq 40(%rsp), %r13
+ movq 48(%rsp), %r14
+ movq 56(%rsp), %r15
+
+ call __rsaz_512_reduce
+___
+$code.=<<___ if ($addx);
+ jmp .Lmul_tail
+
+.align 32
+.Lmulx:
+ movq $bp, %rbp # pass argument
+ movq ($bp), %rdx # pass b[0]
+ call __rsaz_512_mulx
+
+ movq %xmm0, $out
+ movq %xmm1, %rbp
+
+ movq 128(%rsp), %rdx # pull $n0
+ movq (%rsp), %r8
+ movq 8(%rsp), %r9
+ movq 16(%rsp), %r10
+ movq 24(%rsp), %r11
+ movq 32(%rsp), %r12
+ movq 40(%rsp), %r13
+ movq 48(%rsp), %r14
+ movq 56(%rsp), %r15
+
+ call __rsaz_512_reducex
+.Lmul_tail:
+___
+$code.=<<___;
+ addq 64(%rsp), %r8
+ adcq 72(%rsp), %r9
+ adcq 80(%rsp), %r10
+ adcq 88(%rsp), %r11
+ adcq 96(%rsp), %r12
+ adcq 104(%rsp), %r13
+ adcq 112(%rsp), %r14
+ adcq 120(%rsp), %r15
+ sbbq %rcx, %rcx
+
+ call __rsaz_512_subtract
+
+ leaq 128+24+48(%rsp), %rax
+.cfi_def_cfa %rax,8
+ movq -48(%rax), %r15
+.cfi_restore %r15
+ movq -40(%rax), %r14
+.cfi_restore %r14
+ movq -32(%rax), %r13
+.cfi_restore %r13
+ movq -24(%rax), %r12
+.cfi_restore %r12
+ movq -16(%rax), %rbp
+.cfi_restore %rbp
+ movq -8(%rax), %rbx
+.cfi_restore %rbx
+ leaq (%rax), %rsp
+.cfi_def_cfa_register %rsp
+.Lmul_epilogue:
+ ret
+.cfi_endproc
+.size rsaz_512_mul,.-rsaz_512_mul
+___
+}
+{
+my ($out,$ap,$bp,$mod,$n0,$pwr) = ("%rdi","%rsi","%rdx","%rcx","%r8","%r9d");
+$code.=<<___;
+.globl rsaz_512_mul_gather4
+.type rsaz_512_mul_gather4,\@function,6
+.align 32
+rsaz_512_mul_gather4:
+.cfi_startproc
+ push %rbx
+.cfi_push %rbx
+ push %rbp
+.cfi_push %rbp
+ push %r12
+.cfi_push %r12
+ push %r13
+.cfi_push %r13
+ push %r14
+.cfi_push %r14
+ push %r15
+.cfi_push %r15
+
+ subq \$`128+24+($win64?0xb0:0)`, %rsp
+.cfi_adjust_cfa_offset `128+24+($win64?0xb0:0)`
+___
+$code.=<<___ if ($win64);
+ movaps %xmm6,0xa0(%rsp)
+ movaps %xmm7,0xb0(%rsp)
+ movaps %xmm8,0xc0(%rsp)
+ movaps %xmm9,0xd0(%rsp)
+ movaps %xmm10,0xe0(%rsp)
+ movaps %xmm11,0xf0(%rsp)
+ movaps %xmm12,0x100(%rsp)
+ movaps %xmm13,0x110(%rsp)
+ movaps %xmm14,0x120(%rsp)
+ movaps %xmm15,0x130(%rsp)
+___
+$code.=<<___;
+.Lmul_gather4_body:
+ movd $pwr,%xmm8
+ movdqa .Linc+16(%rip),%xmm1 # 00000002000000020000000200000002
+ movdqa .Linc(%rip),%xmm0 # 00000001000000010000000000000000
+
+ pshufd \$0,%xmm8,%xmm8 # broadcast $power
+ movdqa %xmm1,%xmm7
+ movdqa %xmm1,%xmm2
+___
+########################################################################
+# calculate mask by comparing 0..15 to $power
+#
+for($i=0;$i<4;$i++) {
+$code.=<<___;
+ paddd %xmm`$i`,%xmm`$i+1`
+ pcmpeqd %xmm8,%xmm`$i`
+ movdqa %xmm7,%xmm`$i+3`
+___
+}
+for(;$i<7;$i++) {
+$code.=<<___;
+ paddd %xmm`$i`,%xmm`$i+1`
+ pcmpeqd %xmm8,%xmm`$i`
+___
+}
+$code.=<<___;
+ pcmpeqd %xmm8,%xmm7
+
+ movdqa 16*0($bp),%xmm8
+ movdqa 16*1($bp),%xmm9
+ movdqa 16*2($bp),%xmm10
+ movdqa 16*3($bp),%xmm11
+ pand %xmm0,%xmm8
+ movdqa 16*4($bp),%xmm12
+ pand %xmm1,%xmm9
+ movdqa 16*5($bp),%xmm13
+ pand %xmm2,%xmm10
+ movdqa 16*6($bp),%xmm14
+ pand %xmm3,%xmm11
+ movdqa 16*7($bp),%xmm15
+ leaq 128($bp), %rbp
+ pand %xmm4,%xmm12
+ pand %xmm5,%xmm13
+ pand %xmm6,%xmm14
+ pand %xmm7,%xmm15
+ por %xmm10,%xmm8
+ por %xmm11,%xmm9
+ por %xmm12,%xmm8
+ por %xmm13,%xmm9
+ por %xmm14,%xmm8
+ por %xmm15,%xmm9
+
+ por %xmm9,%xmm8
+ pshufd \$0x4e,%xmm8,%xmm9
+ por %xmm9,%xmm8
+___
+$code.=<<___ if ($addx);
+ movl \$0x80100,%r11d
+ andl OPENSSL_ia32cap_P+8(%rip),%r11d
+ cmpl \$0x80100,%r11d # check for MULX and ADO/CX
+ je .Lmulx_gather
+___
+$code.=<<___;
+ movq %xmm8,%rbx
+
+ movq $n0, 128(%rsp) # off-load arguments
+ movq $out, 128+8(%rsp)
+ movq $mod, 128+16(%rsp)
+
+ movq ($ap), %rax
+ movq 8($ap), %rcx
+ mulq %rbx # 0 iteration
+ movq %rax, (%rsp)
+ movq %rcx, %rax
+ movq %rdx, %r8
+
+ mulq %rbx
+ addq %rax, %r8
+ movq 16($ap), %rax
+ movq %rdx, %r9
+ adcq \$0, %r9
+
+ mulq %rbx
+ addq %rax, %r9
+ movq 24($ap), %rax
+ movq %rdx, %r10
+ adcq \$0, %r10
+
+ mulq %rbx
+ addq %rax, %r10
+ movq 32($ap), %rax
+ movq %rdx, %r11
+ adcq \$0, %r11
+
+ mulq %rbx
+ addq %rax, %r11
+ movq 40($ap), %rax
+ movq %rdx, %r12
+ adcq \$0, %r12
+
+ mulq %rbx
+ addq %rax, %r12
+ movq 48($ap), %rax
+ movq %rdx, %r13
+ adcq \$0, %r13
+
+ mulq %rbx
+ addq %rax, %r13
+ movq 56($ap), %rax
+ movq %rdx, %r14
+ adcq \$0, %r14
+
+ mulq %rbx
+ addq %rax, %r14
+ movq ($ap), %rax
+ movq %rdx, %r15
+ adcq \$0, %r15
+
+ leaq 8(%rsp), %rdi
+ movl \$7, %ecx
+ jmp .Loop_mul_gather
+
+.align 32
+.Loop_mul_gather:
+ movdqa 16*0(%rbp),%xmm8
+ movdqa 16*1(%rbp),%xmm9
+ movdqa 16*2(%rbp),%xmm10
+ movdqa 16*3(%rbp),%xmm11
+ pand %xmm0,%xmm8
+ movdqa 16*4(%rbp),%xmm12
+ pand %xmm1,%xmm9
+ movdqa 16*5(%rbp),%xmm13
+ pand %xmm2,%xmm10
+ movdqa 16*6(%rbp),%xmm14
+ pand %xmm3,%xmm11
+ movdqa 16*7(%rbp),%xmm15
+ leaq 128(%rbp), %rbp
+ pand %xmm4,%xmm12
+ pand %xmm5,%xmm13
+ pand %xmm6,%xmm14
+ pand %xmm7,%xmm15
+ por %xmm10,%xmm8
+ por %xmm11,%xmm9
+ por %xmm12,%xmm8
+ por %xmm13,%xmm9
+ por %xmm14,%xmm8
+ por %xmm15,%xmm9
+
+ por %xmm9,%xmm8
+ pshufd \$0x4e,%xmm8,%xmm9
+ por %xmm9,%xmm8
+ movq %xmm8,%rbx
+
+ mulq %rbx
+ addq %rax, %r8
+ movq 8($ap), %rax
+ movq %r8, (%rdi)
+ movq %rdx, %r8
+ adcq \$0, %r8
+
+ mulq %rbx
+ addq %rax, %r9
+ movq 16($ap), %rax
+ adcq \$0, %rdx
+ addq %r9, %r8
+ movq %rdx, %r9
+ adcq \$0, %r9
+
+ mulq %rbx
+ addq %rax, %r10
+ movq 24($ap), %rax
+ adcq \$0, %rdx
+ addq %r10, %r9
+ movq %rdx, %r10
+ adcq \$0, %r10
+
+ mulq %rbx
+ addq %rax, %r11
+ movq 32($ap), %rax
+ adcq \$0, %rdx
+ addq %r11, %r10
+ movq %rdx, %r11
+ adcq \$0, %r11
+
+ mulq %rbx
+ addq %rax, %r12
+ movq 40($ap), %rax
+ adcq \$0, %rdx
+ addq %r12, %r11
+ movq %rdx, %r12
+ adcq \$0, %r12
+
+ mulq %rbx
+ addq %rax, %r13
+ movq 48($ap), %rax
+ adcq \$0, %rdx
+ addq %r13, %r12
+ movq %rdx, %r13
+ adcq \$0, %r13
+
+ mulq %rbx
+ addq %rax, %r14
+ movq 56($ap), %rax
+ adcq \$0, %rdx
+ addq %r14, %r13
+ movq %rdx, %r14
+ adcq \$0, %r14
+
+ mulq %rbx
+ addq %rax, %r15
+ movq ($ap), %rax
+ adcq \$0, %rdx
+ addq %r15, %r14
+ movq %rdx, %r15
+ adcq \$0, %r15
+
+ leaq 8(%rdi), %rdi
+
+ decl %ecx
+ jnz .Loop_mul_gather
+
+ movq %r8, (%rdi)
+ movq %r9, 8(%rdi)
+ movq %r10, 16(%rdi)
+ movq %r11, 24(%rdi)
+ movq %r12, 32(%rdi)
+ movq %r13, 40(%rdi)
+ movq %r14, 48(%rdi)
+ movq %r15, 56(%rdi)
+
+ movq 128+8(%rsp), $out
+ movq 128+16(%rsp), %rbp
+
+ movq (%rsp), %r8
+ movq 8(%rsp), %r9
+ movq 16(%rsp), %r10
+ movq 24(%rsp), %r11
+ movq 32(%rsp), %r12
+ movq 40(%rsp), %r13
+ movq 48(%rsp), %r14
+ movq 56(%rsp), %r15
+
+ call __rsaz_512_reduce
+___
+$code.=<<___ if ($addx);
+ jmp .Lmul_gather_tail
+
+.align 32
+.Lmulx_gather:
+ movq %xmm8,%rdx
+
+ mov $n0, 128(%rsp) # off-load arguments
+ mov $out, 128+8(%rsp)
+ mov $mod, 128+16(%rsp)
+
+ mulx ($ap), %rbx, %r8 # 0 iteration
+ mov %rbx, (%rsp)
+ xor %edi, %edi # cf=0, of=0
+
+ mulx 8($ap), %rax, %r9
+
+ mulx 16($ap), %rbx, %r10
+ adcx %rax, %r8
+
+ mulx 24($ap), %rax, %r11
+ adcx %rbx, %r9
+
+ mulx 32($ap), %rbx, %r12
+ adcx %rax, %r10
+
+ mulx 40($ap), %rax, %r13
+ adcx %rbx, %r11
+
+ mulx 48($ap), %rbx, %r14
+ adcx %rax, %r12
+
+ mulx 56($ap), %rax, %r15
+ adcx %rbx, %r13
+ adcx %rax, %r14
+ .byte 0x67
+ mov %r8, %rbx
+ adcx %rdi, %r15 # %rdi is 0
+
+ mov \$-7, %rcx
+ jmp .Loop_mulx_gather
+
+.align 32
+.Loop_mulx_gather:
+ movdqa 16*0(%rbp),%xmm8
+ movdqa 16*1(%rbp),%xmm9
+ movdqa 16*2(%rbp),%xmm10
+ movdqa 16*3(%rbp),%xmm11
+ pand %xmm0,%xmm8
+ movdqa 16*4(%rbp),%xmm12
+ pand %xmm1,%xmm9
+ movdqa 16*5(%rbp),%xmm13
+ pand %xmm2,%xmm10
+ movdqa 16*6(%rbp),%xmm14
+ pand %xmm3,%xmm11
+ movdqa 16*7(%rbp),%xmm15
+ leaq 128(%rbp), %rbp
+ pand %xmm4,%xmm12
+ pand %xmm5,%xmm13
+ pand %xmm6,%xmm14
+ pand %xmm7,%xmm15
+ por %xmm10,%xmm8
+ por %xmm11,%xmm9
+ por %xmm12,%xmm8
+ por %xmm13,%xmm9
+ por %xmm14,%xmm8
+ por %xmm15,%xmm9
+
+ por %xmm9,%xmm8
+ pshufd \$0x4e,%xmm8,%xmm9
+ por %xmm9,%xmm8
+ movq %xmm8,%rdx
+
+ .byte 0xc4,0x62,0xfb,0xf6,0x86,0x00,0x00,0x00,0x00 # mulx ($ap), %rax, %r8
+ adcx %rax, %rbx
+ adox %r9, %r8
+
+ mulx 8($ap), %rax, %r9
+ adcx %rax, %r8
+ adox %r10, %r9
+
+ mulx 16($ap), %rax, %r10
+ adcx %rax, %r9
+ adox %r11, %r10
+
+ .byte 0xc4,0x62,0xfb,0xf6,0x9e,0x18,0x00,0x00,0x00 # mulx 24($ap), %rax, %r11
+ adcx %rax, %r10
+ adox %r12, %r11
+
+ mulx 32($ap), %rax, %r12
+ adcx %rax, %r11
+ adox %r13, %r12
+
+ mulx 40($ap), %rax, %r13
+ adcx %rax, %r12
+ adox %r14, %r13
+
+ .byte 0xc4,0x62,0xfb,0xf6,0xb6,0x30,0x00,0x00,0x00 # mulx 48($ap), %rax, %r14
+ adcx %rax, %r13
+ .byte 0x67
+ adox %r15, %r14
+
+ mulx 56($ap), %rax, %r15
+ mov %rbx, 64(%rsp,%rcx,8)
+ adcx %rax, %r14
+ adox %rdi, %r15
+ mov %r8, %rbx
+ adcx %rdi, %r15 # cf=0
+
+ inc %rcx # of=0
+ jnz .Loop_mulx_gather
+
+ mov %r8, 64(%rsp)
+ mov %r9, 64+8(%rsp)
+ mov %r10, 64+16(%rsp)
+ mov %r11, 64+24(%rsp)
+ mov %r12, 64+32(%rsp)
+ mov %r13, 64+40(%rsp)
+ mov %r14, 64+48(%rsp)
+ mov %r15, 64+56(%rsp)
+
+ mov 128(%rsp), %rdx # pull arguments
+ mov 128+8(%rsp), $out
+ mov 128+16(%rsp), %rbp
+
+ mov (%rsp), %r8
+ mov 8(%rsp), %r9
+ mov 16(%rsp), %r10
+ mov 24(%rsp), %r11
+ mov 32(%rsp), %r12
+ mov 40(%rsp), %r13
+ mov 48(%rsp), %r14
+ mov 56(%rsp), %r15
+
+ call __rsaz_512_reducex
+
+.Lmul_gather_tail:
+___
+$code.=<<___;
+ addq 64(%rsp), %r8
+ adcq 72(%rsp), %r9
+ adcq 80(%rsp), %r10
+ adcq 88(%rsp), %r11
+ adcq 96(%rsp), %r12
+ adcq 104(%rsp), %r13
+ adcq 112(%rsp), %r14
+ adcq 120(%rsp), %r15
+ sbbq %rcx, %rcx
+
+ call __rsaz_512_subtract
+
+ leaq 128+24+48(%rsp), %rax
+___
+$code.=<<___ if ($win64);
+ movaps 0xa0-0xc8(%rax),%xmm6
+ movaps 0xb0-0xc8(%rax),%xmm7
+ movaps 0xc0-0xc8(%rax),%xmm8
+ movaps 0xd0-0xc8(%rax),%xmm9
+ movaps 0xe0-0xc8(%rax),%xmm10
+ movaps 0xf0-0xc8(%rax),%xmm11
+ movaps 0x100-0xc8(%rax),%xmm12
+ movaps 0x110-0xc8(%rax),%xmm13
+ movaps 0x120-0xc8(%rax),%xmm14
+ movaps 0x130-0xc8(%rax),%xmm15
+ lea 0xb0(%rax),%rax
+___
+$code.=<<___;
+.cfi_def_cfa %rax,8
+ movq -48(%rax), %r15
+.cfi_restore %r15
+ movq -40(%rax), %r14
+.cfi_restore %r14
+ movq -32(%rax), %r13
+.cfi_restore %r13
+ movq -24(%rax), %r12
+.cfi_restore %r12
+ movq -16(%rax), %rbp
+.cfi_restore %rbp
+ movq -8(%rax), %rbx
+.cfi_restore %rbx
+ leaq (%rax), %rsp
+.cfi_def_cfa_register %rsp
+.Lmul_gather4_epilogue:
+ ret
+.cfi_endproc
+.size rsaz_512_mul_gather4,.-rsaz_512_mul_gather4
+___
+}
+{
+my ($out,$ap,$mod,$n0,$tbl,$pwr) = ("%rdi","%rsi","%rdx","%rcx","%r8","%r9d");
+$code.=<<___;
+.globl rsaz_512_mul_scatter4
+.type rsaz_512_mul_scatter4,\@function,6
+.align 32
+rsaz_512_mul_scatter4:
+.cfi_startproc
+ push %rbx
+.cfi_push %rbx
+ push %rbp
+.cfi_push %rbp
+ push %r12
+.cfi_push %r12
+ push %r13
+.cfi_push %r13
+ push %r14
+.cfi_push %r14
+ push %r15
+.cfi_push %r15
+
+ mov $pwr, $pwr
+ subq \$128+24, %rsp
+.cfi_adjust_cfa_offset 128+24
+.Lmul_scatter4_body:
+ leaq ($tbl,$pwr,8), $tbl
+ movq $out, %xmm0 # off-load arguments
+ movq $mod, %xmm1
+ movq $tbl, %xmm2
+ movq $n0, 128(%rsp)
+
+ movq $out, %rbp
+___
+$code.=<<___ if ($addx);
+ movl \$0x80100,%r11d
+ andl OPENSSL_ia32cap_P+8(%rip),%r11d
+ cmpl \$0x80100,%r11d # check for MULX and ADO/CX
+ je .Lmulx_scatter
+___
+$code.=<<___;
+ movq ($out),%rbx # pass b[0]
+ call __rsaz_512_mul
+
+ movq %xmm0, $out
+ movq %xmm1, %rbp
+
+ movq (%rsp), %r8
+ movq 8(%rsp), %r9
+ movq 16(%rsp), %r10
+ movq 24(%rsp), %r11
+ movq 32(%rsp), %r12
+ movq 40(%rsp), %r13
+ movq 48(%rsp), %r14
+ movq 56(%rsp), %r15
+
+ call __rsaz_512_reduce
+___
+$code.=<<___ if ($addx);
+ jmp .Lmul_scatter_tail
+
+.align 32
+.Lmulx_scatter:
+ movq ($out), %rdx # pass b[0]
+ call __rsaz_512_mulx
+
+ movq %xmm0, $out
+ movq %xmm1, %rbp
+
+ movq 128(%rsp), %rdx # pull $n0
+ movq (%rsp), %r8
+ movq 8(%rsp), %r9
+ movq 16(%rsp), %r10
+ movq 24(%rsp), %r11
+ movq 32(%rsp), %r12
+ movq 40(%rsp), %r13
+ movq 48(%rsp), %r14
+ movq 56(%rsp), %r15
+
+ call __rsaz_512_reducex
+
+.Lmul_scatter_tail:
+___
+$code.=<<___;
+ addq 64(%rsp), %r8
+ adcq 72(%rsp), %r9
+ adcq 80(%rsp), %r10
+ adcq 88(%rsp), %r11
+ adcq 96(%rsp), %r12
+ adcq 104(%rsp), %r13
+ adcq 112(%rsp), %r14
+ adcq 120(%rsp), %r15
+ movq %xmm2, $inp
+ sbbq %rcx, %rcx
+
+ call __rsaz_512_subtract
+
+ movq %r8, 128*0($inp) # scatter
+ movq %r9, 128*1($inp)
+ movq %r10, 128*2($inp)
+ movq %r11, 128*3($inp)
+ movq %r12, 128*4($inp)
+ movq %r13, 128*5($inp)
+ movq %r14, 128*6($inp)
+ movq %r15, 128*7($inp)
+
+ leaq 128+24+48(%rsp), %rax
+.cfi_def_cfa %rax,8
+ movq -48(%rax), %r15
+.cfi_restore %r15
+ movq -40(%rax), %r14
+.cfi_restore %r14
+ movq -32(%rax), %r13
+.cfi_restore %r13
+ movq -24(%rax), %r12
+.cfi_restore %r12
+ movq -16(%rax), %rbp
+.cfi_restore %rbp
+ movq -8(%rax), %rbx
+.cfi_restore %rbx
+ leaq (%rax), %rsp
+.cfi_def_cfa_register %rsp
+.Lmul_scatter4_epilogue:
+ ret
+.cfi_endproc
+.size rsaz_512_mul_scatter4,.-rsaz_512_mul_scatter4
+___
+}
+{
+my ($out,$inp,$mod,$n0) = ("%rdi","%rsi","%rdx","%rcx");
+$code.=<<___;
+.globl rsaz_512_mul_by_one
+.type rsaz_512_mul_by_one,\@function,4
+.align 32
+rsaz_512_mul_by_one:
+.cfi_startproc
+ push %rbx
+.cfi_push %rbx
+ push %rbp
+.cfi_push %rbp
+ push %r12
+.cfi_push %r12
+ push %r13
+.cfi_push %r13
+ push %r14
+.cfi_push %r14
+ push %r15
+.cfi_push %r15
+
+ subq \$128+24, %rsp
+.cfi_adjust_cfa_offset 128+24
+.Lmul_by_one_body:
+___
+$code.=<<___ if ($addx);
+ movl OPENSSL_ia32cap_P+8(%rip),%eax
+___
+$code.=<<___;
+ movq $mod, %rbp # reassign argument
+ movq $n0, 128(%rsp)
+
+ movq ($inp), %r8
+ pxor %xmm0, %xmm0
+ movq 8($inp), %r9
+ movq 16($inp), %r10
+ movq 24($inp), %r11
+ movq 32($inp), %r12
+ movq 40($inp), %r13
+ movq 48($inp), %r14
+ movq 56($inp), %r15
+
+ movdqa %xmm0, (%rsp)
+ movdqa %xmm0, 16(%rsp)
+ movdqa %xmm0, 32(%rsp)
+ movdqa %xmm0, 48(%rsp)
+ movdqa %xmm0, 64(%rsp)
+ movdqa %xmm0, 80(%rsp)
+ movdqa %xmm0, 96(%rsp)
+___
+$code.=<<___ if ($addx);
+ andl \$0x80100,%eax
+ cmpl \$0x80100,%eax # check for MULX and ADO/CX
+ je .Lby_one_callx
+___
+$code.=<<___;
+ call __rsaz_512_reduce
+___
+$code.=<<___ if ($addx);
+ jmp .Lby_one_tail
+.align 32
+.Lby_one_callx:
+ movq 128(%rsp), %rdx # pull $n0
+ call __rsaz_512_reducex
+.Lby_one_tail:
+___
+$code.=<<___;
+ movq %r8, ($out)
+ movq %r9, 8($out)
+ movq %r10, 16($out)
+ movq %r11, 24($out)
+ movq %r12, 32($out)
+ movq %r13, 40($out)
+ movq %r14, 48($out)
+ movq %r15, 56($out)
+
+ leaq 128+24+48(%rsp), %rax
+.cfi_def_cfa %rax,8
+ movq -48(%rax), %r15
+.cfi_restore %r15
+ movq -40(%rax), %r14
+.cfi_restore %r14
+ movq -32(%rax), %r13
+.cfi_restore %r13
+ movq -24(%rax), %r12
+.cfi_restore %r12
+ movq -16(%rax), %rbp
+.cfi_restore %rbp
+ movq -8(%rax), %rbx
+.cfi_restore %rbx
+ leaq (%rax), %rsp
+.cfi_def_cfa_register %rsp
+.Lmul_by_one_epilogue:
+ ret
+.cfi_endproc
+.size rsaz_512_mul_by_one,.-rsaz_512_mul_by_one
+___
+}
+{ # __rsaz_512_reduce
+ #
+ # input: %r8-%r15, %rbp - mod, 128(%rsp) - n0
+ # output: %r8-%r15
+ # clobbers: everything except %rbp and %rdi
+$code.=<<___;
+.type __rsaz_512_reduce,\@abi-omnipotent
+.align 32
+__rsaz_512_reduce:
+.cfi_startproc
+ movq %r8, %rbx
+ imulq 128+8(%rsp), %rbx
+ movq 0(%rbp), %rax
+ movl \$8, %ecx
+ jmp .Lreduction_loop
+
+.align 32
+.Lreduction_loop:
+ mulq %rbx
+ movq 8(%rbp), %rax
+ negq %r8
+ movq %rdx, %r8
+ adcq \$0, %r8
+
+ mulq %rbx
+ addq %rax, %r9
+ movq 16(%rbp), %rax
+ adcq \$0, %rdx
+ addq %r9, %r8
+ movq %rdx, %r9
+ adcq \$0, %r9
+
+ mulq %rbx
+ addq %rax, %r10
+ movq 24(%rbp), %rax
+ adcq \$0, %rdx
+ addq %r10, %r9
+ movq %rdx, %r10
+ adcq \$0, %r10
+
+ mulq %rbx
+ addq %rax, %r11
+ movq 32(%rbp), %rax
+ adcq \$0, %rdx
+ addq %r11, %r10
+ movq 128+8(%rsp), %rsi
+ #movq %rdx, %r11
+ #adcq \$0, %r11
+ adcq \$0, %rdx
+ movq %rdx, %r11
+
+ mulq %rbx
+ addq %rax, %r12
+ movq 40(%rbp), %rax
+ adcq \$0, %rdx
+ imulq %r8, %rsi
+ addq %r12, %r11
+ movq %rdx, %r12
+ adcq \$0, %r12
+
+ mulq %rbx
+ addq %rax, %r13
+ movq 48(%rbp), %rax
+ adcq \$0, %rdx
+ addq %r13, %r12
+ movq %rdx, %r13
+ adcq \$0, %r13
+
+ mulq %rbx
+ addq %rax, %r14
+ movq 56(%rbp), %rax
+ adcq \$0, %rdx
+ addq %r14, %r13
+ movq %rdx, %r14
+ adcq \$0, %r14
+
+ mulq %rbx
+ movq %rsi, %rbx
+ addq %rax, %r15
+ movq 0(%rbp), %rax
+ adcq \$0, %rdx
+ addq %r15, %r14
+ movq %rdx, %r15
+ adcq \$0, %r15
+
+ decl %ecx
+ jne .Lreduction_loop
+
+ ret
+.cfi_endproc
+.size __rsaz_512_reduce,.-__rsaz_512_reduce
+___
+}
+if ($addx) {
+ # __rsaz_512_reducex
+ #
+ # input: %r8-%r15, %rbp - mod, 128(%rsp) - n0
+ # output: %r8-%r15
+ # clobbers: everything except %rbp and %rdi
+$code.=<<___;
+.type __rsaz_512_reducex,\@abi-omnipotent
+.align 32
+__rsaz_512_reducex:
+.cfi_startproc
+ #movq 128+8(%rsp), %rdx # pull $n0
+ imulq %r8, %rdx
+ xorq %rsi, %rsi # cf=0,of=0
+ movl \$8, %ecx
+ jmp .Lreduction_loopx
+
+.align 32
+.Lreduction_loopx:
+ mov %r8, %rbx
+ mulx 0(%rbp), %rax, %r8
+ adcx %rbx, %rax
+ adox %r9, %r8
+
+ mulx 8(%rbp), %rax, %r9
+ adcx %rax, %r8
+ adox %r10, %r9
+
+ mulx 16(%rbp), %rbx, %r10
+ adcx %rbx, %r9
+ adox %r11, %r10
+
+ mulx 24(%rbp), %rbx, %r11
+ adcx %rbx, %r10
+ adox %r12, %r11
+
+ .byte 0xc4,0x62,0xe3,0xf6,0xa5,0x20,0x00,0x00,0x00 # mulx 32(%rbp), %rbx, %r12
+ mov %rdx, %rax
+ mov %r8, %rdx
+ adcx %rbx, %r11
+ adox %r13, %r12
+
+ mulx 128+8(%rsp), %rbx, %rdx
+ mov %rax, %rdx
+
+ mulx 40(%rbp), %rax, %r13
+ adcx %rax, %r12
+ adox %r14, %r13
+
+ .byte 0xc4,0x62,0xfb,0xf6,0xb5,0x30,0x00,0x00,0x00 # mulx 48(%rbp), %rax, %r14
+ adcx %rax, %r13
+ adox %r15, %r14
+
+ mulx 56(%rbp), %rax, %r15
+ mov %rbx, %rdx
+ adcx %rax, %r14
+ adox %rsi, %r15 # %rsi is 0
+ adcx %rsi, %r15 # cf=0
+
+ decl %ecx # of=0
+ jne .Lreduction_loopx
+
+ ret
+.cfi_endproc
+.size __rsaz_512_reducex,.-__rsaz_512_reducex
+___
+}
+{ # __rsaz_512_subtract
+ # input: %r8-%r15, %rdi - $out, %rbp - $mod, %rcx - mask
+ # output:
+ # clobbers: everything but %rdi, %rsi and %rbp
+$code.=<<___;
+.type __rsaz_512_subtract,\@abi-omnipotent
+.align 32
+__rsaz_512_subtract:
+.cfi_startproc
+ movq %r8, ($out)
+ movq %r9, 8($out)
+ movq %r10, 16($out)
+ movq %r11, 24($out)
+ movq %r12, 32($out)
+ movq %r13, 40($out)
+ movq %r14, 48($out)
+ movq %r15, 56($out)
+
+ movq 0($mod), %r8
+ movq 8($mod), %r9
+ negq %r8
+ notq %r9
+ andq %rcx, %r8
+ movq 16($mod), %r10
+ andq %rcx, %r9
+ notq %r10
+ movq 24($mod), %r11
+ andq %rcx, %r10
+ notq %r11
+ movq 32($mod), %r12
+ andq %rcx, %r11
+ notq %r12
+ movq 40($mod), %r13
+ andq %rcx, %r12
+ notq %r13
+ movq 48($mod), %r14
+ andq %rcx, %r13
+ notq %r14
+ movq 56($mod), %r15
+ andq %rcx, %r14
+ notq %r15
+ andq %rcx, %r15
+
+ addq ($out), %r8
+ adcq 8($out), %r9
+ adcq 16($out), %r10
+ adcq 24($out), %r11
+ adcq 32($out), %r12
+ adcq 40($out), %r13
+ adcq 48($out), %r14
+ adcq 56($out), %r15
+
+ movq %r8, ($out)
+ movq %r9, 8($out)
+ movq %r10, 16($out)
+ movq %r11, 24($out)
+ movq %r12, 32($out)
+ movq %r13, 40($out)
+ movq %r14, 48($out)
+ movq %r15, 56($out)
+
+ ret
+.cfi_endproc
+.size __rsaz_512_subtract,.-__rsaz_512_subtract
+___
+}
+{ # __rsaz_512_mul
+ #
+ # input: %rsi - ap, %rbp - bp
+ # output:
+ # clobbers: everything
+my ($ap,$bp) = ("%rsi","%rbp");
+$code.=<<___;
+.type __rsaz_512_mul,\@abi-omnipotent
+.align 32
+__rsaz_512_mul:
+.cfi_startproc
+ leaq 8(%rsp), %rdi
+
+ movq ($ap), %rax
+ mulq %rbx
+ movq %rax, (%rdi)
+ movq 8($ap), %rax
+ movq %rdx, %r8
+
+ mulq %rbx
+ addq %rax, %r8
+ movq 16($ap), %rax
+ movq %rdx, %r9
+ adcq \$0, %r9
+
+ mulq %rbx
+ addq %rax, %r9
+ movq 24($ap), %rax
+ movq %rdx, %r10
+ adcq \$0, %r10
+
+ mulq %rbx
+ addq %rax, %r10
+ movq 32($ap), %rax
+ movq %rdx, %r11
+ adcq \$0, %r11
+
+ mulq %rbx
+ addq %rax, %r11
+ movq 40($ap), %rax
+ movq %rdx, %r12
+ adcq \$0, %r12
+
+ mulq %rbx
+ addq %rax, %r12
+ movq 48($ap), %rax
+ movq %rdx, %r13
+ adcq \$0, %r13
+
+ mulq %rbx
+ addq %rax, %r13
+ movq 56($ap), %rax
+ movq %rdx, %r14
+ adcq \$0, %r14
+
+ mulq %rbx
+ addq %rax, %r14
+ movq ($ap), %rax
+ movq %rdx, %r15
+ adcq \$0, %r15
+
+ leaq 8($bp), $bp
+ leaq 8(%rdi), %rdi
+
+ movl \$7, %ecx
+ jmp .Loop_mul
+
+.align 32
+.Loop_mul:
+ movq ($bp), %rbx
+ mulq %rbx
+ addq %rax, %r8
+ movq 8($ap), %rax
+ movq %r8, (%rdi)
+ movq %rdx, %r8
+ adcq \$0, %r8
+
+ mulq %rbx
+ addq %rax, %r9
+ movq 16($ap), %rax
+ adcq \$0, %rdx
+ addq %r9, %r8
+ movq %rdx, %r9
+ adcq \$0, %r9
+
+ mulq %rbx
+ addq %rax, %r10
+ movq 24($ap), %rax
+ adcq \$0, %rdx
+ addq %r10, %r9
+ movq %rdx, %r10
+ adcq \$0, %r10
+
+ mulq %rbx
+ addq %rax, %r11
+ movq 32($ap), %rax
+ adcq \$0, %rdx
+ addq %r11, %r10
+ movq %rdx, %r11
+ adcq \$0, %r11
+
+ mulq %rbx
+ addq %rax, %r12
+ movq 40($ap), %rax
+ adcq \$0, %rdx
+ addq %r12, %r11
+ movq %rdx, %r12
+ adcq \$0, %r12
+
+ mulq %rbx
+ addq %rax, %r13
+ movq 48($ap), %rax
+ adcq \$0, %rdx
+ addq %r13, %r12
+ movq %rdx, %r13
+ adcq \$0, %r13
+
+ mulq %rbx
+ addq %rax, %r14
+ movq 56($ap), %rax
+ adcq \$0, %rdx
+ addq %r14, %r13
+ movq %rdx, %r14
+ leaq 8($bp), $bp
+ adcq \$0, %r14
+
+ mulq %rbx
+ addq %rax, %r15
+ movq ($ap), %rax
+ adcq \$0, %rdx
+ addq %r15, %r14
+ movq %rdx, %r15
+ adcq \$0, %r15
+
+ leaq 8(%rdi), %rdi
+
+ decl %ecx
+ jnz .Loop_mul
+
+ movq %r8, (%rdi)
+ movq %r9, 8(%rdi)
+ movq %r10, 16(%rdi)
+ movq %r11, 24(%rdi)
+ movq %r12, 32(%rdi)
+ movq %r13, 40(%rdi)
+ movq %r14, 48(%rdi)
+ movq %r15, 56(%rdi)
+
+ ret
+.cfi_endproc
+.size __rsaz_512_mul,.-__rsaz_512_mul
+___
+}
+if ($addx) {
+ # __rsaz_512_mulx
+ #
+ # input: %rsi - ap, %rbp - bp
+ # output:
+ # clobbers: everything
+my ($ap,$bp,$zero) = ("%rsi","%rbp","%rdi");
+$code.=<<___;
+.type __rsaz_512_mulx,\@abi-omnipotent
+.align 32
+__rsaz_512_mulx:
+.cfi_startproc
+ mulx ($ap), %rbx, %r8 # initial %rdx preloaded by caller
+ mov \$-6, %rcx
+
+ mulx 8($ap), %rax, %r9
+ movq %rbx, 8(%rsp)
+
+ mulx 16($ap), %rbx, %r10
+ adc %rax, %r8
+
+ mulx 24($ap), %rax, %r11
+ adc %rbx, %r9
+
+ mulx 32($ap), %rbx, %r12
+ adc %rax, %r10
+
+ mulx 40($ap), %rax, %r13
+ adc %rbx, %r11
+
+ mulx 48($ap), %rbx, %r14
+ adc %rax, %r12
+
+ mulx 56($ap), %rax, %r15
+ mov 8($bp), %rdx
+ adc %rbx, %r13
+ adc %rax, %r14
+ adc \$0, %r15
+
+ xor $zero, $zero # cf=0,of=0
+ jmp .Loop_mulx
+
+.align 32
+.Loop_mulx:
+ movq %r8, %rbx
+ mulx ($ap), %rax, %r8
+ adcx %rax, %rbx
+ adox %r9, %r8
+
+ mulx 8($ap), %rax, %r9
+ adcx %rax, %r8
+ adox %r10, %r9
+
+ mulx 16($ap), %rax, %r10
+ adcx %rax, %r9
+ adox %r11, %r10
+
+ mulx 24($ap), %rax, %r11
+ adcx %rax, %r10
+ adox %r12, %r11
+
+ .byte 0x3e,0xc4,0x62,0xfb,0xf6,0xa6,0x20,0x00,0x00,0x00 # mulx 32($ap), %rax, %r12
+ adcx %rax, %r11
+ adox %r13, %r12
+
+ mulx 40($ap), %rax, %r13
+ adcx %rax, %r12
+ adox %r14, %r13
+
+ mulx 48($ap), %rax, %r14
+ adcx %rax, %r13
+ adox %r15, %r14
+
+ mulx 56($ap), %rax, %r15
+ movq 64($bp,%rcx,8), %rdx
+ movq %rbx, 8+64-8(%rsp,%rcx,8)
+ adcx %rax, %r14
+ adox $zero, %r15
+ adcx $zero, %r15 # cf=0
+
+ inc %rcx # of=0
+ jnz .Loop_mulx
+
+ movq %r8, %rbx
+ mulx ($ap), %rax, %r8
+ adcx %rax, %rbx
+ adox %r9, %r8
+
+ .byte 0xc4,0x62,0xfb,0xf6,0x8e,0x08,0x00,0x00,0x00 # mulx 8($ap), %rax, %r9
+ adcx %rax, %r8
+ adox %r10, %r9
+
+ .byte 0xc4,0x62,0xfb,0xf6,0x96,0x10,0x00,0x00,0x00 # mulx 16($ap), %rax, %r10
+ adcx %rax, %r9
+ adox %r11, %r10
+
+ mulx 24($ap), %rax, %r11
+ adcx %rax, %r10
+ adox %r12, %r11
+
+ mulx 32($ap), %rax, %r12
+ adcx %rax, %r11
+ adox %r13, %r12
+
+ mulx 40($ap), %rax, %r13
+ adcx %rax, %r12
+ adox %r14, %r13
+
+ .byte 0xc4,0x62,0xfb,0xf6,0xb6,0x30,0x00,0x00,0x00 # mulx 48($ap), %rax, %r14
+ adcx %rax, %r13
+ adox %r15, %r14
+
+ .byte 0xc4,0x62,0xfb,0xf6,0xbe,0x38,0x00,0x00,0x00 # mulx 56($ap), %rax, %r15
+ adcx %rax, %r14
+ adox $zero, %r15
+ adcx $zero, %r15
+
+ mov %rbx, 8+64-8(%rsp)
+ mov %r8, 8+64(%rsp)
+ mov %r9, 8+64+8(%rsp)
+ mov %r10, 8+64+16(%rsp)
+ mov %r11, 8+64+24(%rsp)
+ mov %r12, 8+64+32(%rsp)
+ mov %r13, 8+64+40(%rsp)
+ mov %r14, 8+64+48(%rsp)
+ mov %r15, 8+64+56(%rsp)
+
+ ret
+.cfi_endproc
+.size __rsaz_512_mulx,.-__rsaz_512_mulx
+___
+}
+{
+my ($out,$inp,$power)= $win64 ? ("%rcx","%rdx","%r8d") : ("%rdi","%rsi","%edx");
+$code.=<<___;
+.globl rsaz_512_scatter4
+.type rsaz_512_scatter4,\@abi-omnipotent
+.align 16
+rsaz_512_scatter4:
+.cfi_startproc
+ leaq ($out,$power,8), $out
+ movl \$8, %r9d
+ jmp .Loop_scatter
+.align 16
+.Loop_scatter:
+ movq ($inp), %rax
+ leaq 8($inp), $inp
+ movq %rax, ($out)
+ leaq 128($out), $out
+ decl %r9d
+ jnz .Loop_scatter
+ ret
+.cfi_endproc
+.size rsaz_512_scatter4,.-rsaz_512_scatter4
+
+.globl rsaz_512_gather4
+.type rsaz_512_gather4,\@abi-omnipotent
+.align 16
+rsaz_512_gather4:
+.cfi_startproc
+___
+$code.=<<___ if ($win64);
+.LSEH_begin_rsaz_512_gather4:
+ .byte 0x48,0x81,0xec,0xa8,0x00,0x00,0x00 # sub $0xa8,%rsp
+ .byte 0x0f,0x29,0x34,0x24 # movaps %xmm6,(%rsp)
+ .byte 0x0f,0x29,0x7c,0x24,0x10 # movaps %xmm7,0x10(%rsp)
+ .byte 0x44,0x0f,0x29,0x44,0x24,0x20 # movaps %xmm8,0x20(%rsp)
+ .byte 0x44,0x0f,0x29,0x4c,0x24,0x30 # movaps %xmm9,0x30(%rsp)
+ .byte 0x44,0x0f,0x29,0x54,0x24,0x40 # movaps %xmm10,0x40(%rsp)
+ .byte 0x44,0x0f,0x29,0x5c,0x24,0x50 # movaps %xmm11,0x50(%rsp)
+ .byte 0x44,0x0f,0x29,0x64,0x24,0x60 # movaps %xmm12,0x60(%rsp)
+ .byte 0x44,0x0f,0x29,0x6c,0x24,0x70 # movaps %xmm13,0x70(%rsp)
+ .byte 0x44,0x0f,0x29,0xb4,0x24,0x80,0,0,0 # movaps %xmm14,0x80(%rsp)
+ .byte 0x44,0x0f,0x29,0xbc,0x24,0x90,0,0,0 # movaps %xmm15,0x90(%rsp)
+___
+$code.=<<___;
+ movd $power,%xmm8
+ movdqa .Linc+16(%rip),%xmm1 # 00000002000000020000000200000002
+ movdqa .Linc(%rip),%xmm0 # 00000001000000010000000000000000
+
+ pshufd \$0,%xmm8,%xmm8 # broadcast $power
+ movdqa %xmm1,%xmm7
+ movdqa %xmm1,%xmm2
+___
+########################################################################
+# calculate mask by comparing 0..15 to $power
+#
+for($i=0;$i<4;$i++) {
+$code.=<<___;
+ paddd %xmm`$i`,%xmm`$i+1`
+ pcmpeqd %xmm8,%xmm`$i`
+ movdqa %xmm7,%xmm`$i+3`
+___
+}
+for(;$i<7;$i++) {
+$code.=<<___;
+ paddd %xmm`$i`,%xmm`$i+1`
+ pcmpeqd %xmm8,%xmm`$i`
+___
+}
+$code.=<<___;
+ pcmpeqd %xmm8,%xmm7
+ movl \$8, %r9d
+ jmp .Loop_gather
+.align 16
+.Loop_gather:
+ movdqa 16*0($inp),%xmm8
+ movdqa 16*1($inp),%xmm9
+ movdqa 16*2($inp),%xmm10
+ movdqa 16*3($inp),%xmm11
+ pand %xmm0,%xmm8
+ movdqa 16*4($inp),%xmm12
+ pand %xmm1,%xmm9
+ movdqa 16*5($inp),%xmm13
+ pand %xmm2,%xmm10
+ movdqa 16*6($inp),%xmm14
+ pand %xmm3,%xmm11
+ movdqa 16*7($inp),%xmm15
+ leaq 128($inp), $inp
+ pand %xmm4,%xmm12
+ pand %xmm5,%xmm13
+ pand %xmm6,%xmm14
+ pand %xmm7,%xmm15
+ por %xmm10,%xmm8
+ por %xmm11,%xmm9
+ por %xmm12,%xmm8
+ por %xmm13,%xmm9
+ por %xmm14,%xmm8
+ por %xmm15,%xmm9
+
+ por %xmm9,%xmm8
+ pshufd \$0x4e,%xmm8,%xmm9
+ por %xmm9,%xmm8
+ movq %xmm8,($out)
+ leaq 8($out), $out
+ decl %r9d
+ jnz .Loop_gather
+___
+$code.=<<___ if ($win64);
+ movaps 0x00(%rsp),%xmm6
+ movaps 0x10(%rsp),%xmm7
+ movaps 0x20(%rsp),%xmm8
+ movaps 0x30(%rsp),%xmm9
+ movaps 0x40(%rsp),%xmm10
+ movaps 0x50(%rsp),%xmm11
+ movaps 0x60(%rsp),%xmm12
+ movaps 0x70(%rsp),%xmm13
+ movaps 0x80(%rsp),%xmm14
+ movaps 0x90(%rsp),%xmm15
+ add \$0xa8,%rsp
+___
+$code.=<<___;
+ ret
+.LSEH_end_rsaz_512_gather4:
+.cfi_endproc
+.size rsaz_512_gather4,.-rsaz_512_gather4
+
+.align 64
+.Linc:
+ .long 0,0, 1,1
+ .long 2,2, 2,2
+___
+}
+
+# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame,
+# CONTEXT *context,DISPATCHER_CONTEXT *disp)
+if ($win64) {
+$rec="%rcx";
+$frame="%rdx";
+$context="%r8";
+$disp="%r9";
+
+$code.=<<___;
+.extern __imp_RtlVirtualUnwind
+.type se_handler,\@abi-omnipotent
+.align 16
+se_handler:
+ push %rsi
+ push %rdi
+ push %rbx
+ push %rbp
+ push %r12
+ push %r13
+ push %r14
+ push %r15
+ pushfq
+ sub \$64,%rsp
+
+ mov 120($context),%rax # pull context->Rax
+ mov 248($context),%rbx # pull context->Rip
+
+ mov 8($disp),%rsi # disp->ImageBase
+ mov 56($disp),%r11 # disp->HandlerData
+
+ mov 0(%r11),%r10d # HandlerData[0]
+ lea (%rsi,%r10),%r10 # end of prologue label
+ cmp %r10,%rbx # context->Rip<end of prologue label
+ jb .Lcommon_seh_tail
+
+ mov 152($context),%rax # pull context->Rsp
+
+ mov 4(%r11),%r10d # HandlerData[1]
+ lea (%rsi,%r10),%r10 # epilogue label
+ cmp %r10,%rbx # context->Rip>=epilogue label
+ jae .Lcommon_seh_tail
+
+ lea 128+24+48(%rax),%rax
+
+ lea .Lmul_gather4_epilogue(%rip),%rbx
+ cmp %r10,%rbx
+ jne .Lse_not_in_mul_gather4
+
+ lea 0xb0(%rax),%rax
+
+ lea -48-0xa8(%rax),%rsi
+ lea 512($context),%rdi
+ mov \$20,%ecx
+ .long 0xa548f3fc # cld; rep movsq
+
+.Lse_not_in_mul_gather4:
+ mov -8(%rax),%rbx
+ mov -16(%rax),%rbp
+ mov -24(%rax),%r12
+ mov -32(%rax),%r13
+ mov -40(%rax),%r14
+ mov -48(%rax),%r15
+ mov %rbx,144($context) # restore context->Rbx
+ mov %rbp,160($context) # restore context->Rbp
+ mov %r12,216($context) # restore context->R12
+ mov %r13,224($context) # restore context->R13
+ mov %r14,232($context) # restore context->R14
+ mov %r15,240($context) # restore context->R15
+
+.Lcommon_seh_tail:
+ mov 8(%rax),%rdi
+ mov 16(%rax),%rsi
+ mov %rax,152($context) # restore context->Rsp
+ mov %rsi,168($context) # restore context->Rsi
+ mov %rdi,176($context) # restore context->Rdi
+
+ mov 40($disp),%rdi # disp->ContextRecord
+ mov $context,%rsi # context
+ mov \$154,%ecx # sizeof(CONTEXT)
+ .long 0xa548f3fc # cld; rep movsq
+
+ mov $disp,%rsi
+ xor %rcx,%rcx # arg1, UNW_FLAG_NHANDLER
+ mov 8(%rsi),%rdx # arg2, disp->ImageBase
+ mov 0(%rsi),%r8 # arg3, disp->ControlPc
+ mov 16(%rsi),%r9 # arg4, disp->FunctionEntry
+ mov 40(%rsi),%r10 # disp->ContextRecord
+ lea 56(%rsi),%r11 # &disp->HandlerData
+ lea 24(%rsi),%r12 # &disp->EstablisherFrame
+ mov %r10,32(%rsp) # arg5
+ mov %r11,40(%rsp) # arg6
+ mov %r12,48(%rsp) # arg7
+ mov %rcx,56(%rsp) # arg8, (NULL)
+ call *__imp_RtlVirtualUnwind(%rip)
+
+ mov \$1,%eax # ExceptionContinueSearch
+ add \$64,%rsp
+ popfq
+ pop %r15
+ pop %r14
+ pop %r13
+ pop %r12
+ pop %rbp
+ pop %rbx
+ pop %rdi
+ pop %rsi
+ ret
+.size se_handler,.-se_handler
+
+.section .pdata
+.align 4
+ .rva .LSEH_begin_rsaz_512_sqr
+ .rva .LSEH_end_rsaz_512_sqr
+ .rva .LSEH_info_rsaz_512_sqr
+
+ .rva .LSEH_begin_rsaz_512_mul
+ .rva .LSEH_end_rsaz_512_mul
+ .rva .LSEH_info_rsaz_512_mul
+
+ .rva .LSEH_begin_rsaz_512_mul_gather4
+ .rva .LSEH_end_rsaz_512_mul_gather4
+ .rva .LSEH_info_rsaz_512_mul_gather4
+
+ .rva .LSEH_begin_rsaz_512_mul_scatter4
+ .rva .LSEH_end_rsaz_512_mul_scatter4
+ .rva .LSEH_info_rsaz_512_mul_scatter4
+
+ .rva .LSEH_begin_rsaz_512_mul_by_one
+ .rva .LSEH_end_rsaz_512_mul_by_one
+ .rva .LSEH_info_rsaz_512_mul_by_one
+
+ .rva .LSEH_begin_rsaz_512_gather4
+ .rva .LSEH_end_rsaz_512_gather4
+ .rva .LSEH_info_rsaz_512_gather4
+
+.section .xdata
+.align 8
+.LSEH_info_rsaz_512_sqr:
+ .byte 9,0,0,0
+ .rva se_handler
+ .rva .Lsqr_body,.Lsqr_epilogue # HandlerData[]
+.LSEH_info_rsaz_512_mul:
+ .byte 9,0,0,0
+ .rva se_handler
+ .rva .Lmul_body,.Lmul_epilogue # HandlerData[]
+.LSEH_info_rsaz_512_mul_gather4:
+ .byte 9,0,0,0
+ .rva se_handler
+ .rva .Lmul_gather4_body,.Lmul_gather4_epilogue # HandlerData[]
+.LSEH_info_rsaz_512_mul_scatter4:
+ .byte 9,0,0,0
+ .rva se_handler
+ .rva .Lmul_scatter4_body,.Lmul_scatter4_epilogue # HandlerData[]
+.LSEH_info_rsaz_512_mul_by_one:
+ .byte 9,0,0,0
+ .rva se_handler
+ .rva .Lmul_by_one_body,.Lmul_by_one_epilogue # HandlerData[]
+.LSEH_info_rsaz_512_gather4:
+ .byte 0x01,0x46,0x16,0x00
+ .byte 0x46,0xf8,0x09,0x00 # vmovaps 0x90(rsp),xmm15
+ .byte 0x3d,0xe8,0x08,0x00 # vmovaps 0x80(rsp),xmm14
+ .byte 0x34,0xd8,0x07,0x00 # vmovaps 0x70(rsp),xmm13
+ .byte 0x2e,0xc8,0x06,0x00 # vmovaps 0x60(rsp),xmm12
+ .byte 0x28,0xb8,0x05,0x00 # vmovaps 0x50(rsp),xmm11
+ .byte 0x22,0xa8,0x04,0x00 # vmovaps 0x40(rsp),xmm10
+ .byte 0x1c,0x98,0x03,0x00 # vmovaps 0x30(rsp),xmm9
+ .byte 0x16,0x88,0x02,0x00 # vmovaps 0x20(rsp),xmm8
+ .byte 0x10,0x78,0x01,0x00 # vmovaps 0x10(rsp),xmm7
+ .byte 0x0b,0x68,0x00,0x00 # vmovaps 0x00(rsp),xmm6
+ .byte 0x07,0x01,0x15,0x00 # sub rsp,0xa8
+___
+}
+
+$code =~ s/\`([^\`]*)\`/eval $1/gem;
+print $code;
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/s390x-gf2m.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/s390x-gf2m.pl
new file mode 100644
index 0000000..a7e4b8a
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/s390x-gf2m.pl
@@ -0,0 +1,228 @@
+#! /usr/bin/env perl
+# Copyright 2011-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+#
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+#
+# May 2011
+#
+# The module implements bn_GF2m_mul_2x2 polynomial multiplication used
+# in bn_gf2m.c. It's kind of low-hanging mechanical port from C for
+# the time being... gcc 4.3 appeared to generate poor code, therefore
+# the effort. And indeed, the module delivers 55%-90%(*) improvement
+# on heaviest ECDSA verify and ECDH benchmarks for 163- and 571-bit
+# key lengths on z990, 30%-55%(*) - on z10, and 70%-110%(*) - on z196.
+# This is for 64-bit build. In 32-bit "highgprs" case improvement is
+# even higher, for example on z990 it was measured 80%-150%. ECDSA
+# sign is modest 9%-12% faster. Keep in mind that these coefficients
+# are not ones for bn_GF2m_mul_2x2 itself, as not all CPU time is
+# burnt in it...
+#
+# (*) gcc 4.1 was observed to deliver better results than gcc 4.3,
+# so that improvement coefficients can vary from one specific
+# setup to another.
+
+$flavour = shift;
+
+if ($flavour =~ /3[12]/) {
+ $SIZE_T=4;
+ $g="";
+} else {
+ $SIZE_T=8;
+ $g="g";
+}
+
+while (($output=shift) && ($output!~/\w[\w\-]*\.\w+$/)) {}
+open STDOUT,">$output";
+
+$stdframe=16*$SIZE_T+4*8;
+
+$rp="%r2";
+$a1="%r3";
+$a0="%r4";
+$b1="%r5";
+$b0="%r6";
+
+$ra="%r14";
+$sp="%r15";
+
+@T=("%r0","%r1");
+@i=("%r12","%r13");
+
+($a1,$a2,$a4,$a8,$a12,$a48)=map("%r$_",(6..11));
+($lo,$hi,$b)=map("%r$_",(3..5)); $a=$lo; $mask=$a8;
+
+$code.=<<___;
+.text
+
+.type _mul_1x1,\@function
+.align 16
+_mul_1x1:
+ lgr $a1,$a
+ sllg $a2,$a,1
+ sllg $a4,$a,2
+ sllg $a8,$a,3
+
+ srag $lo,$a1,63 # broadcast 63rd bit
+ nihh $a1,0x1fff
+ srag @i[0],$a2,63 # broadcast 62nd bit
+ nihh $a2,0x3fff
+ srag @i[1],$a4,63 # broadcast 61st bit
+ nihh $a4,0x7fff
+ ngr $lo,$b
+ ngr @i[0],$b
+ ngr @i[1],$b
+
+ lghi @T[0],0
+ lgr $a12,$a1
+ stg @T[0],`$stdframe+0*8`($sp) # tab[0]=0
+ xgr $a12,$a2
+ stg $a1,`$stdframe+1*8`($sp) # tab[1]=a1
+ lgr $a48,$a4
+ stg $a2,`$stdframe+2*8`($sp) # tab[2]=a2
+ xgr $a48,$a8
+ stg $a12,`$stdframe+3*8`($sp) # tab[3]=a1^a2
+ xgr $a1,$a4
+
+ stg $a4,`$stdframe+4*8`($sp) # tab[4]=a4
+ xgr $a2,$a4
+ stg $a1,`$stdframe+5*8`($sp) # tab[5]=a1^a4
+ xgr $a12,$a4
+ stg $a2,`$stdframe+6*8`($sp) # tab[6]=a2^a4
+ xgr $a1,$a48
+ stg $a12,`$stdframe+7*8`($sp) # tab[7]=a1^a2^a4
+ xgr $a2,$a48
+
+ stg $a8,`$stdframe+8*8`($sp) # tab[8]=a8
+ xgr $a12,$a48
+ stg $a1,`$stdframe+9*8`($sp) # tab[9]=a1^a8
+ xgr $a1,$a4
+ stg $a2,`$stdframe+10*8`($sp) # tab[10]=a2^a8
+ xgr $a2,$a4
+ stg $a12,`$stdframe+11*8`($sp) # tab[11]=a1^a2^a8
+
+ xgr $a12,$a4
+ stg $a48,`$stdframe+12*8`($sp) # tab[12]=a4^a8
+ srlg $hi,$lo,1
+ stg $a1,`$stdframe+13*8`($sp) # tab[13]=a1^a4^a8
+ sllg $lo,$lo,63
+ stg $a2,`$stdframe+14*8`($sp) # tab[14]=a2^a4^a8
+ srlg @T[0],@i[0],2
+ stg $a12,`$stdframe+15*8`($sp) # tab[15]=a1^a2^a4^a8
+
+ lghi $mask,`0xf<<3`
+ sllg $a1,@i[0],62
+ sllg @i[0],$b,3
+ srlg @T[1],@i[1],3
+ ngr @i[0],$mask
+ sllg $a2,@i[1],61
+ srlg @i[1],$b,4-3
+ xgr $hi,@T[0]
+ ngr @i[1],$mask
+ xgr $lo,$a1
+ xgr $hi,@T[1]
+ xgr $lo,$a2
+
+ xg $lo,$stdframe(@i[0],$sp)
+ srlg @i[0],$b,8-3
+ ngr @i[0],$mask
+___
+for($n=1;$n<14;$n++) {
+$code.=<<___;
+ lg @T[1],$stdframe(@i[1],$sp)
+ srlg @i[1],$b,`($n+2)*4`-3
+ sllg @T[0],@T[1],`$n*4`
+ ngr @i[1],$mask
+ srlg @T[1],@T[1],`64-$n*4`
+ xgr $lo,@T[0]
+ xgr $hi,@T[1]
+___
+ push(@i,shift(@i)); push(@T,shift(@T));
+}
+$code.=<<___;
+ lg @T[1],$stdframe(@i[1],$sp)
+ sllg @T[0],@T[1],`$n*4`
+ srlg @T[1],@T[1],`64-$n*4`
+ xgr $lo,@T[0]
+ xgr $hi,@T[1]
+
+ lg @T[0],$stdframe(@i[0],$sp)
+ sllg @T[1],@T[0],`($n+1)*4`
+ srlg @T[0],@T[0],`64-($n+1)*4`
+ xgr $lo,@T[1]
+ xgr $hi,@T[0]
+
+ br $ra
+.size _mul_1x1,.-_mul_1x1
+
+.globl bn_GF2m_mul_2x2
+.type bn_GF2m_mul_2x2,\@function
+.align 16
+bn_GF2m_mul_2x2:
+ stm${g} %r3,%r15,3*$SIZE_T($sp)
+
+ lghi %r1,-$stdframe-128
+ la %r0,0($sp)
+ la $sp,0(%r1,$sp) # alloca
+ st${g} %r0,0($sp) # back chain
+___
+if ($SIZE_T==8) {
+my @r=map("%r$_",(6..9));
+$code.=<<___;
+ bras $ra,_mul_1x1 # a1·b1
+ stmg $lo,$hi,16($rp)
+
+ lg $a,`$stdframe+128+4*$SIZE_T`($sp)
+ lg $b,`$stdframe+128+6*$SIZE_T`($sp)
+ bras $ra,_mul_1x1 # a0·b0
+ stmg $lo,$hi,0($rp)
+
+ lg $a,`$stdframe+128+3*$SIZE_T`($sp)
+ lg $b,`$stdframe+128+5*$SIZE_T`($sp)
+ xg $a,`$stdframe+128+4*$SIZE_T`($sp)
+ xg $b,`$stdframe+128+6*$SIZE_T`($sp)
+ bras $ra,_mul_1x1 # (a0+a1)·(b0+b1)
+ lmg @r[0],@r[3],0($rp)
+
+ xgr $lo,$hi
+ xgr $hi,@r[1]
+ xgr $lo,@r[0]
+ xgr $hi,@r[2]
+ xgr $lo,@r[3]
+ xgr $hi,@r[3]
+ xgr $lo,$hi
+ stg $hi,16($rp)
+ stg $lo,8($rp)
+___
+} else {
+$code.=<<___;
+ sllg %r3,%r3,32
+ sllg %r5,%r5,32
+ or %r3,%r4
+ or %r5,%r6
+ bras $ra,_mul_1x1
+ rllg $lo,$lo,32
+ rllg $hi,$hi,32
+ stmg $lo,$hi,0($rp)
+___
+}
+$code.=<<___;
+ lm${g} %r6,%r15,`$stdframe+128+6*$SIZE_T`($sp)
+ br $ra
+.size bn_GF2m_mul_2x2,.-bn_GF2m_mul_2x2
+.string "GF(2^m) Multiplication for s390x, CRYPTOGAMS by <appro\@openssl.org>"
+___
+
+$code =~ s/\`([^\`]*)\`/eval($1)/gem;
+print $code;
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/s390x-mont.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/s390x-mont.pl
new file mode 100644
index 0000000..bc8c895
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/s390x-mont.pl
@@ -0,0 +1,284 @@
+#! /usr/bin/env perl
+# Copyright 2007-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+
+# April 2007.
+#
+# Performance improvement over vanilla C code varies from 85% to 45%
+# depending on key length and benchmark. Unfortunately in this context
+# these are not very impressive results [for code that utilizes "wide"
+# 64x64=128-bit multiplication, which is not commonly available to C
+# programmers], at least hand-coded bn_asm.c replacement is known to
+# provide 30-40% better results for longest keys. Well, on a second
+# thought it's not very surprising, because z-CPUs are single-issue
+# and _strictly_ in-order execution, while bn_mul_mont is more or less
+# dependent on CPU ability to pipe-line instructions and have several
+# of them "in-flight" at the same time. I mean while other methods,
+# for example Karatsuba, aim to minimize amount of multiplications at
+# the cost of other operations increase, bn_mul_mont aim to neatly
+# "overlap" multiplications and the other operations [and on most
+# platforms even minimize the amount of the other operations, in
+# particular references to memory]. But it's possible to improve this
+# module performance by implementing dedicated squaring code-path and
+# possibly by unrolling loops...
+
+# January 2009.
+#
+# Reschedule to minimize/avoid Address Generation Interlock hazard,
+# make inner loops counter-based.
+
+# November 2010.
+#
+# Adapt for -m31 build. If kernel supports what's called "highgprs"
+# feature on Linux [see /proc/cpuinfo], it's possible to use 64-bit
+# instructions and achieve "64-bit" performance even in 31-bit legacy
+# application context. The feature is not specific to any particular
+# processor, as long as it's "z-CPU". Latter implies that the code
+# remains z/Architecture specific. Compatibility with 32-bit BN_ULONG
+# is achieved by swapping words after 64-bit loads, follow _dswap-s.
+# On z990 it was measured to perform 2.6-2.2 times better than
+# compiler-generated code, less for longer keys...
+
+$flavour = shift;
+
+if ($flavour =~ /3[12]/) {
+ $SIZE_T=4;
+ $g="";
+} else {
+ $SIZE_T=8;
+ $g="g";
+}
+
+while (($output=shift) && ($output!~/\w[\w\-]*\.\w+$/)) {}
+open STDOUT,">$output";
+
+$stdframe=16*$SIZE_T+4*8;
+
+$mn0="%r0";
+$num="%r1";
+
+# int bn_mul_mont(
+$rp="%r2"; # BN_ULONG *rp,
+$ap="%r3"; # const BN_ULONG *ap,
+$bp="%r4"; # const BN_ULONG *bp,
+$np="%r5"; # const BN_ULONG *np,
+$n0="%r6"; # const BN_ULONG *n0,
+#$num="160(%r15)" # int num);
+
+$bi="%r2"; # zaps rp
+$j="%r7";
+
+$ahi="%r8";
+$alo="%r9";
+$nhi="%r10";
+$nlo="%r11";
+$AHI="%r12";
+$NHI="%r13";
+$count="%r14";
+$sp="%r15";
+
+$code.=<<___;
+.text
+.globl bn_mul_mont
+.type bn_mul_mont,\@function
+bn_mul_mont:
+ lgf $num,`$stdframe+$SIZE_T-4`($sp) # pull $num
+ sla $num,`log($SIZE_T)/log(2)` # $num to enumerate bytes
+ la $bp,0($num,$bp)
+
+ st${g} %r2,2*$SIZE_T($sp)
+
+ cghi $num,16 #
+ lghi %r2,0 #
+ blr %r14 # if($num<16) return 0;
+___
+$code.=<<___ if ($flavour =~ /3[12]/);
+ tmll $num,4
+ bnzr %r14 # if ($num&1) return 0;
+___
+$code.=<<___ if ($flavour !~ /3[12]/);
+ cghi $num,96 #
+ bhr %r14 # if($num>96) return 0;
+___
+$code.=<<___;
+ stm${g} %r3,%r15,3*$SIZE_T($sp)
+
+ lghi $rp,-$stdframe-8 # leave room for carry bit
+ lcgr $j,$num # -$num
+ lgr %r0,$sp
+ la $rp,0($rp,$sp)
+ la $sp,0($j,$rp) # alloca
+ st${g} %r0,0($sp) # back chain
+
+ sra $num,3 # restore $num
+ la $bp,0($j,$bp) # restore $bp
+ ahi $num,-1 # adjust $num for inner loop
+ lg $n0,0($n0) # pull n0
+ _dswap $n0
+
+ lg $bi,0($bp)
+ _dswap $bi
+ lg $alo,0($ap)
+ _dswap $alo
+ mlgr $ahi,$bi # ap[0]*bp[0]
+ lgr $AHI,$ahi
+
+ lgr $mn0,$alo # "tp[0]"*n0
+ msgr $mn0,$n0
+
+ lg $nlo,0($np) #
+ _dswap $nlo
+ mlgr $nhi,$mn0 # np[0]*m1
+ algr $nlo,$alo # +="tp[0]"
+ lghi $NHI,0
+ alcgr $NHI,$nhi
+
+ la $j,8 # j=1
+ lr $count,$num
+
+.align 16
+.L1st:
+ lg $alo,0($j,$ap)
+ _dswap $alo
+ mlgr $ahi,$bi # ap[j]*bp[0]
+ algr $alo,$AHI
+ lghi $AHI,0
+ alcgr $AHI,$ahi
+
+ lg $nlo,0($j,$np)
+ _dswap $nlo
+ mlgr $nhi,$mn0 # np[j]*m1
+ algr $nlo,$NHI
+ lghi $NHI,0
+ alcgr $nhi,$NHI # +="tp[j]"
+ algr $nlo,$alo
+ alcgr $NHI,$nhi
+
+ stg $nlo,$stdframe-8($j,$sp) # tp[j-1]=
+ la $j,8($j) # j++
+ brct $count,.L1st
+
+ algr $NHI,$AHI
+ lghi $AHI,0
+ alcgr $AHI,$AHI # upmost overflow bit
+ stg $NHI,$stdframe-8($j,$sp)
+ stg $AHI,$stdframe($j,$sp)
+ la $bp,8($bp) # bp++
+
+.Louter:
+ lg $bi,0($bp) # bp[i]
+ _dswap $bi
+ lg $alo,0($ap)
+ _dswap $alo
+ mlgr $ahi,$bi # ap[0]*bp[i]
+ alg $alo,$stdframe($sp) # +=tp[0]
+ lghi $AHI,0
+ alcgr $AHI,$ahi
+
+ lgr $mn0,$alo
+ msgr $mn0,$n0 # tp[0]*n0
+
+ lg $nlo,0($np) # np[0]
+ _dswap $nlo
+ mlgr $nhi,$mn0 # np[0]*m1
+ algr $nlo,$alo # +="tp[0]"
+ lghi $NHI,0
+ alcgr $NHI,$nhi
+
+ la $j,8 # j=1
+ lr $count,$num
+
+.align 16
+.Linner:
+ lg $alo,0($j,$ap)
+ _dswap $alo
+ mlgr $ahi,$bi # ap[j]*bp[i]
+ algr $alo,$AHI
+ lghi $AHI,0
+ alcgr $ahi,$AHI
+ alg $alo,$stdframe($j,$sp)# +=tp[j]
+ alcgr $AHI,$ahi
+
+ lg $nlo,0($j,$np)
+ _dswap $nlo
+ mlgr $nhi,$mn0 # np[j]*m1
+ algr $nlo,$NHI
+ lghi $NHI,0
+ alcgr $nhi,$NHI
+ algr $nlo,$alo # +="tp[j]"
+ alcgr $NHI,$nhi
+
+ stg $nlo,$stdframe-8($j,$sp) # tp[j-1]=
+ la $j,8($j) # j++
+ brct $count,.Linner
+
+ algr $NHI,$AHI
+ lghi $AHI,0
+ alcgr $AHI,$AHI
+ alg $NHI,$stdframe($j,$sp)# accumulate previous upmost overflow bit
+ lghi $ahi,0
+ alcgr $AHI,$ahi # new upmost overflow bit
+ stg $NHI,$stdframe-8($j,$sp)
+ stg $AHI,$stdframe($j,$sp)
+
+ la $bp,8($bp) # bp++
+ cl${g} $bp,`$stdframe+8+4*$SIZE_T`($j,$sp) # compare to &bp[num]
+ jne .Louter
+
+ l${g} $rp,`$stdframe+8+2*$SIZE_T`($j,$sp) # reincarnate rp
+ la $ap,$stdframe($sp)
+ ahi $num,1 # restore $num, incidentally clears "borrow"
+
+ la $j,0
+ lr $count,$num
+.Lsub: lg $alo,0($j,$ap)
+ lg $nlo,0($j,$np)
+ _dswap $nlo
+ slbgr $alo,$nlo
+ stg $alo,0($j,$rp)
+ la $j,8($j)
+ brct $count,.Lsub
+ lghi $ahi,0
+ slbgr $AHI,$ahi # handle upmost carry
+ lghi $NHI,-1
+ xgr $NHI,$AHI
+
+ la $j,0
+ lgr $count,$num
+.Lcopy: lg $ahi,$stdframe($j,$sp) # conditional copy
+ lg $alo,0($j,$rp)
+ ngr $ahi,$AHI
+ ngr $alo,$NHI
+ ogr $alo,$ahi
+ _dswap $alo
+ stg $j,$stdframe($j,$sp) # zap tp
+ stg $alo,0($j,$rp)
+ la $j,8($j)
+ brct $count,.Lcopy
+
+ la %r1,`$stdframe+8+6*$SIZE_T`($j,$sp)
+ lm${g} %r6,%r15,0(%r1)
+ lghi %r2,1 # signal "processed"
+ br %r14
+.size bn_mul_mont,.-bn_mul_mont
+.string "Montgomery Multiplication for s390x, CRYPTOGAMS by <appro\@openssl.org>"
+___
+
+foreach (split("\n",$code)) {
+ s/\`([^\`]*)\`/eval $1/ge;
+ s/_dswap\s+(%r[0-9]+)/sprintf("rllg\t%s,%s,32",$1,$1) if($SIZE_T==4)/e;
+ print $_,"\n";
+}
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/s390x.S b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/s390x.S
new file mode 100644
index 0000000..b666c41
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/s390x.S
@@ -0,0 +1,713 @@
+.ident "s390x.S, version 1.1"
+// ====================================================================
+// Copyright 2007-2016 The OpenSSL Project Authors. All Rights Reserved.
+//
+// Licensed under the OpenSSL license (the "License"). You may not use
+// this file except in compliance with the License. You can obtain a copy
+// in the file LICENSE in the source distribution or at
+// https://www.openssl.org/source/license.html
+// ====================================================================
+
+.text
+
+#define zero %r0
+
+// BN_ULONG bn_mul_add_words(BN_ULONG *r2,BN_ULONG *r3,int r4,BN_ULONG r5);
+.globl bn_mul_add_words
+.type bn_mul_add_words,@function
+.align 4
+bn_mul_add_words:
+ lghi zero,0 // zero = 0
+ la %r1,0(%r2) // put rp aside [to give way to]
+ lghi %r2,0 // return value
+ ltgfr %r4,%r4
+ bler %r14 // if (len<=0) return 0;
+
+ stmg %r6,%r13,48(%r15)
+ lghi %r2,3
+ lghi %r12,0 // carry = 0
+ slgr %r1,%r3 // rp-=ap
+ nr %r2,%r4 // len%4
+ sra %r4,2 // cnt=len/4
+ jz .Loop1_madd // carry is incidentally cleared if branch taken
+ algr zero,zero // clear carry
+
+ lg %r7,0(%r3) // ap[0]
+ lg %r9,8(%r3) // ap[1]
+ mlgr %r6,%r5 // *=w
+ brct %r4,.Loop4_madd
+ j .Loop4_madd_tail
+
+.Loop4_madd:
+ mlgr %r8,%r5
+ lg %r11,16(%r3) // ap[i+2]
+ alcgr %r7,%r12 // +=carry
+ alcgr %r6,zero
+ alg %r7,0(%r3,%r1) // +=rp[i]
+ stg %r7,0(%r3,%r1) // rp[i]=
+
+ mlgr %r10,%r5
+ lg %r13,24(%r3)
+ alcgr %r9,%r6
+ alcgr %r8,zero
+ alg %r9,8(%r3,%r1)
+ stg %r9,8(%r3,%r1)
+
+ mlgr %r12,%r5
+ lg %r7,32(%r3)
+ alcgr %r11,%r8
+ alcgr %r10,zero
+ alg %r11,16(%r3,%r1)
+ stg %r11,16(%r3,%r1)
+
+ mlgr %r6,%r5
+ lg %r9,40(%r3)
+ alcgr %r13,%r10
+ alcgr %r12,zero
+ alg %r13,24(%r3,%r1)
+ stg %r13,24(%r3,%r1)
+
+ la %r3,32(%r3) // i+=4
+ brct %r4,.Loop4_madd
+
+.Loop4_madd_tail:
+ mlgr %r8,%r5
+ lg %r11,16(%r3)
+ alcgr %r7,%r12 // +=carry
+ alcgr %r6,zero
+ alg %r7,0(%r3,%r1) // +=rp[i]
+ stg %r7,0(%r3,%r1) // rp[i]=
+
+ mlgr %r10,%r5
+ lg %r13,24(%r3)
+ alcgr %r9,%r6
+ alcgr %r8,zero
+ alg %r9,8(%r3,%r1)
+ stg %r9,8(%r3,%r1)
+
+ mlgr %r12,%r5
+ alcgr %r11,%r8
+ alcgr %r10,zero
+ alg %r11,16(%r3,%r1)
+ stg %r11,16(%r3,%r1)
+
+ alcgr %r13,%r10
+ alcgr %r12,zero
+ alg %r13,24(%r3,%r1)
+ stg %r13,24(%r3,%r1)
+
+ la %r3,32(%r3) // i+=4
+
+ la %r2,1(%r2) // see if len%4 is zero ...
+ brct %r2,.Loop1_madd // without touching condition code:-)
+
+.Lend_madd:
+ lgr %r2,zero // return value
+ alcgr %r2,%r12 // collect even carry bit
+ lmg %r6,%r13,48(%r15)
+ br %r14
+
+.Loop1_madd:
+ lg %r7,0(%r3) // ap[i]
+ mlgr %r6,%r5 // *=w
+ alcgr %r7,%r12 // +=carry
+ alcgr %r6,zero
+ alg %r7,0(%r3,%r1) // +=rp[i]
+ stg %r7,0(%r3,%r1) // rp[i]=
+
+ lgr %r12,%r6
+ la %r3,8(%r3) // i++
+ brct %r2,.Loop1_madd
+
+ j .Lend_madd
+.size bn_mul_add_words,.-bn_mul_add_words
+
+// BN_ULONG bn_mul_words(BN_ULONG *r2,BN_ULONG *r3,int r4,BN_ULONG r5);
+.globl bn_mul_words
+.type bn_mul_words,@function
+.align 4
+bn_mul_words:
+ lghi zero,0 // zero = 0
+ la %r1,0(%r2) // put rp aside
+ lghi %r2,0 // i=0;
+ ltgfr %r4,%r4
+ bler %r14 // if (len<=0) return 0;
+
+ stmg %r6,%r10,48(%r15)
+ lghi %r10,3
+ lghi %r8,0 // carry = 0
+ nr %r10,%r4 // len%4
+ sra %r4,2 // cnt=len/4
+ jz .Loop1_mul // carry is incidentally cleared if branch taken
+ algr zero,zero // clear carry
+
+.Loop4_mul:
+ lg %r7,0(%r2,%r3) // ap[i]
+ mlgr %r6,%r5 // *=w
+ alcgr %r7,%r8 // +=carry
+ stg %r7,0(%r2,%r1) // rp[i]=
+
+ lg %r9,8(%r2,%r3)
+ mlgr %r8,%r5
+ alcgr %r9,%r6
+ stg %r9,8(%r2,%r1)
+
+ lg %r7,16(%r2,%r3)
+ mlgr %r6,%r5
+ alcgr %r7,%r8
+ stg %r7,16(%r2,%r1)
+
+ lg %r9,24(%r2,%r3)
+ mlgr %r8,%r5
+ alcgr %r9,%r6
+ stg %r9,24(%r2,%r1)
+
+ la %r2,32(%r2) // i+=4
+ brct %r4,.Loop4_mul
+
+ la %r10,1(%r10) // see if len%4 is zero ...
+ brct %r10,.Loop1_mul // without touching condition code:-)
+
+.Lend_mul:
+ alcgr %r8,zero // collect carry bit
+ lgr %r2,%r8
+ lmg %r6,%r10,48(%r15)
+ br %r14
+
+.Loop1_mul:
+ lg %r7,0(%r2,%r3) // ap[i]
+ mlgr %r6,%r5 // *=w
+ alcgr %r7,%r8 // +=carry
+ stg %r7,0(%r2,%r1) // rp[i]=
+
+ lgr %r8,%r6
+ la %r2,8(%r2) // i++
+ brct %r10,.Loop1_mul
+
+ j .Lend_mul
+.size bn_mul_words,.-bn_mul_words
+
+// void bn_sqr_words(BN_ULONG *r2,BN_ULONG *r2,int r4)
+.globl bn_sqr_words
+.type bn_sqr_words,@function
+.align 4
+bn_sqr_words:
+ ltgfr %r4,%r4
+ bler %r14
+
+ stmg %r6,%r7,48(%r15)
+ srag %r1,%r4,2 // cnt=len/4
+ jz .Loop1_sqr
+
+.Loop4_sqr:
+ lg %r7,0(%r3)
+ mlgr %r6,%r7
+ stg %r7,0(%r2)
+ stg %r6,8(%r2)
+
+ lg %r7,8(%r3)
+ mlgr %r6,%r7
+ stg %r7,16(%r2)
+ stg %r6,24(%r2)
+
+ lg %r7,16(%r3)
+ mlgr %r6,%r7
+ stg %r7,32(%r2)
+ stg %r6,40(%r2)
+
+ lg %r7,24(%r3)
+ mlgr %r6,%r7
+ stg %r7,48(%r2)
+ stg %r6,56(%r2)
+
+ la %r3,32(%r3)
+ la %r2,64(%r2)
+ brct %r1,.Loop4_sqr
+
+ lghi %r1,3
+ nr %r4,%r1 // cnt=len%4
+ jz .Lend_sqr
+
+.Loop1_sqr:
+ lg %r7,0(%r3)
+ mlgr %r6,%r7
+ stg %r7,0(%r2)
+ stg %r6,8(%r2)
+
+ la %r3,8(%r3)
+ la %r2,16(%r2)
+ brct %r4,.Loop1_sqr
+
+.Lend_sqr:
+ lmg %r6,%r7,48(%r15)
+ br %r14
+.size bn_sqr_words,.-bn_sqr_words
+
+// BN_ULONG bn_div_words(BN_ULONG h,BN_ULONG l,BN_ULONG d);
+.globl bn_div_words
+.type bn_div_words,@function
+.align 4
+bn_div_words:
+ dlgr %r2,%r4
+ lgr %r2,%r3
+ br %r14
+.size bn_div_words,.-bn_div_words
+
+// BN_ULONG bn_add_words(BN_ULONG *r2,BN_ULONG *r3,BN_ULONG *r4,int r5);
+.globl bn_add_words
+.type bn_add_words,@function
+.align 4
+bn_add_words:
+ la %r1,0(%r2) // put rp aside
+ lghi %r2,0 // i=0
+ ltgfr %r5,%r5
+ bler %r14 // if (len<=0) return 0;
+
+ stg %r6,48(%r15)
+ lghi %r6,3
+ nr %r6,%r5 // len%4
+ sra %r5,2 // len/4, use sra because it sets condition code
+ jz .Loop1_add // carry is incidentally cleared if branch taken
+ algr %r2,%r2 // clear carry
+
+.Loop4_add:
+ lg %r0,0(%r2,%r3)
+ alcg %r0,0(%r2,%r4)
+ stg %r0,0(%r2,%r1)
+ lg %r0,8(%r2,%r3)
+ alcg %r0,8(%r2,%r4)
+ stg %r0,8(%r2,%r1)
+ lg %r0,16(%r2,%r3)
+ alcg %r0,16(%r2,%r4)
+ stg %r0,16(%r2,%r1)
+ lg %r0,24(%r2,%r3)
+ alcg %r0,24(%r2,%r4)
+ stg %r0,24(%r2,%r1)
+
+ la %r2,32(%r2) // i+=4
+ brct %r5,.Loop4_add
+
+ la %r6,1(%r6) // see if len%4 is zero ...
+ brct %r6,.Loop1_add // without touching condition code:-)
+
+.Lexit_add:
+ lghi %r2,0
+ alcgr %r2,%r2
+ lg %r6,48(%r15)
+ br %r14
+
+.Loop1_add:
+ lg %r0,0(%r2,%r3)
+ alcg %r0,0(%r2,%r4)
+ stg %r0,0(%r2,%r1)
+
+ la %r2,8(%r2) // i++
+ brct %r6,.Loop1_add
+
+ j .Lexit_add
+.size bn_add_words,.-bn_add_words
+
+// BN_ULONG bn_sub_words(BN_ULONG *r2,BN_ULONG *r3,BN_ULONG *r4,int r5);
+.globl bn_sub_words
+.type bn_sub_words,@function
+.align 4
+bn_sub_words:
+ la %r1,0(%r2) // put rp aside
+ lghi %r2,0 // i=0
+ ltgfr %r5,%r5
+ bler %r14 // if (len<=0) return 0;
+
+ stg %r6,48(%r15)
+ lghi %r6,3
+ nr %r6,%r5 // len%4
+ sra %r5,2 // len/4, use sra because it sets condition code
+ jnz .Loop4_sub // borrow is incidentally cleared if branch taken
+ slgr %r2,%r2 // clear borrow
+
+.Loop1_sub:
+ lg %r0,0(%r2,%r3)
+ slbg %r0,0(%r2,%r4)
+ stg %r0,0(%r2,%r1)
+
+ la %r2,8(%r2) // i++
+ brct %r6,.Loop1_sub
+ j .Lexit_sub
+
+.Loop4_sub:
+ lg %r0,0(%r2,%r3)
+ slbg %r0,0(%r2,%r4)
+ stg %r0,0(%r2,%r1)
+ lg %r0,8(%r2,%r3)
+ slbg %r0,8(%r2,%r4)
+ stg %r0,8(%r2,%r1)
+ lg %r0,16(%r2,%r3)
+ slbg %r0,16(%r2,%r4)
+ stg %r0,16(%r2,%r1)
+ lg %r0,24(%r2,%r3)
+ slbg %r0,24(%r2,%r4)
+ stg %r0,24(%r2,%r1)
+
+ la %r2,32(%r2) // i+=4
+ brct %r5,.Loop4_sub
+
+ la %r6,1(%r6) // see if len%4 is zero ...
+ brct %r6,.Loop1_sub // without touching condition code:-)
+
+.Lexit_sub:
+ lghi %r2,0
+ slbgr %r2,%r2
+ lcgr %r2,%r2
+ lg %r6,48(%r15)
+ br %r14
+.size bn_sub_words,.-bn_sub_words
+
+#define c1 %r1
+#define c2 %r5
+#define c3 %r8
+
+#define mul_add_c(ai,bi,c1,c2,c3) \
+ lg %r7,ai*8(%r3); \
+ mlg %r6,bi*8(%r4); \
+ algr c1,%r7; \
+ alcgr c2,%r6; \
+ alcgr c3,zero
+
+// void bn_mul_comba8(BN_ULONG *r2,BN_ULONG *r3,BN_ULONG *r4);
+.globl bn_mul_comba8
+.type bn_mul_comba8,@function
+.align 4
+bn_mul_comba8:
+ stmg %r6,%r8,48(%r15)
+
+ lghi c1,0
+ lghi c2,0
+ lghi c3,0
+ lghi zero,0
+
+ mul_add_c(0,0,c1,c2,c3);
+ stg c1,0*8(%r2)
+ lghi c1,0
+
+ mul_add_c(0,1,c2,c3,c1);
+ mul_add_c(1,0,c2,c3,c1);
+ stg c2,1*8(%r2)
+ lghi c2,0
+
+ mul_add_c(2,0,c3,c1,c2);
+ mul_add_c(1,1,c3,c1,c2);
+ mul_add_c(0,2,c3,c1,c2);
+ stg c3,2*8(%r2)
+ lghi c3,0
+
+ mul_add_c(0,3,c1,c2,c3);
+ mul_add_c(1,2,c1,c2,c3);
+ mul_add_c(2,1,c1,c2,c3);
+ mul_add_c(3,0,c1,c2,c3);
+ stg c1,3*8(%r2)
+ lghi c1,0
+
+ mul_add_c(4,0,c2,c3,c1);
+ mul_add_c(3,1,c2,c3,c1);
+ mul_add_c(2,2,c2,c3,c1);
+ mul_add_c(1,3,c2,c3,c1);
+ mul_add_c(0,4,c2,c3,c1);
+ stg c2,4*8(%r2)
+ lghi c2,0
+
+ mul_add_c(0,5,c3,c1,c2);
+ mul_add_c(1,4,c3,c1,c2);
+ mul_add_c(2,3,c3,c1,c2);
+ mul_add_c(3,2,c3,c1,c2);
+ mul_add_c(4,1,c3,c1,c2);
+ mul_add_c(5,0,c3,c1,c2);
+ stg c3,5*8(%r2)
+ lghi c3,0
+
+ mul_add_c(6,0,c1,c2,c3);
+ mul_add_c(5,1,c1,c2,c3);
+ mul_add_c(4,2,c1,c2,c3);
+ mul_add_c(3,3,c1,c2,c3);
+ mul_add_c(2,4,c1,c2,c3);
+ mul_add_c(1,5,c1,c2,c3);
+ mul_add_c(0,6,c1,c2,c3);
+ stg c1,6*8(%r2)
+ lghi c1,0
+
+ mul_add_c(0,7,c2,c3,c1);
+ mul_add_c(1,6,c2,c3,c1);
+ mul_add_c(2,5,c2,c3,c1);
+ mul_add_c(3,4,c2,c3,c1);
+ mul_add_c(4,3,c2,c3,c1);
+ mul_add_c(5,2,c2,c3,c1);
+ mul_add_c(6,1,c2,c3,c1);
+ mul_add_c(7,0,c2,c3,c1);
+ stg c2,7*8(%r2)
+ lghi c2,0
+
+ mul_add_c(7,1,c3,c1,c2);
+ mul_add_c(6,2,c3,c1,c2);
+ mul_add_c(5,3,c3,c1,c2);
+ mul_add_c(4,4,c3,c1,c2);
+ mul_add_c(3,5,c3,c1,c2);
+ mul_add_c(2,6,c3,c1,c2);
+ mul_add_c(1,7,c3,c1,c2);
+ stg c3,8*8(%r2)
+ lghi c3,0
+
+ mul_add_c(2,7,c1,c2,c3);
+ mul_add_c(3,6,c1,c2,c3);
+ mul_add_c(4,5,c1,c2,c3);
+ mul_add_c(5,4,c1,c2,c3);
+ mul_add_c(6,3,c1,c2,c3);
+ mul_add_c(7,2,c1,c2,c3);
+ stg c1,9*8(%r2)
+ lghi c1,0
+
+ mul_add_c(7,3,c2,c3,c1);
+ mul_add_c(6,4,c2,c3,c1);
+ mul_add_c(5,5,c2,c3,c1);
+ mul_add_c(4,6,c2,c3,c1);
+ mul_add_c(3,7,c2,c3,c1);
+ stg c2,10*8(%r2)
+ lghi c2,0
+
+ mul_add_c(4,7,c3,c1,c2);
+ mul_add_c(5,6,c3,c1,c2);
+ mul_add_c(6,5,c3,c1,c2);
+ mul_add_c(7,4,c3,c1,c2);
+ stg c3,11*8(%r2)
+ lghi c3,0
+
+ mul_add_c(7,5,c1,c2,c3);
+ mul_add_c(6,6,c1,c2,c3);
+ mul_add_c(5,7,c1,c2,c3);
+ stg c1,12*8(%r2)
+ lghi c1,0
+
+
+ mul_add_c(6,7,c2,c3,c1);
+ mul_add_c(7,6,c2,c3,c1);
+ stg c2,13*8(%r2)
+ lghi c2,0
+
+ mul_add_c(7,7,c3,c1,c2);
+ stg c3,14*8(%r2)
+ stg c1,15*8(%r2)
+
+ lmg %r6,%r8,48(%r15)
+ br %r14
+.size bn_mul_comba8,.-bn_mul_comba8
+
+// void bn_mul_comba4(BN_ULONG *r2,BN_ULONG *r3,BN_ULONG *r4);
+.globl bn_mul_comba4
+.type bn_mul_comba4,@function
+.align 4
+bn_mul_comba4:
+ stmg %r6,%r8,48(%r15)
+
+ lghi c1,0
+ lghi c2,0
+ lghi c3,0
+ lghi zero,0
+
+ mul_add_c(0,0,c1,c2,c3);
+ stg c1,0*8(%r2)
+ lghi c1,0
+
+ mul_add_c(0,1,c2,c3,c1);
+ mul_add_c(1,0,c2,c3,c1);
+ stg c2,1*8(%r2)
+ lghi c2,0
+
+ mul_add_c(2,0,c3,c1,c2);
+ mul_add_c(1,1,c3,c1,c2);
+ mul_add_c(0,2,c3,c1,c2);
+ stg c3,2*8(%r2)
+ lghi c3,0
+
+ mul_add_c(0,3,c1,c2,c3);
+ mul_add_c(1,2,c1,c2,c3);
+ mul_add_c(2,1,c1,c2,c3);
+ mul_add_c(3,0,c1,c2,c3);
+ stg c1,3*8(%r2)
+ lghi c1,0
+
+ mul_add_c(3,1,c2,c3,c1);
+ mul_add_c(2,2,c2,c3,c1);
+ mul_add_c(1,3,c2,c3,c1);
+ stg c2,4*8(%r2)
+ lghi c2,0
+
+ mul_add_c(2,3,c3,c1,c2);
+ mul_add_c(3,2,c3,c1,c2);
+ stg c3,5*8(%r2)
+ lghi c3,0
+
+ mul_add_c(3,3,c1,c2,c3);
+ stg c1,6*8(%r2)
+ stg c2,7*8(%r2)
+
+ stmg %r6,%r8,48(%r15)
+ br %r14
+.size bn_mul_comba4,.-bn_mul_comba4
+
+#define sqr_add_c(ai,c1,c2,c3) \
+ lg %r7,ai*8(%r3); \
+ mlgr %r6,%r7; \
+ algr c1,%r7; \
+ alcgr c2,%r6; \
+ alcgr c3,zero
+
+#define sqr_add_c2(ai,aj,c1,c2,c3) \
+ lg %r7,ai*8(%r3); \
+ mlg %r6,aj*8(%r3); \
+ algr c1,%r7; \
+ alcgr c2,%r6; \
+ alcgr c3,zero; \
+ algr c1,%r7; \
+ alcgr c2,%r6; \
+ alcgr c3,zero
+
+// void bn_sqr_comba8(BN_ULONG *r2,BN_ULONG *r3);
+.globl bn_sqr_comba8
+.type bn_sqr_comba8,@function
+.align 4
+bn_sqr_comba8:
+ stmg %r6,%r8,48(%r15)
+
+ lghi c1,0
+ lghi c2,0
+ lghi c3,0
+ lghi zero,0
+
+ sqr_add_c(0,c1,c2,c3);
+ stg c1,0*8(%r2)
+ lghi c1,0
+
+ sqr_add_c2(1,0,c2,c3,c1);
+ stg c2,1*8(%r2)
+ lghi c2,0
+
+ sqr_add_c(1,c3,c1,c2);
+ sqr_add_c2(2,0,c3,c1,c2);
+ stg c3,2*8(%r2)
+ lghi c3,0
+
+ sqr_add_c2(3,0,c1,c2,c3);
+ sqr_add_c2(2,1,c1,c2,c3);
+ stg c1,3*8(%r2)
+ lghi c1,0
+
+ sqr_add_c(2,c2,c3,c1);
+ sqr_add_c2(3,1,c2,c3,c1);
+ sqr_add_c2(4,0,c2,c3,c1);
+ stg c2,4*8(%r2)
+ lghi c2,0
+
+ sqr_add_c2(5,0,c3,c1,c2);
+ sqr_add_c2(4,1,c3,c1,c2);
+ sqr_add_c2(3,2,c3,c1,c2);
+ stg c3,5*8(%r2)
+ lghi c3,0
+
+ sqr_add_c(3,c1,c2,c3);
+ sqr_add_c2(4,2,c1,c2,c3);
+ sqr_add_c2(5,1,c1,c2,c3);
+ sqr_add_c2(6,0,c1,c2,c3);
+ stg c1,6*8(%r2)
+ lghi c1,0
+
+ sqr_add_c2(7,0,c2,c3,c1);
+ sqr_add_c2(6,1,c2,c3,c1);
+ sqr_add_c2(5,2,c2,c3,c1);
+ sqr_add_c2(4,3,c2,c3,c1);
+ stg c2,7*8(%r2)
+ lghi c2,0
+
+ sqr_add_c(4,c3,c1,c2);
+ sqr_add_c2(5,3,c3,c1,c2);
+ sqr_add_c2(6,2,c3,c1,c2);
+ sqr_add_c2(7,1,c3,c1,c2);
+ stg c3,8*8(%r2)
+ lghi c3,0
+
+ sqr_add_c2(7,2,c1,c2,c3);
+ sqr_add_c2(6,3,c1,c2,c3);
+ sqr_add_c2(5,4,c1,c2,c3);
+ stg c1,9*8(%r2)
+ lghi c1,0
+
+ sqr_add_c(5,c2,c3,c1);
+ sqr_add_c2(6,4,c2,c3,c1);
+ sqr_add_c2(7,3,c2,c3,c1);
+ stg c2,10*8(%r2)
+ lghi c2,0
+
+ sqr_add_c2(7,4,c3,c1,c2);
+ sqr_add_c2(6,5,c3,c1,c2);
+ stg c3,11*8(%r2)
+ lghi c3,0
+
+ sqr_add_c(6,c1,c2,c3);
+ sqr_add_c2(7,5,c1,c2,c3);
+ stg c1,12*8(%r2)
+ lghi c1,0
+
+ sqr_add_c2(7,6,c2,c3,c1);
+ stg c2,13*8(%r2)
+ lghi c2,0
+
+ sqr_add_c(7,c3,c1,c2);
+ stg c3,14*8(%r2)
+ stg c1,15*8(%r2)
+
+ lmg %r6,%r8,48(%r15)
+ br %r14
+.size bn_sqr_comba8,.-bn_sqr_comba8
+
+// void bn_sqr_comba4(BN_ULONG *r2,BN_ULONG *r3);
+.globl bn_sqr_comba4
+.type bn_sqr_comba4,@function
+.align 4
+bn_sqr_comba4:
+ stmg %r6,%r8,48(%r15)
+
+ lghi c1,0
+ lghi c2,0
+ lghi c3,0
+ lghi zero,0
+
+ sqr_add_c(0,c1,c2,c3);
+ stg c1,0*8(%r2)
+ lghi c1,0
+
+ sqr_add_c2(1,0,c2,c3,c1);
+ stg c2,1*8(%r2)
+ lghi c2,0
+
+ sqr_add_c(1,c3,c1,c2);
+ sqr_add_c2(2,0,c3,c1,c2);
+ stg c3,2*8(%r2)
+ lghi c3,0
+
+ sqr_add_c2(3,0,c1,c2,c3);
+ sqr_add_c2(2,1,c1,c2,c3);
+ stg c1,3*8(%r2)
+ lghi c1,0
+
+ sqr_add_c(2,c2,c3,c1);
+ sqr_add_c2(3,1,c2,c3,c1);
+ stg c2,4*8(%r2)
+ lghi c2,0
+
+ sqr_add_c2(3,2,c3,c1,c2);
+ stg c3,5*8(%r2)
+ lghi c3,0
+
+ sqr_add_c(3,c1,c2,c3);
+ stg c1,6*8(%r2)
+ stg c2,7*8(%r2)
+
+ lmg %r6,%r8,48(%r15)
+ br %r14
+.size bn_sqr_comba4,.-bn_sqr_comba4
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/sparct4-mont.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/sparct4-mont.pl
new file mode 100755
index 0000000..62e297a
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/sparct4-mont.pl
@@ -0,0 +1,1228 @@
+#! /usr/bin/env perl
+# Copyright 2012-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+
+# ====================================================================
+# Written by David S. Miller and Andy Polyakov
+# The module is licensed under 2-clause BSD license.
+# November 2012. All rights reserved.
+# ====================================================================
+
+######################################################################
+# Montgomery squaring-n-multiplication module for SPARC T4.
+#
+# The module consists of three parts:
+#
+# 1) collection of "single-op" subroutines that perform single
+# operation, Montgomery squaring or multiplication, on 512-,
+# 1024-, 1536- and 2048-bit operands;
+# 2) collection of "multi-op" subroutines that perform 5 squaring and
+# 1 multiplication operations on operands of above lengths;
+# 3) fall-back and helper VIS3 subroutines.
+#
+# RSA sign is dominated by multi-op subroutine, while RSA verify and
+# DSA - by single-op. Special note about 4096-bit RSA verify result.
+# Operands are too long for dedicated hardware and it's handled by
+# VIS3 code, which is why you don't see any improvement. It's surely
+# possible to improve it [by deploying 'mpmul' instruction], maybe in
+# the future...
+#
+# Performance improvement.
+#
+# 64-bit process, VIS3:
+# sign verify sign/s verify/s
+# rsa 1024 bits 0.000628s 0.000028s 1592.4 35434.4
+# rsa 2048 bits 0.003282s 0.000106s 304.7 9438.3
+# rsa 4096 bits 0.025866s 0.000340s 38.7 2940.9
+# dsa 1024 bits 0.000301s 0.000332s 3323.7 3013.9
+# dsa 2048 bits 0.001056s 0.001233s 946.9 810.8
+#
+# 64-bit process, this module:
+# sign verify sign/s verify/s
+# rsa 1024 bits 0.000256s 0.000016s 3904.4 61411.9
+# rsa 2048 bits 0.000946s 0.000029s 1056.8 34292.7
+# rsa 4096 bits 0.005061s 0.000340s 197.6 2940.5
+# dsa 1024 bits 0.000176s 0.000195s 5674.7 5130.5
+# dsa 2048 bits 0.000296s 0.000354s 3383.2 2827.6
+#
+######################################################################
+# 32-bit process, VIS3:
+# sign verify sign/s verify/s
+# rsa 1024 bits 0.000665s 0.000028s 1504.8 35233.3
+# rsa 2048 bits 0.003349s 0.000106s 298.6 9433.4
+# rsa 4096 bits 0.025959s 0.000341s 38.5 2934.8
+# dsa 1024 bits 0.000320s 0.000341s 3123.3 2929.6
+# dsa 2048 bits 0.001101s 0.001260s 908.2 793.4
+#
+# 32-bit process, this module:
+# sign verify sign/s verify/s
+# rsa 1024 bits 0.000301s 0.000017s 3317.1 60240.0
+# rsa 2048 bits 0.001034s 0.000030s 966.9 33812.7
+# rsa 4096 bits 0.005244s 0.000341s 190.7 2935.4
+# dsa 1024 bits 0.000201s 0.000205s 4976.1 4879.2
+# dsa 2048 bits 0.000328s 0.000360s 3051.1 2774.2
+#
+# 32-bit code is prone to performance degradation as interrupt rate
+# dispatched to CPU executing the code grows. This is because in
+# standard process of handling interrupt in 32-bit process context
+# upper halves of most integer registers used as input or output are
+# zeroed. This renders result invalid, and operation has to be re-run.
+# If CPU is "bothered" with timer interrupts only, the penalty is
+# hardly measurable. But in order to mitigate this problem for higher
+# interrupt rates contemporary Linux kernel recognizes biased stack
+# even in 32-bit process context and preserves full register contents.
+# See http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=517ffce4e1a03aea979fe3a18a3dd1761a24fafb
+# for details.
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+push(@INC,"${dir}","${dir}../../perlasm");
+require "sparcv9_modes.pl";
+
+$output = pop;
+open STDOUT,">$output";
+
+$code.=<<___;
+#include "sparc_arch.h"
+
+#ifdef __arch64__
+.register %g2,#scratch
+.register %g3,#scratch
+#endif
+
+.section ".text",#alloc,#execinstr
+
+#ifdef __PIC__
+SPARC_PIC_THUNK(%g1)
+#endif
+___
+
+########################################################################
+# Register layout for mont[mul|sqr] instructions.
+# For details see "Oracle SPARC Architecture 2011" manual at
+# http://www.oracle.com/technetwork/server-storage/sun-sparc-enterprise/documentation/.
+#
+my @R=map("%f".2*$_,(0..11,30,31,12..29));
+my @N=(map("%l$_",(0..7)),map("%o$_",(0..5))); @N=(@N,@N,@N[0..3]);
+my @A=(@N[0..13],@R[14..31]);
+my @B=(map("%i$_",(0..5)),map("%l$_",(0..7))); @B=(@B,@B,map("%o$_",(0..3)));
+
+########################################################################
+# int bn_mul_mont_t4_$NUM(u64 *rp,const u64 *ap,const u64 *bp,
+# const u64 *np,const BN_ULONG *n0);
+#
+sub generate_bn_mul_mont_t4() {
+my $NUM=shift;
+my ($rp,$ap,$bp,$np,$sentinel)=map("%g$_",(1..5));
+
+$code.=<<___;
+.globl bn_mul_mont_t4_$NUM
+.align 32
+bn_mul_mont_t4_$NUM:
+#ifdef __arch64__
+ mov 0,$sentinel
+ mov -128,%g4
+#elif defined(SPARCV9_64BIT_STACK)
+ SPARC_LOAD_ADDRESS_LEAF(OPENSSL_sparcv9cap_P,%g1,%g5)
+ ld [%g1+0],%g1 ! OPENSSL_sparcv9_P[0]
+ mov -2047,%g4
+ and %g1,SPARCV9_64BIT_STACK,%g1
+ movrz %g1,0,%g4
+ mov -1,$sentinel
+ add %g4,-128,%g4
+#else
+ mov -1,$sentinel
+ mov -128,%g4
+#endif
+ sllx $sentinel,32,$sentinel
+ save %sp,%g4,%sp
+#ifndef __arch64__
+ save %sp,-128,%sp ! warm it up
+ save %sp,-128,%sp
+ save %sp,-128,%sp
+ save %sp,-128,%sp
+ save %sp,-128,%sp
+ save %sp,-128,%sp
+ restore
+ restore
+ restore
+ restore
+ restore
+ restore
+#endif
+ and %sp,1,%g4
+ or $sentinel,%fp,%fp
+ or %g4,$sentinel,$sentinel
+
+ ! copy arguments to global registers
+ mov %i0,$rp
+ mov %i1,$ap
+ mov %i2,$bp
+ mov %i3,$np
+ ld [%i4+0],%f1 ! load *n0
+ ld [%i4+4],%f0
+ fsrc2 %f0,%f60
+___
+
+# load ap[$NUM] ########################################################
+$code.=<<___;
+ save %sp,-128,%sp; or $sentinel,%fp,%fp
+___
+for($i=0; $i<14 && $i<$NUM; $i++) {
+my $lo=$i<13?@A[$i+1]:"%o7";
+$code.=<<___;
+ ld [$ap+$i*8+0],$lo
+ ld [$ap+$i*8+4],@A[$i]
+ sllx @A[$i],32,@A[$i]
+ or $lo,@A[$i],@A[$i]
+___
+}
+for(; $i<$NUM; $i++) {
+my ($hi,$lo)=("%f".2*($i%4),"%f".(2*($i%4)+1));
+$code.=<<___;
+ ld [$ap+$i*8+0],$lo
+ ld [$ap+$i*8+4],$hi
+ fsrc2 $hi,@A[$i]
+___
+}
+# load np[$NUM] ########################################################
+$code.=<<___;
+ save %sp,-128,%sp; or $sentinel,%fp,%fp
+___
+for($i=0; $i<14 && $i<$NUM; $i++) {
+my $lo=$i<13?@N[$i+1]:"%o7";
+$code.=<<___;
+ ld [$np+$i*8+0],$lo
+ ld [$np+$i*8+4],@N[$i]
+ sllx @N[$i],32,@N[$i]
+ or $lo,@N[$i],@N[$i]
+___
+}
+$code.=<<___;
+ save %sp,-128,%sp; or $sentinel,%fp,%fp
+___
+for(; $i<28 && $i<$NUM; $i++) {
+my $lo=$i<27?@N[$i+1]:"%o7";
+$code.=<<___;
+ ld [$np+$i*8+0],$lo
+ ld [$np+$i*8+4],@N[$i]
+ sllx @N[$i],32,@N[$i]
+ or $lo,@N[$i],@N[$i]
+___
+}
+$code.=<<___;
+ save %sp,-128,%sp; or $sentinel,%fp,%fp
+___
+for(; $i<$NUM; $i++) {
+my $lo=($i<$NUM-1)?@N[$i+1]:"%o7";
+$code.=<<___;
+ ld [$np+$i*8+0],$lo
+ ld [$np+$i*8+4],@N[$i]
+ sllx @N[$i],32,@N[$i]
+ or $lo,@N[$i],@N[$i]
+___
+}
+$code.=<<___;
+ cmp $ap,$bp
+ be SIZE_T_CC,.Lmsquare_$NUM
+ nop
+___
+
+# load bp[$NUM] ########################################################
+$code.=<<___;
+ save %sp,-128,%sp; or $sentinel,%fp,%fp
+___
+for($i=0; $i<14 && $i<$NUM; $i++) {
+my $lo=$i<13?@B[$i+1]:"%o7";
+$code.=<<___;
+ ld [$bp+$i*8+0],$lo
+ ld [$bp+$i*8+4],@B[$i]
+ sllx @B[$i],32,@B[$i]
+ or $lo,@B[$i],@B[$i]
+___
+}
+$code.=<<___;
+ save %sp,-128,%sp; or $sentinel,%fp,%fp
+___
+for(; $i<$NUM; $i++) {
+my $lo=($i<$NUM-1)?@B[$i+1]:"%o7";
+$code.=<<___;
+ ld [$bp+$i*8+0],$lo
+ ld [$bp+$i*8+4],@B[$i]
+ sllx @B[$i],32,@B[$i]
+ or $lo,@B[$i],@B[$i]
+___
+}
+# magic ################################################################
+$code.=<<___;
+ .word 0x81b02920+$NUM-1 ! montmul $NUM-1
+.Lmresume_$NUM:
+ fbu,pn %fcc3,.Lmabort_$NUM
+#ifndef __arch64__
+ and %fp,$sentinel,$sentinel
+ brz,pn $sentinel,.Lmabort_$NUM
+#endif
+ nop
+#ifdef __arch64__
+ restore
+ restore
+ restore
+ restore
+ restore
+#else
+ restore; and %fp,$sentinel,$sentinel
+ restore; and %fp,$sentinel,$sentinel
+ restore; and %fp,$sentinel,$sentinel
+ restore; and %fp,$sentinel,$sentinel
+ brz,pn $sentinel,.Lmabort1_$NUM
+ restore
+#endif
+___
+
+# save tp[$NUM] ########################################################
+for($i=0; $i<14 && $i<$NUM; $i++) {
+$code.=<<___;
+ movxtod @A[$i],@R[$i]
+___
+}
+$code.=<<___;
+#ifdef __arch64__
+ restore
+#else
+ and %fp,$sentinel,$sentinel
+ restore
+ and $sentinel,1,%o7
+ and %fp,$sentinel,$sentinel
+ srl %fp,0,%fp ! just in case?
+ or %o7,$sentinel,$sentinel
+ brz,a,pn $sentinel,.Lmdone_$NUM
+ mov 0,%i0 ! return failure
+#endif
+___
+for($i=0; $i<12 && $i<$NUM; $i++) {
+@R[$i] =~ /%f([0-9]+)/;
+my $lo = "%f".($1+1);
+$code.=<<___;
+ st $lo,[$rp+$i*8+0]
+ st @R[$i],[$rp+$i*8+4]
+___
+}
+for(; $i<$NUM; $i++) {
+my ($hi,$lo)=("%f".2*($i%4),"%f".(2*($i%4)+1));
+$code.=<<___;
+ fsrc2 @R[$i],$hi
+ st $lo,[$rp+$i*8+0]
+ st $hi,[$rp+$i*8+4]
+___
+}
+$code.=<<___;
+ mov 1,%i0 ! return success
+.Lmdone_$NUM:
+ ret
+ restore
+
+.Lmabort_$NUM:
+ restore
+ restore
+ restore
+ restore
+ restore
+.Lmabort1_$NUM:
+ restore
+
+ mov 0,%i0 ! return failure
+ ret
+ restore
+
+.align 32
+.Lmsquare_$NUM:
+ save %sp,-128,%sp; or $sentinel,%fp,%fp
+ save %sp,-128,%sp; or $sentinel,%fp,%fp
+ .word 0x81b02940+$NUM-1 ! montsqr $NUM-1
+ ba .Lmresume_$NUM
+ nop
+.type bn_mul_mont_t4_$NUM, #function
+.size bn_mul_mont_t4_$NUM, .-bn_mul_mont_t4_$NUM
+___
+}
+
+for ($i=8;$i<=32;$i+=8) {
+ &generate_bn_mul_mont_t4($i);
+}
+
+########################################################################
+#
+sub load_ccr {
+my ($ptbl,$pwr,$ccr,$skip_wr)=@_;
+$code.=<<___;
+ srl $pwr, 2, %o4
+ and $pwr, 3, %o5
+ and %o4, 7, %o4
+ sll %o5, 3, %o5 ! offset within first cache line
+ add %o5, $ptbl, $ptbl ! of the pwrtbl
+ or %g0, 1, %o5
+ sll %o5, %o4, $ccr
+___
+$code.=<<___ if (!$skip_wr);
+ wr $ccr, %g0, %ccr
+___
+}
+sub load_b_pair {
+my ($pwrtbl,$B0,$B1)=@_;
+
+$code.=<<___;
+ ldx [$pwrtbl+0*32], $B0
+ ldx [$pwrtbl+8*32], $B1
+ ldx [$pwrtbl+1*32], %o4
+ ldx [$pwrtbl+9*32], %o5
+ movvs %icc, %o4, $B0
+ ldx [$pwrtbl+2*32], %o4
+ movvs %icc, %o5, $B1
+ ldx [$pwrtbl+10*32],%o5
+ move %icc, %o4, $B0
+ ldx [$pwrtbl+3*32], %o4
+ move %icc, %o5, $B1
+ ldx [$pwrtbl+11*32],%o5
+ movneg %icc, %o4, $B0
+ ldx [$pwrtbl+4*32], %o4
+ movneg %icc, %o5, $B1
+ ldx [$pwrtbl+12*32],%o5
+ movcs %xcc, %o4, $B0
+ ldx [$pwrtbl+5*32],%o4
+ movcs %xcc, %o5, $B1
+ ldx [$pwrtbl+13*32],%o5
+ movvs %xcc, %o4, $B0
+ ldx [$pwrtbl+6*32], %o4
+ movvs %xcc, %o5, $B1
+ ldx [$pwrtbl+14*32],%o5
+ move %xcc, %o4, $B0
+ ldx [$pwrtbl+7*32], %o4
+ move %xcc, %o5, $B1
+ ldx [$pwrtbl+15*32],%o5
+ movneg %xcc, %o4, $B0
+ add $pwrtbl,16*32, $pwrtbl
+ movneg %xcc, %o5, $B1
+___
+}
+sub load_b {
+my ($pwrtbl,$Bi)=@_;
+
+$code.=<<___;
+ ldx [$pwrtbl+0*32], $Bi
+ ldx [$pwrtbl+1*32], %o4
+ ldx [$pwrtbl+2*32], %o5
+ movvs %icc, %o4, $Bi
+ ldx [$pwrtbl+3*32], %o4
+ move %icc, %o5, $Bi
+ ldx [$pwrtbl+4*32], %o5
+ movneg %icc, %o4, $Bi
+ ldx [$pwrtbl+5*32], %o4
+ movcs %xcc, %o5, $Bi
+ ldx [$pwrtbl+6*32], %o5
+ movvs %xcc, %o4, $Bi
+ ldx [$pwrtbl+7*32], %o4
+ move %xcc, %o5, $Bi
+ add $pwrtbl,8*32, $pwrtbl
+ movneg %xcc, %o4, $Bi
+___
+}
+
+########################################################################
+# int bn_pwr5_mont_t4_$NUM(u64 *tp,const u64 *np,const BN_ULONG *n0,
+# const u64 *pwrtbl,int pwr,int stride);
+#
+sub generate_bn_pwr5_mont_t4() {
+my $NUM=shift;
+my ($tp,$np,$pwrtbl,$pwr,$sentinel)=map("%g$_",(1..5));
+
+$code.=<<___;
+.globl bn_pwr5_mont_t4_$NUM
+.align 32
+bn_pwr5_mont_t4_$NUM:
+#ifdef __arch64__
+ mov 0,$sentinel
+ mov -128,%g4
+#elif defined(SPARCV9_64BIT_STACK)
+ SPARC_LOAD_ADDRESS_LEAF(OPENSSL_sparcv9cap_P,%g1,%g5)
+ ld [%g1+0],%g1 ! OPENSSL_sparcv9_P[0]
+ mov -2047,%g4
+ and %g1,SPARCV9_64BIT_STACK,%g1
+ movrz %g1,0,%g4
+ mov -1,$sentinel
+ add %g4,-128,%g4
+#else
+ mov -1,$sentinel
+ mov -128,%g4
+#endif
+ sllx $sentinel,32,$sentinel
+ save %sp,%g4,%sp
+#ifndef __arch64__
+ save %sp,-128,%sp ! warm it up
+ save %sp,-128,%sp
+ save %sp,-128,%sp
+ save %sp,-128,%sp
+ save %sp,-128,%sp
+ save %sp,-128,%sp
+ restore
+ restore
+ restore
+ restore
+ restore
+ restore
+#endif
+ and %sp,1,%g4
+ or $sentinel,%fp,%fp
+ or %g4,$sentinel,$sentinel
+
+ ! copy arguments to global registers
+ mov %i0,$tp
+ mov %i1,$np
+ ld [%i2+0],%f1 ! load *n0
+ ld [%i2+4],%f0
+ mov %i3,$pwrtbl
+ srl %i4,%g0,%i4 ! pack last arguments
+ sllx %i5,32,$pwr
+ or %i4,$pwr,$pwr
+ fsrc2 %f0,%f60
+___
+
+# load tp[$NUM] ########################################################
+$code.=<<___;
+ save %sp,-128,%sp; or $sentinel,%fp,%fp
+___
+for($i=0; $i<14 && $i<$NUM; $i++) {
+$code.=<<___;
+ ldx [$tp+$i*8],@A[$i]
+___
+}
+for(; $i<$NUM; $i++) {
+$code.=<<___;
+ ldd [$tp+$i*8],@A[$i]
+___
+}
+# load np[$NUM] ########################################################
+$code.=<<___;
+ save %sp,-128,%sp; or $sentinel,%fp,%fp
+___
+for($i=0; $i<14 && $i<$NUM; $i++) {
+$code.=<<___;
+ ldx [$np+$i*8],@N[$i]
+___
+}
+$code.=<<___;
+ save %sp,-128,%sp; or $sentinel,%fp,%fp
+___
+for(; $i<28 && $i<$NUM; $i++) {
+$code.=<<___;
+ ldx [$np+$i*8],@N[$i]
+___
+}
+$code.=<<___;
+ save %sp,-128,%sp; or $sentinel,%fp,%fp
+___
+for(; $i<$NUM; $i++) {
+$code.=<<___;
+ ldx [$np+$i*8],@N[$i]
+___
+}
+# load pwrtbl[pwr] ########################################################
+$code.=<<___;
+ save %sp,-128,%sp; or $sentinel,%fp,%fp
+
+ srlx $pwr, 32, %o4 ! unpack $pwr
+ srl $pwr, %g0, %o5
+ sub %o4, 5, %o4
+ mov $pwrtbl, %o7
+ sllx %o4, 32, $pwr ! re-pack $pwr
+ or %o5, $pwr, $pwr
+ srl %o5, %o4, %o5
+___
+ &load_ccr("%o7","%o5","%o4");
+$code.=<<___;
+ b .Lstride_$NUM
+ nop
+.align 16
+.Lstride_$NUM:
+___
+for($i=0; $i<14 && $i<$NUM; $i+=2) {
+ &load_b_pair("%o7",@B[$i],@B[$i+1]);
+}
+$code.=<<___;
+ save %sp,-128,%sp; or $sentinel,%fp,%fp
+___
+for(; $i<$NUM; $i+=2) {
+ &load_b_pair("%i7",@B[$i],@B[$i+1]);
+}
+$code.=<<___;
+ srax $pwr, 32, %o4 ! unpack $pwr
+ srl $pwr, %g0, %o5
+ sub %o4, 5, %o4
+ mov $pwrtbl, %i7
+ sllx %o4, 32, $pwr ! re-pack $pwr
+ or %o5, $pwr, $pwr
+ srl %o5, %o4, %o5
+___
+ &load_ccr("%i7","%o5","%o4",1);
+
+# magic ################################################################
+for($i=0; $i<5; $i++) {
+$code.=<<___;
+ .word 0x81b02940+$NUM-1 ! montsqr $NUM-1
+ fbu,pn %fcc3,.Labort_$NUM
+#ifndef __arch64__
+ and %fp,$sentinel,$sentinel
+ brz,pn $sentinel,.Labort_$NUM
+#endif
+ nop
+___
+}
+$code.=<<___;
+ wr %o4, %g0, %ccr
+ .word 0x81b02920+$NUM-1 ! montmul $NUM-1
+ fbu,pn %fcc3,.Labort_$NUM
+#ifndef __arch64__
+ and %fp,$sentinel,$sentinel
+ brz,pn $sentinel,.Labort_$NUM
+#endif
+
+ srax $pwr, 32, %o4
+#ifdef __arch64__
+ brgez %o4,.Lstride_$NUM
+ restore
+ restore
+ restore
+ restore
+ restore
+#else
+ brgez %o4,.Lstride_$NUM
+ restore; and %fp,$sentinel,$sentinel
+ restore; and %fp,$sentinel,$sentinel
+ restore; and %fp,$sentinel,$sentinel
+ restore; and %fp,$sentinel,$sentinel
+ brz,pn $sentinel,.Labort1_$NUM
+ restore
+#endif
+___
+
+# save tp[$NUM] ########################################################
+for($i=0; $i<14 && $i<$NUM; $i++) {
+$code.=<<___;
+ movxtod @A[$i],@R[$i]
+___
+}
+$code.=<<___;
+#ifdef __arch64__
+ restore
+#else
+ and %fp,$sentinel,$sentinel
+ restore
+ and $sentinel,1,%o7
+ and %fp,$sentinel,$sentinel
+ srl %fp,0,%fp ! just in case?
+ or %o7,$sentinel,$sentinel
+ brz,a,pn $sentinel,.Ldone_$NUM
+ mov 0,%i0 ! return failure
+#endif
+___
+for($i=0; $i<$NUM; $i++) {
+$code.=<<___;
+ std @R[$i],[$tp+$i*8]
+___
+}
+$code.=<<___;
+ mov 1,%i0 ! return success
+.Ldone_$NUM:
+ ret
+ restore
+
+.Labort_$NUM:
+ restore
+ restore
+ restore
+ restore
+ restore
+.Labort1_$NUM:
+ restore
+
+ mov 0,%i0 ! return failure
+ ret
+ restore
+.type bn_pwr5_mont_t4_$NUM, #function
+.size bn_pwr5_mont_t4_$NUM, .-bn_pwr5_mont_t4_$NUM
+___
+}
+
+for ($i=8;$i<=32;$i+=8) {
+ &generate_bn_pwr5_mont_t4($i);
+}
+
+{
+########################################################################
+# Fall-back subroutines
+#
+# copy of bn_mul_mont_vis3 adjusted for vectors of 64-bit values
+#
+($n0,$m0,$m1,$lo0,$hi0, $lo1,$hi1,$aj,$alo,$nj,$nlo,$tj)=
+ (map("%g$_",(1..5)),map("%o$_",(0..5,7)));
+
+# int bn_mul_mont(
+$rp="%o0"; # u64 *rp,
+$ap="%o1"; # const u64 *ap,
+$bp="%o2"; # const u64 *bp,
+$np="%o3"; # const u64 *np,
+$n0p="%o4"; # const BN_ULONG *n0,
+$num="%o5"; # int num); # caller ensures that num is >=3
+$code.=<<___;
+.globl bn_mul_mont_t4
+.align 32
+bn_mul_mont_t4:
+ add %sp, STACK_BIAS, %g4 ! real top of stack
+ sll $num, 3, $num ! size in bytes
+ add $num, 63, %g1
+ andn %g1, 63, %g1 ! buffer size rounded up to 64 bytes
+ sub %g4, %g1, %g1
+ andn %g1, 63, %g1 ! align at 64 byte
+ sub %g1, STACK_FRAME, %g1 ! new top of stack
+ sub %g1, %g4, %g1
+
+ save %sp, %g1, %sp
+___
+# +-------------------------------+<----- %sp
+# . .
+# +-------------------------------+<----- aligned at 64 bytes
+# | __int64 tmp[0] |
+# +-------------------------------+
+# . .
+# . .
+# +-------------------------------+<----- aligned at 64 bytes
+# . .
+($rp,$ap,$bp,$np,$n0p,$num)=map("%i$_",(0..5));
+($t0,$t1,$t2,$t3,$cnt,$tp,$bufsz)=map("%l$_",(0..7));
+($ovf,$i)=($t0,$t1);
+$code.=<<___;
+ ld [$n0p+0], $t0 ! pull n0[0..1] value
+ ld [$n0p+4], $t1
+ add %sp, STACK_BIAS+STACK_FRAME, $tp
+ ldx [$bp+0], $m0 ! m0=bp[0]
+ sllx $t1, 32, $n0
+ add $bp, 8, $bp
+ or $t0, $n0, $n0
+
+ ldx [$ap+0], $aj ! ap[0]
+
+ mulx $aj, $m0, $lo0 ! ap[0]*bp[0]
+ umulxhi $aj, $m0, $hi0
+
+ ldx [$ap+8], $aj ! ap[1]
+ add $ap, 16, $ap
+ ldx [$np+0], $nj ! np[0]
+
+ mulx $lo0, $n0, $m1 ! "tp[0]"*n0
+
+ mulx $aj, $m0, $alo ! ap[1]*bp[0]
+ umulxhi $aj, $m0, $aj ! ahi=aj
+
+ mulx $nj, $m1, $lo1 ! np[0]*m1
+ umulxhi $nj, $m1, $hi1
+
+ ldx [$np+8], $nj ! np[1]
+
+ addcc $lo0, $lo1, $lo1
+ add $np, 16, $np
+ addxc %g0, $hi1, $hi1
+
+ mulx $nj, $m1, $nlo ! np[1]*m1
+ umulxhi $nj, $m1, $nj ! nhi=nj
+
+ ba .L1st
+ sub $num, 24, $cnt ! cnt=num-3
+
+.align 16
+.L1st:
+ addcc $alo, $hi0, $lo0
+ addxc $aj, %g0, $hi0
+
+ ldx [$ap+0], $aj ! ap[j]
+ addcc $nlo, $hi1, $lo1
+ add $ap, 8, $ap
+ addxc $nj, %g0, $hi1 ! nhi=nj
+
+ ldx [$np+0], $nj ! np[j]
+ mulx $aj, $m0, $alo ! ap[j]*bp[0]
+ add $np, 8, $np
+ umulxhi $aj, $m0, $aj ! ahi=aj
+
+ mulx $nj, $m1, $nlo ! np[j]*m1
+ addcc $lo0, $lo1, $lo1 ! np[j]*m1+ap[j]*bp[0]
+ umulxhi $nj, $m1, $nj ! nhi=nj
+ addxc %g0, $hi1, $hi1
+ stxa $lo1, [$tp]0xe2 ! tp[j-1]
+ add $tp, 8, $tp ! tp++
+
+ brnz,pt $cnt, .L1st
+ sub $cnt, 8, $cnt ! j--
+!.L1st
+ addcc $alo, $hi0, $lo0
+ addxc $aj, %g0, $hi0 ! ahi=aj
+
+ addcc $nlo, $hi1, $lo1
+ addxc $nj, %g0, $hi1
+ addcc $lo0, $lo1, $lo1 ! np[j]*m1+ap[j]*bp[0]
+ addxc %g0, $hi1, $hi1
+ stxa $lo1, [$tp]0xe2 ! tp[j-1]
+ add $tp, 8, $tp
+
+ addcc $hi0, $hi1, $hi1
+ addxc %g0, %g0, $ovf ! upmost overflow bit
+ stxa $hi1, [$tp]0xe2
+ add $tp, 8, $tp
+
+ ba .Louter
+ sub $num, 16, $i ! i=num-2
+
+.align 16
+.Louter:
+ ldx [$bp+0], $m0 ! m0=bp[i]
+ add $bp, 8, $bp
+
+ sub $ap, $num, $ap ! rewind
+ sub $np, $num, $np
+ sub $tp, $num, $tp
+
+ ldx [$ap+0], $aj ! ap[0]
+ ldx [$np+0], $nj ! np[0]
+
+ mulx $aj, $m0, $lo0 ! ap[0]*bp[i]
+ ldx [$tp], $tj ! tp[0]
+ umulxhi $aj, $m0, $hi0
+ ldx [$ap+8], $aj ! ap[1]
+ addcc $lo0, $tj, $lo0 ! ap[0]*bp[i]+tp[0]
+ mulx $aj, $m0, $alo ! ap[1]*bp[i]
+ addxc %g0, $hi0, $hi0
+ mulx $lo0, $n0, $m1 ! tp[0]*n0
+ umulxhi $aj, $m0, $aj ! ahi=aj
+ mulx $nj, $m1, $lo1 ! np[0]*m1
+ add $ap, 16, $ap
+ umulxhi $nj, $m1, $hi1
+ ldx [$np+8], $nj ! np[1]
+ add $np, 16, $np
+ addcc $lo1, $lo0, $lo1
+ mulx $nj, $m1, $nlo ! np[1]*m1
+ addxc %g0, $hi1, $hi1
+ umulxhi $nj, $m1, $nj ! nhi=nj
+
+ ba .Linner
+ sub $num, 24, $cnt ! cnt=num-3
+.align 16
+.Linner:
+ addcc $alo, $hi0, $lo0
+ ldx [$tp+8], $tj ! tp[j]
+ addxc $aj, %g0, $hi0 ! ahi=aj
+ ldx [$ap+0], $aj ! ap[j]
+ add $ap, 8, $ap
+ addcc $nlo, $hi1, $lo1
+ mulx $aj, $m0, $alo ! ap[j]*bp[i]
+ addxc $nj, %g0, $hi1 ! nhi=nj
+ ldx [$np+0], $nj ! np[j]
+ add $np, 8, $np
+ umulxhi $aj, $m0, $aj ! ahi=aj
+ addcc $lo0, $tj, $lo0 ! ap[j]*bp[i]+tp[j]
+ mulx $nj, $m1, $nlo ! np[j]*m1
+ addxc %g0, $hi0, $hi0
+ umulxhi $nj, $m1, $nj ! nhi=nj
+ addcc $lo1, $lo0, $lo1 ! np[j]*m1+ap[j]*bp[i]+tp[j]
+ addxc %g0, $hi1, $hi1
+ stx $lo1, [$tp] ! tp[j-1]
+ add $tp, 8, $tp
+ brnz,pt $cnt, .Linner
+ sub $cnt, 8, $cnt
+!.Linner
+ ldx [$tp+8], $tj ! tp[j]
+ addcc $alo, $hi0, $lo0
+ addxc $aj, %g0, $hi0 ! ahi=aj
+ addcc $lo0, $tj, $lo0 ! ap[j]*bp[i]+tp[j]
+ addxc %g0, $hi0, $hi0
+
+ addcc $nlo, $hi1, $lo1
+ addxc $nj, %g0, $hi1 ! nhi=nj
+ addcc $lo1, $lo0, $lo1 ! np[j]*m1+ap[j]*bp[i]+tp[j]
+ addxc %g0, $hi1, $hi1
+ stx $lo1, [$tp] ! tp[j-1]
+
+ subcc %g0, $ovf, %g0 ! move upmost overflow to CCR.xcc
+ addxccc $hi1, $hi0, $hi1
+ addxc %g0, %g0, $ovf
+ stx $hi1, [$tp+8]
+ add $tp, 16, $tp
+
+ brnz,pt $i, .Louter
+ sub $i, 8, $i
+
+ sub $ap, $num, $ap ! rewind
+ sub $np, $num, $np
+ sub $tp, $num, $tp
+ ba .Lsub
+ subcc $num, 8, $cnt ! cnt=num-1 and clear CCR.xcc
+
+.align 16
+.Lsub:
+ ldx [$tp], $tj
+ add $tp, 8, $tp
+ ldx [$np+0], $nj
+ add $np, 8, $np
+ subccc $tj, $nj, $t2 ! tp[j]-np[j]
+ srlx $tj, 32, $tj
+ srlx $nj, 32, $nj
+ subccc $tj, $nj, $t3
+ add $rp, 8, $rp
+ st $t2, [$rp-4] ! reverse order
+ st $t3, [$rp-8]
+ brnz,pt $cnt, .Lsub
+ sub $cnt, 8, $cnt
+
+ sub $np, $num, $np ! rewind
+ sub $tp, $num, $tp
+ sub $rp, $num, $rp
+
+ subccc $ovf, %g0, $ovf ! handle upmost overflow bit
+ ba .Lcopy
+ sub $num, 8, $cnt
+
+.align 16
+.Lcopy: ! conditional copy
+ ldx [$tp], $tj
+ ldx [$rp+0], $t2
+ stx %g0, [$tp] ! zap
+ add $tp, 8, $tp
+ movcs %icc, $tj, $t2
+ stx $t2, [$rp+0]
+ add $rp, 8, $rp
+ brnz $cnt, .Lcopy
+ sub $cnt, 8, $cnt
+
+ mov 1, %o0
+ ret
+ restore
+.type bn_mul_mont_t4, #function
+.size bn_mul_mont_t4, .-bn_mul_mont_t4
+___
+
+# int bn_mul_mont_gather5(
+$rp="%o0"; # u64 *rp,
+$ap="%o1"; # const u64 *ap,
+$bp="%o2"; # const u64 *pwrtbl,
+$np="%o3"; # const u64 *np,
+$n0p="%o4"; # const BN_ULONG *n0,
+$num="%o5"; # int num, # caller ensures that num is >=3
+ # int power);
+$code.=<<___;
+.globl bn_mul_mont_gather5_t4
+.align 32
+bn_mul_mont_gather5_t4:
+ add %sp, STACK_BIAS, %g4 ! real top of stack
+ sll $num, 3, $num ! size in bytes
+ add $num, 63, %g1
+ andn %g1, 63, %g1 ! buffer size rounded up to 64 bytes
+ sub %g4, %g1, %g1
+ andn %g1, 63, %g1 ! align at 64 byte
+ sub %g1, STACK_FRAME, %g1 ! new top of stack
+ sub %g1, %g4, %g1
+ LDPTR [%sp+STACK_7thARG], %g4 ! load power, 7th argument
+
+ save %sp, %g1, %sp
+___
+# +-------------------------------+<----- %sp
+# . .
+# +-------------------------------+<----- aligned at 64 bytes
+# | __int64 tmp[0] |
+# +-------------------------------+
+# . .
+# . .
+# +-------------------------------+<----- aligned at 64 bytes
+# . .
+($rp,$ap,$bp,$np,$n0p,$num)=map("%i$_",(0..5));
+($t0,$t1,$t2,$t3,$cnt,$tp,$bufsz,$ccr)=map("%l$_",(0..7));
+($ovf,$i)=($t0,$t1);
+ &load_ccr($bp,"%g4",$ccr);
+ &load_b($bp,$m0,"%o7"); # m0=bp[0]
+
+$code.=<<___;
+ ld [$n0p+0], $t0 ! pull n0[0..1] value
+ ld [$n0p+4], $t1
+ add %sp, STACK_BIAS+STACK_FRAME, $tp
+ sllx $t1, 32, $n0
+ or $t0, $n0, $n0
+
+ ldx [$ap+0], $aj ! ap[0]
+
+ mulx $aj, $m0, $lo0 ! ap[0]*bp[0]
+ umulxhi $aj, $m0, $hi0
+
+ ldx [$ap+8], $aj ! ap[1]
+ add $ap, 16, $ap
+ ldx [$np+0], $nj ! np[0]
+
+ mulx $lo0, $n0, $m1 ! "tp[0]"*n0
+
+ mulx $aj, $m0, $alo ! ap[1]*bp[0]
+ umulxhi $aj, $m0, $aj ! ahi=aj
+
+ mulx $nj, $m1, $lo1 ! np[0]*m1
+ umulxhi $nj, $m1, $hi1
+
+ ldx [$np+8], $nj ! np[1]
+
+ addcc $lo0, $lo1, $lo1
+ add $np, 16, $np
+ addxc %g0, $hi1, $hi1
+
+ mulx $nj, $m1, $nlo ! np[1]*m1
+ umulxhi $nj, $m1, $nj ! nhi=nj
+
+ ba .L1st_g5
+ sub $num, 24, $cnt ! cnt=num-3
+
+.align 16
+.L1st_g5:
+ addcc $alo, $hi0, $lo0
+ addxc $aj, %g0, $hi0
+
+ ldx [$ap+0], $aj ! ap[j]
+ addcc $nlo, $hi1, $lo1
+ add $ap, 8, $ap
+ addxc $nj, %g0, $hi1 ! nhi=nj
+
+ ldx [$np+0], $nj ! np[j]
+ mulx $aj, $m0, $alo ! ap[j]*bp[0]
+ add $np, 8, $np
+ umulxhi $aj, $m0, $aj ! ahi=aj
+
+ mulx $nj, $m1, $nlo ! np[j]*m1
+ addcc $lo0, $lo1, $lo1 ! np[j]*m1+ap[j]*bp[0]
+ umulxhi $nj, $m1, $nj ! nhi=nj
+ addxc %g0, $hi1, $hi1
+ stxa $lo1, [$tp]0xe2 ! tp[j-1]
+ add $tp, 8, $tp ! tp++
+
+ brnz,pt $cnt, .L1st_g5
+ sub $cnt, 8, $cnt ! j--
+!.L1st_g5
+ addcc $alo, $hi0, $lo0
+ addxc $aj, %g0, $hi0 ! ahi=aj
+
+ addcc $nlo, $hi1, $lo1
+ addxc $nj, %g0, $hi1
+ addcc $lo0, $lo1, $lo1 ! np[j]*m1+ap[j]*bp[0]
+ addxc %g0, $hi1, $hi1
+ stxa $lo1, [$tp]0xe2 ! tp[j-1]
+ add $tp, 8, $tp
+
+ addcc $hi0, $hi1, $hi1
+ addxc %g0, %g0, $ovf ! upmost overflow bit
+ stxa $hi1, [$tp]0xe2
+ add $tp, 8, $tp
+
+ ba .Louter_g5
+ sub $num, 16, $i ! i=num-2
+
+.align 16
+.Louter_g5:
+ wr $ccr, %g0, %ccr
+___
+ &load_b($bp,$m0); # m0=bp[i]
+$code.=<<___;
+ sub $ap, $num, $ap ! rewind
+ sub $np, $num, $np
+ sub $tp, $num, $tp
+
+ ldx [$ap+0], $aj ! ap[0]
+ ldx [$np+0], $nj ! np[0]
+
+ mulx $aj, $m0, $lo0 ! ap[0]*bp[i]
+ ldx [$tp], $tj ! tp[0]
+ umulxhi $aj, $m0, $hi0
+ ldx [$ap+8], $aj ! ap[1]
+ addcc $lo0, $tj, $lo0 ! ap[0]*bp[i]+tp[0]
+ mulx $aj, $m0, $alo ! ap[1]*bp[i]
+ addxc %g0, $hi0, $hi0
+ mulx $lo0, $n0, $m1 ! tp[0]*n0
+ umulxhi $aj, $m0, $aj ! ahi=aj
+ mulx $nj, $m1, $lo1 ! np[0]*m1
+ add $ap, 16, $ap
+ umulxhi $nj, $m1, $hi1
+ ldx [$np+8], $nj ! np[1]
+ add $np, 16, $np
+ addcc $lo1, $lo0, $lo1
+ mulx $nj, $m1, $nlo ! np[1]*m1
+ addxc %g0, $hi1, $hi1
+ umulxhi $nj, $m1, $nj ! nhi=nj
+
+ ba .Linner_g5
+ sub $num, 24, $cnt ! cnt=num-3
+.align 16
+.Linner_g5:
+ addcc $alo, $hi0, $lo0
+ ldx [$tp+8], $tj ! tp[j]
+ addxc $aj, %g0, $hi0 ! ahi=aj
+ ldx [$ap+0], $aj ! ap[j]
+ add $ap, 8, $ap
+ addcc $nlo, $hi1, $lo1
+ mulx $aj, $m0, $alo ! ap[j]*bp[i]
+ addxc $nj, %g0, $hi1 ! nhi=nj
+ ldx [$np+0], $nj ! np[j]
+ add $np, 8, $np
+ umulxhi $aj, $m0, $aj ! ahi=aj
+ addcc $lo0, $tj, $lo0 ! ap[j]*bp[i]+tp[j]
+ mulx $nj, $m1, $nlo ! np[j]*m1
+ addxc %g0, $hi0, $hi0
+ umulxhi $nj, $m1, $nj ! nhi=nj
+ addcc $lo1, $lo0, $lo1 ! np[j]*m1+ap[j]*bp[i]+tp[j]
+ addxc %g0, $hi1, $hi1
+ stx $lo1, [$tp] ! tp[j-1]
+ add $tp, 8, $tp
+ brnz,pt $cnt, .Linner_g5
+ sub $cnt, 8, $cnt
+!.Linner_g5
+ ldx [$tp+8], $tj ! tp[j]
+ addcc $alo, $hi0, $lo0
+ addxc $aj, %g0, $hi0 ! ahi=aj
+ addcc $lo0, $tj, $lo0 ! ap[j]*bp[i]+tp[j]
+ addxc %g0, $hi0, $hi0
+
+ addcc $nlo, $hi1, $lo1
+ addxc $nj, %g0, $hi1 ! nhi=nj
+ addcc $lo1, $lo0, $lo1 ! np[j]*m1+ap[j]*bp[i]+tp[j]
+ addxc %g0, $hi1, $hi1
+ stx $lo1, [$tp] ! tp[j-1]
+
+ subcc %g0, $ovf, %g0 ! move upmost overflow to CCR.xcc
+ addxccc $hi1, $hi0, $hi1
+ addxc %g0, %g0, $ovf
+ stx $hi1, [$tp+8]
+ add $tp, 16, $tp
+
+ brnz,pt $i, .Louter_g5
+ sub $i, 8, $i
+
+ sub $ap, $num, $ap ! rewind
+ sub $np, $num, $np
+ sub $tp, $num, $tp
+ ba .Lsub_g5
+ subcc $num, 8, $cnt ! cnt=num-1 and clear CCR.xcc
+
+.align 16
+.Lsub_g5:
+ ldx [$tp], $tj
+ add $tp, 8, $tp
+ ldx [$np+0], $nj
+ add $np, 8, $np
+ subccc $tj, $nj, $t2 ! tp[j]-np[j]
+ srlx $tj, 32, $tj
+ srlx $nj, 32, $nj
+ subccc $tj, $nj, $t3
+ add $rp, 8, $rp
+ st $t2, [$rp-4] ! reverse order
+ st $t3, [$rp-8]
+ brnz,pt $cnt, .Lsub_g5
+ sub $cnt, 8, $cnt
+
+ sub $np, $num, $np ! rewind
+ sub $tp, $num, $tp
+ sub $rp, $num, $rp
+
+ subccc $ovf, %g0, $ovf ! handle upmost overflow bit
+ ba .Lcopy_g5
+ sub $num, 8, $cnt
+
+.align 16
+.Lcopy_g5: ! conditional copy
+ ldx [$tp], $tj
+ ldx [$rp+0], $t2
+ stx %g0, [$tp] ! zap
+ add $tp, 8, $tp
+ movcs %icc, $tj, $t2
+ stx $t2, [$rp+0]
+ add $rp, 8, $rp
+ brnz $cnt, .Lcopy_g5
+ sub $cnt, 8, $cnt
+
+ mov 1, %o0
+ ret
+ restore
+.type bn_mul_mont_gather5_t4, #function
+.size bn_mul_mont_gather5_t4, .-bn_mul_mont_gather5_t4
+___
+}
+
+$code.=<<___;
+.globl bn_flip_t4
+.align 32
+bn_flip_t4:
+.Loop_flip:
+ ld [%o1+0], %o4
+ sub %o2, 1, %o2
+ ld [%o1+4], %o5
+ add %o1, 8, %o1
+ st %o5, [%o0+0]
+ st %o4, [%o0+4]
+ brnz %o2, .Loop_flip
+ add %o0, 8, %o0
+ retl
+ nop
+.type bn_flip_t4, #function
+.size bn_flip_t4, .-bn_flip_t4
+
+.globl bn_flip_n_scatter5_t4
+.align 32
+bn_flip_n_scatter5_t4:
+ sll %o3, 3, %o3
+ srl %o1, 1, %o1
+ add %o3, %o2, %o2 ! &pwrtbl[pwr]
+ sub %o1, 1, %o1
+.Loop_flip_n_scatter5:
+ ld [%o0+0], %o4 ! inp[i]
+ ld [%o0+4], %o5
+ add %o0, 8, %o0
+ sllx %o5, 32, %o5
+ or %o4, %o5, %o5
+ stx %o5, [%o2]
+ add %o2, 32*8, %o2
+ brnz %o1, .Loop_flip_n_scatter5
+ sub %o1, 1, %o1
+ retl
+ nop
+.type bn_flip_n_scatter5_t4, #function
+.size bn_flip_n_scatter5_t4, .-bn_flip_n_scatter5_t4
+
+.globl bn_gather5_t4
+.align 32
+bn_gather5_t4:
+___
+ &load_ccr("%o2","%o3","%g1");
+$code.=<<___;
+ sub %o1, 1, %o1
+.Loop_gather5:
+___
+ &load_b("%o2","%g1");
+$code.=<<___;
+ stx %g1, [%o0]
+ add %o0, 8, %o0
+ brnz %o1, .Loop_gather5
+ sub %o1, 1, %o1
+
+ retl
+ nop
+.type bn_gather5_t4, #function
+.size bn_gather5_t4, .-bn_gather5_t4
+
+.asciz "Montgomery Multiplication for SPARC T4, David S. Miller, Andy Polyakov"
+.align 4
+___
+
+&emit_assembler();
+
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/sparcv8.S b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/sparcv8.S
new file mode 100644
index 0000000..75d72eb
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/sparcv8.S
@@ -0,0 +1,1458 @@
+.ident "sparcv8.s, Version 1.4"
+.ident "SPARC v8 ISA artwork by Andy Polyakov <appro@openssl.org>"
+
+/*
+ * ====================================================================
+ * Copyright 1999-2016 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ * ====================================================================
+ */
+
+/*
+ * This is my modest contribution to OpenSSL project (see
+ * http://www.openssl.org/ for more information about it) and is
+ * a drop-in SuperSPARC ISA replacement for crypto/bn/bn_asm.c
+ * module. For updates see http://fy.chalmers.se/~appro/hpe/.
+ *
+ * See bn_asm.sparc.v8plus.S for more details.
+ */
+
+/*
+ * Revision history.
+ *
+ * 1.1 - new loop unrolling model(*);
+ * 1.2 - made gas friendly;
+ * 1.3 - fixed problem with /usr/ccs/lib/cpp;
+ * 1.4 - some retunes;
+ *
+ * (*) see bn_asm.sparc.v8plus.S for details
+ */
+
+.section ".text",#alloc,#execinstr
+.file "bn_asm.sparc.v8.S"
+
+.align 32
+
+.global bn_mul_add_words
+/*
+ * BN_ULONG bn_mul_add_words(rp,ap,num,w)
+ * BN_ULONG *rp,*ap;
+ * int num;
+ * BN_ULONG w;
+ */
+bn_mul_add_words:
+ cmp %o2,0
+ bg,a .L_bn_mul_add_words_proceed
+ ld [%o1],%g2
+ retl
+ clr %o0
+
+.L_bn_mul_add_words_proceed:
+ andcc %o2,-4,%g0
+ bz .L_bn_mul_add_words_tail
+ clr %o5
+
+.L_bn_mul_add_words_loop:
+ ld [%o0],%o4
+ ld [%o1+4],%g3
+ umul %o3,%g2,%g2
+ rd %y,%g1
+ addcc %o4,%o5,%o4
+ addx %g1,0,%g1
+ addcc %o4,%g2,%o4
+ st %o4,[%o0]
+ addx %g1,0,%o5
+
+ ld [%o0+4],%o4
+ ld [%o1+8],%g2
+ umul %o3,%g3,%g3
+ dec 4,%o2
+ rd %y,%g1
+ addcc %o4,%o5,%o4
+ addx %g1,0,%g1
+ addcc %o4,%g3,%o4
+ st %o4,[%o0+4]
+ addx %g1,0,%o5
+
+ ld [%o0+8],%o4
+ ld [%o1+12],%g3
+ umul %o3,%g2,%g2
+ inc 16,%o1
+ rd %y,%g1
+ addcc %o4,%o5,%o4
+ addx %g1,0,%g1
+ addcc %o4,%g2,%o4
+ st %o4,[%o0+8]
+ addx %g1,0,%o5
+
+ ld [%o0+12],%o4
+ umul %o3,%g3,%g3
+ inc 16,%o0
+ rd %y,%g1
+ addcc %o4,%o5,%o4
+ addx %g1,0,%g1
+ addcc %o4,%g3,%o4
+ st %o4,[%o0-4]
+ addx %g1,0,%o5
+ andcc %o2,-4,%g0
+ bnz,a .L_bn_mul_add_words_loop
+ ld [%o1],%g2
+
+ tst %o2
+ bnz,a .L_bn_mul_add_words_tail
+ ld [%o1],%g2
+.L_bn_mul_add_words_return:
+ retl
+ mov %o5,%o0
+ nop
+
+.L_bn_mul_add_words_tail:
+ ld [%o0],%o4
+ umul %o3,%g2,%g2
+ addcc %o4,%o5,%o4
+ rd %y,%g1
+ addx %g1,0,%g1
+ addcc %o4,%g2,%o4
+ addx %g1,0,%o5
+ deccc %o2
+ bz .L_bn_mul_add_words_return
+ st %o4,[%o0]
+
+ ld [%o1+4],%g2
+ ld [%o0+4],%o4
+ umul %o3,%g2,%g2
+ rd %y,%g1
+ addcc %o4,%o5,%o4
+ addx %g1,0,%g1
+ addcc %o4,%g2,%o4
+ addx %g1,0,%o5
+ deccc %o2
+ bz .L_bn_mul_add_words_return
+ st %o4,[%o0+4]
+
+ ld [%o1+8],%g2
+ ld [%o0+8],%o4
+ umul %o3,%g2,%g2
+ rd %y,%g1
+ addcc %o4,%o5,%o4
+ addx %g1,0,%g1
+ addcc %o4,%g2,%o4
+ st %o4,[%o0+8]
+ retl
+ addx %g1,0,%o0
+
+.type bn_mul_add_words,#function
+.size bn_mul_add_words,(.-bn_mul_add_words)
+
+.align 32
+
+.global bn_mul_words
+/*
+ * BN_ULONG bn_mul_words(rp,ap,num,w)
+ * BN_ULONG *rp,*ap;
+ * int num;
+ * BN_ULONG w;
+ */
+bn_mul_words:
+ cmp %o2,0
+ bg,a .L_bn_mul_words_proceed
+ ld [%o1],%g2
+ retl
+ clr %o0
+
+.L_bn_mul_words_proceed:
+ andcc %o2,-4,%g0
+ bz .L_bn_mul_words_tail
+ clr %o5
+
+.L_bn_mul_words_loop:
+ ld [%o1+4],%g3
+ umul %o3,%g2,%g2
+ addcc %g2,%o5,%g2
+ rd %y,%g1
+ addx %g1,0,%o5
+ st %g2,[%o0]
+
+ ld [%o1+8],%g2
+ umul %o3,%g3,%g3
+ addcc %g3,%o5,%g3
+ rd %y,%g1
+ dec 4,%o2
+ addx %g1,0,%o5
+ st %g3,[%o0+4]
+
+ ld [%o1+12],%g3
+ umul %o3,%g2,%g2
+ addcc %g2,%o5,%g2
+ rd %y,%g1
+ inc 16,%o1
+ st %g2,[%o0+8]
+ addx %g1,0,%o5
+
+ umul %o3,%g3,%g3
+ addcc %g3,%o5,%g3
+ rd %y,%g1
+ inc 16,%o0
+ addx %g1,0,%o5
+ st %g3,[%o0-4]
+ andcc %o2,-4,%g0
+ nop
+ bnz,a .L_bn_mul_words_loop
+ ld [%o1],%g2
+
+ tst %o2
+ bnz,a .L_bn_mul_words_tail
+ ld [%o1],%g2
+.L_bn_mul_words_return:
+ retl
+ mov %o5,%o0
+ nop
+
+.L_bn_mul_words_tail:
+ umul %o3,%g2,%g2
+ addcc %g2,%o5,%g2
+ rd %y,%g1
+ addx %g1,0,%o5
+ deccc %o2
+ bz .L_bn_mul_words_return
+ st %g2,[%o0]
+ nop
+
+ ld [%o1+4],%g2
+ umul %o3,%g2,%g2
+ addcc %g2,%o5,%g2
+ rd %y,%g1
+ addx %g1,0,%o5
+ deccc %o2
+ bz .L_bn_mul_words_return
+ st %g2,[%o0+4]
+
+ ld [%o1+8],%g2
+ umul %o3,%g2,%g2
+ addcc %g2,%o5,%g2
+ rd %y,%g1
+ st %g2,[%o0+8]
+ retl
+ addx %g1,0,%o0
+
+.type bn_mul_words,#function
+.size bn_mul_words,(.-bn_mul_words)
+
+.align 32
+.global bn_sqr_words
+/*
+ * void bn_sqr_words(r,a,n)
+ * BN_ULONG *r,*a;
+ * int n;
+ */
+bn_sqr_words:
+ cmp %o2,0
+ bg,a .L_bn_sqr_words_proceed
+ ld [%o1],%g2
+ retl
+ clr %o0
+
+.L_bn_sqr_words_proceed:
+ andcc %o2,-4,%g0
+ bz .L_bn_sqr_words_tail
+ clr %o5
+
+.L_bn_sqr_words_loop:
+ ld [%o1+4],%g3
+ umul %g2,%g2,%o4
+ st %o4,[%o0]
+ rd %y,%o5
+ st %o5,[%o0+4]
+
+ ld [%o1+8],%g2
+ umul %g3,%g3,%o4
+ dec 4,%o2
+ st %o4,[%o0+8]
+ rd %y,%o5
+ st %o5,[%o0+12]
+ nop
+
+ ld [%o1+12],%g3
+ umul %g2,%g2,%o4
+ st %o4,[%o0+16]
+ rd %y,%o5
+ inc 16,%o1
+ st %o5,[%o0+20]
+
+ umul %g3,%g3,%o4
+ inc 32,%o0
+ st %o4,[%o0-8]
+ rd %y,%o5
+ st %o5,[%o0-4]
+ andcc %o2,-4,%g2
+ bnz,a .L_bn_sqr_words_loop
+ ld [%o1],%g2
+
+ tst %o2
+ nop
+ bnz,a .L_bn_sqr_words_tail
+ ld [%o1],%g2
+.L_bn_sqr_words_return:
+ retl
+ clr %o0
+
+.L_bn_sqr_words_tail:
+ umul %g2,%g2,%o4
+ st %o4,[%o0]
+ deccc %o2
+ rd %y,%o5
+ bz .L_bn_sqr_words_return
+ st %o5,[%o0+4]
+
+ ld [%o1+4],%g2
+ umul %g2,%g2,%o4
+ st %o4,[%o0+8]
+ deccc %o2
+ rd %y,%o5
+ nop
+ bz .L_bn_sqr_words_return
+ st %o5,[%o0+12]
+
+ ld [%o1+8],%g2
+ umul %g2,%g2,%o4
+ st %o4,[%o0+16]
+ rd %y,%o5
+ st %o5,[%o0+20]
+ retl
+ clr %o0
+
+.type bn_sqr_words,#function
+.size bn_sqr_words,(.-bn_sqr_words)
+
+.align 32
+
+.global bn_div_words
+/*
+ * BN_ULONG bn_div_words(h,l,d)
+ * BN_ULONG h,l,d;
+ */
+bn_div_words:
+ wr %o0,%y
+ udiv %o1,%o2,%o0
+ retl
+ nop
+
+.type bn_div_words,#function
+.size bn_div_words,(.-bn_div_words)
+
+.align 32
+
+.global bn_add_words
+/*
+ * BN_ULONG bn_add_words(rp,ap,bp,n)
+ * BN_ULONG *rp,*ap,*bp;
+ * int n;
+ */
+bn_add_words:
+ cmp %o3,0
+ bg,a .L_bn_add_words_proceed
+ ld [%o1],%o4
+ retl
+ clr %o0
+
+.L_bn_add_words_proceed:
+ andcc %o3,-4,%g0
+ bz .L_bn_add_words_tail
+ clr %g1
+ ba .L_bn_add_words_warn_loop
+ addcc %g0,0,%g0 ! clear carry flag
+
+.L_bn_add_words_loop:
+ ld [%o1],%o4
+.L_bn_add_words_warn_loop:
+ ld [%o2],%o5
+ ld [%o1+4],%g3
+ ld [%o2+4],%g4
+ dec 4,%o3
+ addxcc %o5,%o4,%o5
+ st %o5,[%o0]
+
+ ld [%o1+8],%o4
+ ld [%o2+8],%o5
+ inc 16,%o1
+ addxcc %g3,%g4,%g3
+ st %g3,[%o0+4]
+
+ ld [%o1-4],%g3
+ ld [%o2+12],%g4
+ inc 16,%o2
+ addxcc %o5,%o4,%o5
+ st %o5,[%o0+8]
+
+ inc 16,%o0
+ addxcc %g3,%g4,%g3
+ st %g3,[%o0-4]
+ addx %g0,0,%g1
+ andcc %o3,-4,%g0
+ bnz,a .L_bn_add_words_loop
+ addcc %g1,-1,%g0
+
+ tst %o3
+ bnz,a .L_bn_add_words_tail
+ ld [%o1],%o4
+.L_bn_add_words_return:
+ retl
+ mov %g1,%o0
+
+.L_bn_add_words_tail:
+ addcc %g1,-1,%g0
+ ld [%o2],%o5
+ addxcc %o5,%o4,%o5
+ addx %g0,0,%g1
+ deccc %o3
+ bz .L_bn_add_words_return
+ st %o5,[%o0]
+
+ ld [%o1+4],%o4
+ addcc %g1,-1,%g0
+ ld [%o2+4],%o5
+ addxcc %o5,%o4,%o5
+ addx %g0,0,%g1
+ deccc %o3
+ bz .L_bn_add_words_return
+ st %o5,[%o0+4]
+
+ ld [%o1+8],%o4
+ addcc %g1,-1,%g0
+ ld [%o2+8],%o5
+ addxcc %o5,%o4,%o5
+ st %o5,[%o0+8]
+ retl
+ addx %g0,0,%o0
+
+.type bn_add_words,#function
+.size bn_add_words,(.-bn_add_words)
+
+.align 32
+
+.global bn_sub_words
+/*
+ * BN_ULONG bn_sub_words(rp,ap,bp,n)
+ * BN_ULONG *rp,*ap,*bp;
+ * int n;
+ */
+bn_sub_words:
+ cmp %o3,0
+ bg,a .L_bn_sub_words_proceed
+ ld [%o1],%o4
+ retl
+ clr %o0
+
+.L_bn_sub_words_proceed:
+ andcc %o3,-4,%g0
+ bz .L_bn_sub_words_tail
+ clr %g1
+ ba .L_bn_sub_words_warm_loop
+ addcc %g0,0,%g0 ! clear carry flag
+
+.L_bn_sub_words_loop:
+ ld [%o1],%o4
+.L_bn_sub_words_warm_loop:
+ ld [%o2],%o5
+ ld [%o1+4],%g3
+ ld [%o2+4],%g4
+ dec 4,%o3
+ subxcc %o4,%o5,%o5
+ st %o5,[%o0]
+
+ ld [%o1+8],%o4
+ ld [%o2+8],%o5
+ inc 16,%o1
+ subxcc %g3,%g4,%g4
+ st %g4,[%o0+4]
+
+ ld [%o1-4],%g3
+ ld [%o2+12],%g4
+ inc 16,%o2
+ subxcc %o4,%o5,%o5
+ st %o5,[%o0+8]
+
+ inc 16,%o0
+ subxcc %g3,%g4,%g4
+ st %g4,[%o0-4]
+ addx %g0,0,%g1
+ andcc %o3,-4,%g0
+ bnz,a .L_bn_sub_words_loop
+ addcc %g1,-1,%g0
+
+ tst %o3
+ nop
+ bnz,a .L_bn_sub_words_tail
+ ld [%o1],%o4
+.L_bn_sub_words_return:
+ retl
+ mov %g1,%o0
+
+.L_bn_sub_words_tail:
+ addcc %g1,-1,%g0
+ ld [%o2],%o5
+ subxcc %o4,%o5,%o5
+ addx %g0,0,%g1
+ deccc %o3
+ bz .L_bn_sub_words_return
+ st %o5,[%o0]
+ nop
+
+ ld [%o1+4],%o4
+ addcc %g1,-1,%g0
+ ld [%o2+4],%o5
+ subxcc %o4,%o5,%o5
+ addx %g0,0,%g1
+ deccc %o3
+ bz .L_bn_sub_words_return
+ st %o5,[%o0+4]
+
+ ld [%o1+8],%o4
+ addcc %g1,-1,%g0
+ ld [%o2+8],%o5
+ subxcc %o4,%o5,%o5
+ st %o5,[%o0+8]
+ retl
+ addx %g0,0,%o0
+
+.type bn_sub_words,#function
+.size bn_sub_words,(.-bn_sub_words)
+
+#define FRAME_SIZE -96
+
+/*
+ * Here is register usage map for *all* routines below.
+ */
+#define t_1 %o0
+#define t_2 %o1
+#define c_1 %o2
+#define c_2 %o3
+#define c_3 %o4
+
+#define ap(I) [%i1+4*I]
+#define bp(I) [%i2+4*I]
+#define rp(I) [%i0+4*I]
+
+#define a_0 %l0
+#define a_1 %l1
+#define a_2 %l2
+#define a_3 %l3
+#define a_4 %l4
+#define a_5 %l5
+#define a_6 %l6
+#define a_7 %l7
+
+#define b_0 %i3
+#define b_1 %i4
+#define b_2 %i5
+#define b_3 %o5
+#define b_4 %g1
+#define b_5 %g2
+#define b_6 %g3
+#define b_7 %g4
+
+.align 32
+.global bn_mul_comba8
+/*
+ * void bn_mul_comba8(r,a,b)
+ * BN_ULONG *r,*a,*b;
+ */
+bn_mul_comba8:
+ save %sp,FRAME_SIZE,%sp
+ ld ap(0),a_0
+ ld bp(0),b_0
+ umul a_0,b_0,c_1 !=!mul_add_c(a[0],b[0],c1,c2,c3);
+ ld bp(1),b_1
+ rd %y,c_2
+ st c_1,rp(0) !r[0]=c1;
+
+ umul a_0,b_1,t_1 !=!mul_add_c(a[0],b[1],c2,c3,c1);
+ ld ap(1),a_1
+ addcc c_2,t_1,c_2
+ rd %y,t_2
+ addxcc %g0,t_2,c_3 !=
+ addx %g0,%g0,c_1
+ ld ap(2),a_2
+ umul a_1,b_0,t_1 !mul_add_c(a[1],b[0],c2,c3,c1);
+ addcc c_2,t_1,c_2 !=
+ rd %y,t_2
+ addxcc c_3,t_2,c_3
+ st c_2,rp(1) !r[1]=c2;
+ addx c_1,%g0,c_1 !=
+
+ umul a_2,b_0,t_1 !mul_add_c(a[2],b[0],c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2
+ addxcc c_1,t_2,c_1 !=
+ addx %g0,%g0,c_2
+ ld bp(2),b_2
+ umul a_1,b_1,t_1 !mul_add_c(a[1],b[1],c3,c1,c2);
+ addcc c_3,t_1,c_3 !=
+ rd %y,t_2
+ addxcc c_1,t_2,c_1
+ ld bp(3),b_3
+ addx c_2,%g0,c_2 !=
+ umul a_0,b_2,t_1 !mul_add_c(a[0],b[2],c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2
+ addxcc c_1,t_2,c_1 !=
+ addx c_2,%g0,c_2
+ st c_3,rp(2) !r[2]=c3;
+
+ umul a_0,b_3,t_1 !mul_add_c(a[0],b[3],c1,c2,c3);
+ addcc c_1,t_1,c_1 !=
+ rd %y,t_2
+ addxcc c_2,t_2,c_2
+ addx %g0,%g0,c_3
+ umul a_1,b_2,t_1 !=!mul_add_c(a[1],b[2],c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3 !=
+ ld ap(3),a_3
+ umul a_2,b_1,t_1 !mul_add_c(a[2],b[1],c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2 !=
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3
+ ld ap(4),a_4
+ umul a_3,b_0,t_1 !mul_add_c(a[3],b[0],c1,c2,c3);!=
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3 !=
+ st c_1,rp(3) !r[3]=c1;
+
+ umul a_4,b_0,t_1 !mul_add_c(a[4],b[0],c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2 !=
+ addxcc c_3,t_2,c_3
+ addx %g0,%g0,c_1
+ umul a_3,b_1,t_1 !mul_add_c(a[3],b[1],c2,c3,c1);
+ addcc c_2,t_1,c_2 !=
+ rd %y,t_2
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1
+ umul a_2,b_2,t_1 !=!mul_add_c(a[2],b[2],c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1 !=
+ ld bp(4),b_4
+ umul a_1,b_3,t_1 !mul_add_c(a[1],b[3],c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2 !=
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1
+ ld bp(5),b_5
+ umul a_0,b_4,t_1 !=!mul_add_c(a[0],b[4],c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1 !=
+ st c_2,rp(4) !r[4]=c2;
+
+ umul a_0,b_5,t_1 !mul_add_c(a[0],b[5],c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2 !=
+ addxcc c_1,t_2,c_1
+ addx %g0,%g0,c_2
+ umul a_1,b_4,t_1 !mul_add_c(a[1],b[4],c3,c1,c2);
+ addcc c_3,t_1,c_3 !=
+ rd %y,t_2
+ addxcc c_1,t_2,c_1
+ addx c_2,%g0,c_2
+ umul a_2,b_3,t_1 !=!mul_add_c(a[2],b[3],c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2
+ addxcc c_1,t_2,c_1
+ addx c_2,%g0,c_2 !=
+ umul a_3,b_2,t_1 !mul_add_c(a[3],b[2],c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2
+ addxcc c_1,t_2,c_1 !=
+ addx c_2,%g0,c_2
+ ld ap(5),a_5
+ umul a_4,b_1,t_1 !mul_add_c(a[4],b[1],c3,c1,c2);
+ addcc c_3,t_1,c_3 !=
+ rd %y,t_2
+ addxcc c_1,t_2,c_1
+ ld ap(6),a_6
+ addx c_2,%g0,c_2 !=
+ umul a_5,b_0,t_1 !mul_add_c(a[5],b[0],c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2
+ addxcc c_1,t_2,c_1 !=
+ addx c_2,%g0,c_2
+ st c_3,rp(5) !r[5]=c3;
+
+ umul a_6,b_0,t_1 !mul_add_c(a[6],b[0],c1,c2,c3);
+ addcc c_1,t_1,c_1 !=
+ rd %y,t_2
+ addxcc c_2,t_2,c_2
+ addx %g0,%g0,c_3
+ umul a_5,b_1,t_1 !=!mul_add_c(a[5],b[1],c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3 !=
+ umul a_4,b_2,t_1 !mul_add_c(a[4],b[2],c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2 !=
+ addx c_3,%g0,c_3
+ umul a_3,b_3,t_1 !mul_add_c(a[3],b[3],c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2 !=
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3
+ umul a_2,b_4,t_1 !mul_add_c(a[2],b[4],c1,c2,c3);
+ addcc c_1,t_1,c_1 !=
+ rd %y,t_2
+ addxcc c_2,t_2,c_2
+ ld bp(6),b_6
+ addx c_3,%g0,c_3 !=
+ umul a_1,b_5,t_1 !mul_add_c(a[1],b[5],c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2 !=
+ addx c_3,%g0,c_3
+ ld bp(7),b_7
+ umul a_0,b_6,t_1 !mul_add_c(a[0],b[6],c1,c2,c3);
+ addcc c_1,t_1,c_1 !=
+ rd %y,t_2
+ addxcc c_2,t_2,c_2
+ st c_1,rp(6) !r[6]=c1;
+ addx c_3,%g0,c_3 !=
+
+ umul a_0,b_7,t_1 !mul_add_c(a[0],b[7],c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2
+ addxcc c_3,t_2,c_3 !=
+ addx %g0,%g0,c_1
+ umul a_1,b_6,t_1 !mul_add_c(a[1],b[6],c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2 !=
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1
+ umul a_2,b_5,t_1 !mul_add_c(a[2],b[5],c2,c3,c1);
+ addcc c_2,t_1,c_2 !=
+ rd %y,t_2
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1
+ umul a_3,b_4,t_1 !=!mul_add_c(a[3],b[4],c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1 !=
+ umul a_4,b_3,t_1 !mul_add_c(a[4],b[3],c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2
+ addxcc c_3,t_2,c_3 !=
+ addx c_1,%g0,c_1
+ umul a_5,b_2,t_1 !mul_add_c(a[5],b[2],c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2 !=
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1
+ ld ap(7),a_7
+ umul a_6,b_1,t_1 !=!mul_add_c(a[6],b[1],c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1 !=
+ umul a_7,b_0,t_1 !mul_add_c(a[7],b[0],c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2
+ addxcc c_3,t_2,c_3 !=
+ addx c_1,%g0,c_1
+ st c_2,rp(7) !r[7]=c2;
+
+ umul a_7,b_1,t_1 !mul_add_c(a[7],b[1],c3,c1,c2);
+ addcc c_3,t_1,c_3 !=
+ rd %y,t_2
+ addxcc c_1,t_2,c_1
+ addx %g0,%g0,c_2
+ umul a_6,b_2,t_1 !=!mul_add_c(a[6],b[2],c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2
+ addxcc c_1,t_2,c_1
+ addx c_2,%g0,c_2 !=
+ umul a_5,b_3,t_1 !mul_add_c(a[5],b[3],c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2
+ addxcc c_1,t_2,c_1 !=
+ addx c_2,%g0,c_2
+ umul a_4,b_4,t_1 !mul_add_c(a[4],b[4],c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2 !=
+ addxcc c_1,t_2,c_1
+ addx c_2,%g0,c_2
+ umul a_3,b_5,t_1 !mul_add_c(a[3],b[5],c3,c1,c2);
+ addcc c_3,t_1,c_3 !=
+ rd %y,t_2
+ addxcc c_1,t_2,c_1
+ addx c_2,%g0,c_2
+ umul a_2,b_6,t_1 !=!mul_add_c(a[2],b[6],c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2
+ addxcc c_1,t_2,c_1
+ addx c_2,%g0,c_2 !=
+ umul a_1,b_7,t_1 !mul_add_c(a[1],b[7],c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2
+ addxcc c_1,t_2,c_1 !
+ addx c_2,%g0,c_2
+ st c_3,rp(8) !r[8]=c3;
+
+ umul a_2,b_7,t_1 !mul_add_c(a[2],b[7],c1,c2,c3);
+ addcc c_1,t_1,c_1 !=
+ rd %y,t_2
+ addxcc c_2,t_2,c_2
+ addx %g0,%g0,c_3
+ umul a_3,b_6,t_1 !=!mul_add_c(a[3],b[6],c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3 !=
+ umul a_4,b_5,t_1 !mul_add_c(a[4],b[5],c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2 !=
+ addx c_3,%g0,c_3
+ umul a_5,b_4,t_1 !mul_add_c(a[5],b[4],c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2 !=
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3
+ umul a_6,b_3,t_1 !mul_add_c(a[6],b[3],c1,c2,c3);
+ addcc c_1,t_1,c_1 !=
+ rd %y,t_2
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3
+ umul a_7,b_2,t_1 !=!mul_add_c(a[7],b[2],c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3 !=
+ st c_1,rp(9) !r[9]=c1;
+
+ umul a_7,b_3,t_1 !mul_add_c(a[7],b[3],c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2 !=
+ addxcc c_3,t_2,c_3
+ addx %g0,%g0,c_1
+ umul a_6,b_4,t_1 !mul_add_c(a[6],b[4],c2,c3,c1);
+ addcc c_2,t_1,c_2 !=
+ rd %y,t_2
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1
+ umul a_5,b_5,t_1 !=!mul_add_c(a[5],b[5],c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1 !=
+ umul a_4,b_6,t_1 !mul_add_c(a[4],b[6],c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2
+ addxcc c_3,t_2,c_3 !=
+ addx c_1,%g0,c_1
+ umul a_3,b_7,t_1 !mul_add_c(a[3],b[7],c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2 !=
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1
+ st c_2,rp(10) !r[10]=c2;
+
+ umul a_4,b_7,t_1 !=!mul_add_c(a[4],b[7],c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2
+ addxcc c_1,t_2,c_1
+ addx %g0,%g0,c_2 !=
+ umul a_5,b_6,t_1 !mul_add_c(a[5],b[6],c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2
+ addxcc c_1,t_2,c_1 !=
+ addx c_2,%g0,c_2
+ umul a_6,b_5,t_1 !mul_add_c(a[6],b[5],c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2 !=
+ addxcc c_1,t_2,c_1
+ addx c_2,%g0,c_2
+ umul a_7,b_4,t_1 !mul_add_c(a[7],b[4],c3,c1,c2);
+ addcc c_3,t_1,c_3 !=
+ rd %y,t_2
+ addxcc c_1,t_2,c_1
+ st c_3,rp(11) !r[11]=c3;
+ addx c_2,%g0,c_2 !=
+
+ umul a_7,b_5,t_1 !mul_add_c(a[7],b[5],c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2 !=
+ addx %g0,%g0,c_3
+ umul a_6,b_6,t_1 !mul_add_c(a[6],b[6],c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2 !=
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3
+ umul a_5,b_7,t_1 !mul_add_c(a[5],b[7],c1,c2,c3);
+ addcc c_1,t_1,c_1 !=
+ rd %y,t_2
+ addxcc c_2,t_2,c_2
+ st c_1,rp(12) !r[12]=c1;
+ addx c_3,%g0,c_3 !=
+
+ umul a_6,b_7,t_1 !mul_add_c(a[6],b[7],c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2
+ addxcc c_3,t_2,c_3 !=
+ addx %g0,%g0,c_1
+ umul a_7,b_6,t_1 !mul_add_c(a[7],b[6],c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2 !=
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1
+ st c_2,rp(13) !r[13]=c2;
+
+ umul a_7,b_7,t_1 !=!mul_add_c(a[7],b[7],c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2
+ addxcc c_1,t_2,c_1
+ nop !=
+ st c_3,rp(14) !r[14]=c3;
+ st c_1,rp(15) !r[15]=c1;
+
+ ret
+ restore %g0,%g0,%o0
+
+.type bn_mul_comba8,#function
+.size bn_mul_comba8,(.-bn_mul_comba8)
+
+.align 32
+
+.global bn_mul_comba4
+/*
+ * void bn_mul_comba4(r,a,b)
+ * BN_ULONG *r,*a,*b;
+ */
+bn_mul_comba4:
+ save %sp,FRAME_SIZE,%sp
+ ld ap(0),a_0
+ ld bp(0),b_0
+ umul a_0,b_0,c_1 !=!mul_add_c(a[0],b[0],c1,c2,c3);
+ ld bp(1),b_1
+ rd %y,c_2
+ st c_1,rp(0) !r[0]=c1;
+
+ umul a_0,b_1,t_1 !=!mul_add_c(a[0],b[1],c2,c3,c1);
+ ld ap(1),a_1
+ addcc c_2,t_1,c_2
+ rd %y,t_2 !=
+ addxcc %g0,t_2,c_3
+ addx %g0,%g0,c_1
+ ld ap(2),a_2
+ umul a_1,b_0,t_1 !=!mul_add_c(a[1],b[0],c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1 !=
+ st c_2,rp(1) !r[1]=c2;
+
+ umul a_2,b_0,t_1 !mul_add_c(a[2],b[0],c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2 !=
+ addxcc c_1,t_2,c_1
+ addx %g0,%g0,c_2
+ ld bp(2),b_2
+ umul a_1,b_1,t_1 !=!mul_add_c(a[1],b[1],c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2
+ addxcc c_1,t_2,c_1
+ addx c_2,%g0,c_2 !=
+ ld bp(3),b_3
+ umul a_0,b_2,t_1 !mul_add_c(a[0],b[2],c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2 !=
+ addxcc c_1,t_2,c_1
+ addx c_2,%g0,c_2
+ st c_3,rp(2) !r[2]=c3;
+
+ umul a_0,b_3,t_1 !=!mul_add_c(a[0],b[3],c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2
+ addx %g0,%g0,c_3 !=
+ umul a_1,b_2,t_1 !mul_add_c(a[1],b[2],c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2 !=
+ addx c_3,%g0,c_3
+ ld ap(3),a_3
+ umul a_2,b_1,t_1 !mul_add_c(a[2],b[1],c1,c2,c3);
+ addcc c_1,t_1,c_1 !=
+ rd %y,t_2
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3
+ umul a_3,b_0,t_1 !=!mul_add_c(a[3],b[0],c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3 !=
+ st c_1,rp(3) !r[3]=c1;
+
+ umul a_3,b_1,t_1 !mul_add_c(a[3],b[1],c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2 !=
+ addxcc c_3,t_2,c_3
+ addx %g0,%g0,c_1
+ umul a_2,b_2,t_1 !mul_add_c(a[2],b[2],c2,c3,c1);
+ addcc c_2,t_1,c_2 !=
+ rd %y,t_2
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1
+ umul a_1,b_3,t_1 !=!mul_add_c(a[1],b[3],c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1 !=
+ st c_2,rp(4) !r[4]=c2;
+
+ umul a_2,b_3,t_1 !mul_add_c(a[2],b[3],c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2 !=
+ addxcc c_1,t_2,c_1
+ addx %g0,%g0,c_2
+ umul a_3,b_2,t_1 !mul_add_c(a[3],b[2],c3,c1,c2);
+ addcc c_3,t_1,c_3 !=
+ rd %y,t_2
+ addxcc c_1,t_2,c_1
+ st c_3,rp(5) !r[5]=c3;
+ addx c_2,%g0,c_2 !=
+
+ umul a_3,b_3,t_1 !mul_add_c(a[3],b[3],c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2 !=
+ st c_1,rp(6) !r[6]=c1;
+ st c_2,rp(7) !r[7]=c2;
+
+ ret
+ restore %g0,%g0,%o0
+
+.type bn_mul_comba4,#function
+.size bn_mul_comba4,(.-bn_mul_comba4)
+
+.align 32
+
+.global bn_sqr_comba8
+bn_sqr_comba8:
+ save %sp,FRAME_SIZE,%sp
+ ld ap(0),a_0
+ ld ap(1),a_1
+ umul a_0,a_0,c_1 !=!sqr_add_c(a,0,c1,c2,c3);
+ rd %y,c_2
+ st c_1,rp(0) !r[0]=c1;
+
+ ld ap(2),a_2
+ umul a_0,a_1,t_1 !=!sqr_add_c2(a,1,0,c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2
+ addxcc %g0,t_2,c_3
+ addx %g0,%g0,c_1 !=
+ addcc c_2,t_1,c_2
+ addxcc c_3,t_2,c_3
+ st c_2,rp(1) !r[1]=c2;
+ addx c_1,%g0,c_1 !=
+
+ umul a_2,a_0,t_1 !sqr_add_c2(a,2,0,c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2
+ addxcc c_1,t_2,c_1 !=
+ addx %g0,%g0,c_2
+ addcc c_3,t_1,c_3
+ addxcc c_1,t_2,c_1
+ addx c_2,%g0,c_2 !=
+ ld ap(3),a_3
+ umul a_1,a_1,t_1 !sqr_add_c(a,1,c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2 !=
+ addxcc c_1,t_2,c_1
+ addx c_2,%g0,c_2
+ st c_3,rp(2) !r[2]=c3;
+
+ umul a_0,a_3,t_1 !=!sqr_add_c2(a,3,0,c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2
+ addx %g0,%g0,c_3 !=
+ addcc c_1,t_1,c_1
+ addxcc c_2,t_2,c_2
+ ld ap(4),a_4
+ addx c_3,%g0,c_3 !=
+ umul a_1,a_2,t_1 !sqr_add_c2(a,2,1,c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2 !=
+ addx c_3,%g0,c_3
+ addcc c_1,t_1,c_1
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3 !=
+ st c_1,rp(3) !r[3]=c1;
+
+ umul a_4,a_0,t_1 !sqr_add_c2(a,4,0,c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2 !=
+ addxcc c_3,t_2,c_3
+ addx %g0,%g0,c_1
+ addcc c_2,t_1,c_2
+ addxcc c_3,t_2,c_3 !=
+ addx c_1,%g0,c_1
+ umul a_3,a_1,t_1 !sqr_add_c2(a,3,1,c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2 !=
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1
+ addcc c_2,t_1,c_2
+ addxcc c_3,t_2,c_3 !=
+ addx c_1,%g0,c_1
+ ld ap(5),a_5
+ umul a_2,a_2,t_1 !sqr_add_c(a,2,c2,c3,c1);
+ addcc c_2,t_1,c_2 !=
+ rd %y,t_2
+ addxcc c_3,t_2,c_3
+ st c_2,rp(4) !r[4]=c2;
+ addx c_1,%g0,c_1 !=
+
+ umul a_0,a_5,t_1 !sqr_add_c2(a,5,0,c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2
+ addxcc c_1,t_2,c_1 !=
+ addx %g0,%g0,c_2
+ addcc c_3,t_1,c_3
+ addxcc c_1,t_2,c_1
+ addx c_2,%g0,c_2 !=
+ umul a_1,a_4,t_1 !sqr_add_c2(a,4,1,c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2
+ addxcc c_1,t_2,c_1 !=
+ addx c_2,%g0,c_2
+ addcc c_3,t_1,c_3
+ addxcc c_1,t_2,c_1
+ addx c_2,%g0,c_2 !=
+ ld ap(6),a_6
+ umul a_2,a_3,t_1 !sqr_add_c2(a,3,2,c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2 !=
+ addxcc c_1,t_2,c_1
+ addx c_2,%g0,c_2
+ addcc c_3,t_1,c_3
+ addxcc c_1,t_2,c_1 !=
+ addx c_2,%g0,c_2
+ st c_3,rp(5) !r[5]=c3;
+
+ umul a_6,a_0,t_1 !sqr_add_c2(a,6,0,c1,c2,c3);
+ addcc c_1,t_1,c_1 !=
+ rd %y,t_2
+ addxcc c_2,t_2,c_2
+ addx %g0,%g0,c_3
+ addcc c_1,t_1,c_1 !=
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3
+ umul a_5,a_1,t_1 !sqr_add_c2(a,5,1,c1,c2,c3);
+ addcc c_1,t_1,c_1 !=
+ rd %y,t_2
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3
+ addcc c_1,t_1,c_1 !=
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3
+ umul a_4,a_2,t_1 !sqr_add_c2(a,4,2,c1,c2,c3);
+ addcc c_1,t_1,c_1 !=
+ rd %y,t_2
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3
+ addcc c_1,t_1,c_1 !=
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3
+ ld ap(7),a_7
+ umul a_3,a_3,t_1 !=!sqr_add_c(a,3,c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3 !=
+ st c_1,rp(6) !r[6]=c1;
+
+ umul a_0,a_7,t_1 !sqr_add_c2(a,7,0,c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2 !=
+ addxcc c_3,t_2,c_3
+ addx %g0,%g0,c_1
+ addcc c_2,t_1,c_2
+ addxcc c_3,t_2,c_3 !=
+ addx c_1,%g0,c_1
+ umul a_1,a_6,t_1 !sqr_add_c2(a,6,1,c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2 !=
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1
+ addcc c_2,t_1,c_2
+ addxcc c_3,t_2,c_3 !=
+ addx c_1,%g0,c_1
+ umul a_2,a_5,t_1 !sqr_add_c2(a,5,2,c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2 !=
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1
+ addcc c_2,t_1,c_2
+ addxcc c_3,t_2,c_3 !=
+ addx c_1,%g0,c_1
+ umul a_3,a_4,t_1 !sqr_add_c2(a,4,3,c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2 !=
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1
+ addcc c_2,t_1,c_2
+ addxcc c_3,t_2,c_3 !=
+ addx c_1,%g0,c_1
+ st c_2,rp(7) !r[7]=c2;
+
+ umul a_7,a_1,t_1 !sqr_add_c2(a,7,1,c3,c1,c2);
+ addcc c_3,t_1,c_3 !=
+ rd %y,t_2
+ addxcc c_1,t_2,c_1
+ addx %g0,%g0,c_2
+ addcc c_3,t_1,c_3 !=
+ addxcc c_1,t_2,c_1
+ addx c_2,%g0,c_2
+ umul a_6,a_2,t_1 !sqr_add_c2(a,6,2,c3,c1,c2);
+ addcc c_3,t_1,c_3 !=
+ rd %y,t_2
+ addxcc c_1,t_2,c_1
+ addx c_2,%g0,c_2
+ addcc c_3,t_1,c_3 !=
+ addxcc c_1,t_2,c_1
+ addx c_2,%g0,c_2
+ umul a_5,a_3,t_1 !sqr_add_c2(a,5,3,c3,c1,c2);
+ addcc c_3,t_1,c_3 !=
+ rd %y,t_2
+ addxcc c_1,t_2,c_1
+ addx c_2,%g0,c_2
+ addcc c_3,t_1,c_3 !=
+ addxcc c_1,t_2,c_1
+ addx c_2,%g0,c_2
+ umul a_4,a_4,t_1 !sqr_add_c(a,4,c3,c1,c2);
+ addcc c_3,t_1,c_3 !=
+ rd %y,t_2
+ addxcc c_1,t_2,c_1
+ st c_3,rp(8) !r[8]=c3;
+ addx c_2,%g0,c_2 !=
+
+ umul a_2,a_7,t_1 !sqr_add_c2(a,7,2,c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2 !=
+ addx %g0,%g0,c_3
+ addcc c_1,t_1,c_1
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3 !=
+ umul a_3,a_6,t_1 !sqr_add_c2(a,6,3,c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2 !=
+ addx c_3,%g0,c_3
+ addcc c_1,t_1,c_1
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3 !=
+ umul a_4,a_5,t_1 !sqr_add_c2(a,5,4,c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2 !=
+ addx c_3,%g0,c_3
+ addcc c_1,t_1,c_1
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3 !=
+ st c_1,rp(9) !r[9]=c1;
+
+ umul a_7,a_3,t_1 !sqr_add_c2(a,7,3,c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2 !=
+ addxcc c_3,t_2,c_3
+ addx %g0,%g0,c_1
+ addcc c_2,t_1,c_2
+ addxcc c_3,t_2,c_3 !=
+ addx c_1,%g0,c_1
+ umul a_6,a_4,t_1 !sqr_add_c2(a,6,4,c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2 !=
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1
+ addcc c_2,t_1,c_2
+ addxcc c_3,t_2,c_3 !=
+ addx c_1,%g0,c_1
+ umul a_5,a_5,t_1 !sqr_add_c(a,5,c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2 !=
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1
+ st c_2,rp(10) !r[10]=c2;
+
+ umul a_4,a_7,t_1 !=!sqr_add_c2(a,7,4,c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2
+ addxcc c_1,t_2,c_1
+ addx %g0,%g0,c_2 !=
+ addcc c_3,t_1,c_3
+ addxcc c_1,t_2,c_1
+ addx c_2,%g0,c_2
+ umul a_5,a_6,t_1 !=!sqr_add_c2(a,6,5,c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2
+ addxcc c_1,t_2,c_1
+ addx c_2,%g0,c_2 !=
+ addcc c_3,t_1,c_3
+ addxcc c_1,t_2,c_1
+ st c_3,rp(11) !r[11]=c3;
+ addx c_2,%g0,c_2 !=
+
+ umul a_7,a_5,t_1 !sqr_add_c2(a,7,5,c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2 !=
+ addx %g0,%g0,c_3
+ addcc c_1,t_1,c_1
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3 !=
+ umul a_6,a_6,t_1 !sqr_add_c(a,6,c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2 !=
+ addx c_3,%g0,c_3
+ st c_1,rp(12) !r[12]=c1;
+
+ umul a_6,a_7,t_1 !sqr_add_c2(a,7,6,c2,c3,c1);
+ addcc c_2,t_1,c_2 !=
+ rd %y,t_2
+ addxcc c_3,t_2,c_3
+ addx %g0,%g0,c_1
+ addcc c_2,t_1,c_2 !=
+ addxcc c_3,t_2,c_3
+ st c_2,rp(13) !r[13]=c2;
+ addx c_1,%g0,c_1 !=
+
+ umul a_7,a_7,t_1 !sqr_add_c(a,7,c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2
+ addxcc c_1,t_2,c_1 !=
+ st c_3,rp(14) !r[14]=c3;
+ st c_1,rp(15) !r[15]=c1;
+
+ ret
+ restore %g0,%g0,%o0
+
+.type bn_sqr_comba8,#function
+.size bn_sqr_comba8,(.-bn_sqr_comba8)
+
+.align 32
+
+.global bn_sqr_comba4
+/*
+ * void bn_sqr_comba4(r,a)
+ * BN_ULONG *r,*a;
+ */
+bn_sqr_comba4:
+ save %sp,FRAME_SIZE,%sp
+ ld ap(0),a_0
+ umul a_0,a_0,c_1 !sqr_add_c(a,0,c1,c2,c3);
+ ld ap(1),a_1 !=
+ rd %y,c_2
+ st c_1,rp(0) !r[0]=c1;
+
+ ld ap(2),a_2
+ umul a_0,a_1,t_1 !=!sqr_add_c2(a,1,0,c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2
+ addxcc %g0,t_2,c_3
+ addx %g0,%g0,c_1 !=
+ addcc c_2,t_1,c_2
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1 !=
+ st c_2,rp(1) !r[1]=c2;
+
+ umul a_2,a_0,t_1 !sqr_add_c2(a,2,0,c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2 !=
+ addxcc c_1,t_2,c_1
+ addx %g0,%g0,c_2
+ addcc c_3,t_1,c_3
+ addxcc c_1,t_2,c_1 !=
+ addx c_2,%g0,c_2
+ ld ap(3),a_3
+ umul a_1,a_1,t_1 !sqr_add_c(a,1,c3,c1,c2);
+ addcc c_3,t_1,c_3 !=
+ rd %y,t_2
+ addxcc c_1,t_2,c_1
+ st c_3,rp(2) !r[2]=c3;
+ addx c_2,%g0,c_2 !=
+
+ umul a_0,a_3,t_1 !sqr_add_c2(a,3,0,c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2 !=
+ addx %g0,%g0,c_3
+ addcc c_1,t_1,c_1
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3 !=
+ umul a_1,a_2,t_1 !sqr_add_c2(a,2,1,c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2 !=
+ addx c_3,%g0,c_3
+ addcc c_1,t_1,c_1
+ addxcc c_2,t_2,c_2
+ addx c_3,%g0,c_3 !=
+ st c_1,rp(3) !r[3]=c1;
+
+ umul a_3,a_1,t_1 !sqr_add_c2(a,3,1,c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2 !=
+ addxcc c_3,t_2,c_3
+ addx %g0,%g0,c_1
+ addcc c_2,t_1,c_2
+ addxcc c_3,t_2,c_3 !=
+ addx c_1,%g0,c_1
+ umul a_2,a_2,t_1 !sqr_add_c(a,2,c2,c3,c1);
+ addcc c_2,t_1,c_2
+ rd %y,t_2 !=
+ addxcc c_3,t_2,c_3
+ addx c_1,%g0,c_1
+ st c_2,rp(4) !r[4]=c2;
+
+ umul a_2,a_3,t_1 !=!sqr_add_c2(a,3,2,c3,c1,c2);
+ addcc c_3,t_1,c_3
+ rd %y,t_2
+ addxcc c_1,t_2,c_1
+ addx %g0,%g0,c_2 !=
+ addcc c_3,t_1,c_3
+ addxcc c_1,t_2,c_1
+ st c_3,rp(5) !r[5]=c3;
+ addx c_2,%g0,c_2 !=
+
+ umul a_3,a_3,t_1 !sqr_add_c(a,3,c1,c2,c3);
+ addcc c_1,t_1,c_1
+ rd %y,t_2
+ addxcc c_2,t_2,c_2 !=
+ st c_1,rp(6) !r[6]=c1;
+ st c_2,rp(7) !r[7]=c2;
+
+ ret
+ restore %g0,%g0,%o0
+
+.type bn_sqr_comba4,#function
+.size bn_sqr_comba4,(.-bn_sqr_comba4)
+
+.align 32
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/sparcv8plus.S b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/sparcv8plus.S
new file mode 100644
index 0000000..d520ffa
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/sparcv8plus.S
@@ -0,0 +1,1558 @@
+.ident "sparcv8plus.s, Version 1.4"
+.ident "SPARC v9 ISA artwork by Andy Polyakov <appro@openssl.org>"
+
+/*
+ * ====================================================================
+ * Copyright 1999-2019 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ * ====================================================================
+ */
+
+/*
+ * This is my modest contribution to OpenSSL project (see
+ * http://www.openssl.org/ for more information about it) and is
+ * a drop-in UltraSPARC ISA replacement for crypto/bn/bn_asm.c
+ * module. For updates see http://fy.chalmers.se/~appro/hpe/.
+ *
+ * Questions-n-answers.
+ *
+ * Q. How to compile?
+ * A. With SC4.x/SC5.x:
+ *
+ * cc -xarch=v8plus -c bn_asm.sparc.v8plus.S -o bn_asm.o
+ *
+ * and with gcc:
+ *
+ * gcc -mcpu=ultrasparc -c bn_asm.sparc.v8plus.S -o bn_asm.o
+ *
+ * or if above fails (it does if you have gas installed):
+ *
+ * gcc -E bn_asm.sparc.v8plus.S | as -xarch=v8plus /dev/fd/0 -o bn_asm.o
+ *
+ * Quick-n-dirty way to fuse the module into the library.
+ * Provided that the library is already configured and built
+ * (in 0.9.2 case with no-asm option):
+ *
+ * # cd crypto/bn
+ * # cp /some/place/bn_asm.sparc.v8plus.S .
+ * # cc -xarch=v8plus -c bn_asm.sparc.v8plus.S -o bn_asm.o
+ * # make
+ * # cd ../..
+ * # make; make test
+ *
+ * Quick-n-dirty way to get rid of it:
+ *
+ * # cd crypto/bn
+ * # touch bn_asm.c
+ * # make
+ * # cd ../..
+ * # make; make test
+ *
+ * Q. V8plus architecture? What kind of beast is that?
+ * A. Well, it's rather a programming model than an architecture...
+ * It's actually v9-compliant, i.e. *any* UltraSPARC, CPU under
+ * special conditions, namely when kernel doesn't preserve upper
+ * 32 bits of otherwise 64-bit registers during a context switch.
+ *
+ * Q. Why just UltraSPARC? What about SuperSPARC?
+ * A. Original release did target UltraSPARC only. Now SuperSPARC
+ * version is provided along. Both version share bn_*comba[48]
+ * implementations (see comment later in code for explanation).
+ * But what's so special about this UltraSPARC implementation?
+ * Why didn't I let compiler do the job? Trouble is that most of
+ * available compilers (well, SC5.0 is the only exception) don't
+ * attempt to take advantage of UltraSPARC's 64-bitness under
+ * 32-bit kernels even though it's perfectly possible (see next
+ * question).
+ *
+ * Q. 64-bit registers under 32-bit kernels? Didn't you just say it
+ * doesn't work?
+ * A. You can't address *all* registers as 64-bit wide:-( The catch is
+ * that you actually may rely upon %o0-%o5 and %g1-%g4 being fully
+ * preserved if you're in a leaf function, i.e. such never calling
+ * any other functions. All functions in this module are leaf and
+ * 10 registers is a handful. And as a matter of fact none-"comba"
+ * routines don't require even that much and I could even afford to
+ * not allocate own stack frame for 'em:-)
+ *
+ * Q. What about 64-bit kernels?
+ * A. What about 'em? Just kidding:-) Pure 64-bit version is currently
+ * under evaluation and development...
+ *
+ * Q. What about shared libraries?
+ * A. What about 'em? Kidding again:-) Code does *not* contain any
+ * code position dependencies and it's safe to include it into
+ * shared library as is.
+ *
+ * Q. How much faster does it go?
+ * A. Do you have a good benchmark? In either case below is what I
+ * experience with crypto/bn/expspeed.c test program:
+ *
+ * v8plus module on U10/300MHz against bn_asm.c compiled with:
+ *
+ * cc-5.0 -xarch=v8plus -xO5 -xdepend +7-12%
+ * cc-4.2 -xarch=v8plus -xO5 -xdepend +25-35%
+ * egcs-1.1.2 -mcpu=ultrasparc -O3 +35-45%
+ *
+ * v8 module on SS10/60MHz against bn_asm.c compiled with:
+ *
+ * cc-5.0 -xarch=v8 -xO5 -xdepend +7-10%
+ * cc-4.2 -xarch=v8 -xO5 -xdepend +10%
+ * egcs-1.1.2 -mv8 -O3 +35-45%
+ *
+ * As you can see it's damn hard to beat the new Sun C compiler
+ * and it's in first place GNU C users who will appreciate this
+ * assembler implementation:-)
+ */
+
+/*
+ * Revision history.
+ *
+ * 1.0 - initial release;
+ * 1.1 - new loop unrolling model(*);
+ * - some more fine tuning;
+ * 1.2 - made gas friendly;
+ * - updates to documentation concerning v9;
+ * - new performance comparison matrix;
+ * 1.3 - fixed problem with /usr/ccs/lib/cpp;
+ * 1.4 - native V9 bn_*_comba[48] implementation (15% more efficient)
+ * resulting in slight overall performance kick;
+ * - some retunes;
+ * - support for GNU as added;
+ *
+ * (*) Originally unrolled loop looked like this:
+ * for (;;) {
+ * op(p+0); if (--n==0) break;
+ * op(p+1); if (--n==0) break;
+ * op(p+2); if (--n==0) break;
+ * op(p+3); if (--n==0) break;
+ * p+=4;
+ * }
+ * I unroll according to following:
+ * while (n&~3) {
+ * op(p+0); op(p+1); op(p+2); op(p+3);
+ * p+=4; n=-4;
+ * }
+ * if (n) {
+ * op(p+0); if (--n==0) return;
+ * op(p+2); if (--n==0) return;
+ * op(p+3); return;
+ * }
+ */
+
+#if defined(__SUNPRO_C) && defined(__sparcv9)
+ /* They've said -xarch=v9 at command line */
+ .register %g2,#scratch
+ .register %g3,#scratch
+# define FRAME_SIZE -192
+#elif defined(__GNUC__) && defined(__arch64__)
+ /* They've said -m64 at command line */
+ .register %g2,#scratch
+ .register %g3,#scratch
+# define FRAME_SIZE -192
+#else
+# define FRAME_SIZE -96
+#endif
+/*
+ * GNU assembler can't stand stuw:-(
+ */
+#define stuw st
+
+.section ".text",#alloc,#execinstr
+.file "bn_asm.sparc.v8plus.S"
+
+.align 32
+
+.global bn_mul_add_words
+/*
+ * BN_ULONG bn_mul_add_words(rp,ap,num,w)
+ * BN_ULONG *rp,*ap;
+ * int num;
+ * BN_ULONG w;
+ */
+bn_mul_add_words:
+ sra %o2,%g0,%o2 ! signx %o2
+ brgz,a %o2,.L_bn_mul_add_words_proceed
+ lduw [%o1],%g2
+ retl
+ clr %o0
+ nop
+ nop
+ nop
+
+.L_bn_mul_add_words_proceed:
+ srl %o3,%g0,%o3 ! clruw %o3
+ andcc %o2,-4,%g0
+ bz,pn %icc,.L_bn_mul_add_words_tail
+ clr %o5
+
+.L_bn_mul_add_words_loop: ! wow! 32 aligned!
+ lduw [%o0],%g1
+ lduw [%o1+4],%g3
+ mulx %o3,%g2,%g2
+ add %g1,%o5,%o4
+ nop
+ add %o4,%g2,%o4
+ stuw %o4,[%o0]
+ srlx %o4,32,%o5
+
+ lduw [%o0+4],%g1
+ lduw [%o1+8],%g2
+ mulx %o3,%g3,%g3
+ add %g1,%o5,%o4
+ dec 4,%o2
+ add %o4,%g3,%o4
+ stuw %o4,[%o0+4]
+ srlx %o4,32,%o5
+
+ lduw [%o0+8],%g1
+ lduw [%o1+12],%g3
+ mulx %o3,%g2,%g2
+ add %g1,%o5,%o4
+ inc 16,%o1
+ add %o4,%g2,%o4
+ stuw %o4,[%o0+8]
+ srlx %o4,32,%o5
+
+ lduw [%o0+12],%g1
+ mulx %o3,%g3,%g3
+ add %g1,%o5,%o4
+ inc 16,%o0
+ add %o4,%g3,%o4
+ andcc %o2,-4,%g0
+ stuw %o4,[%o0-4]
+ srlx %o4,32,%o5
+ bnz,a,pt %icc,.L_bn_mul_add_words_loop
+ lduw [%o1],%g2
+
+ brnz,a,pn %o2,.L_bn_mul_add_words_tail
+ lduw [%o1],%g2
+.L_bn_mul_add_words_return:
+ retl
+ mov %o5,%o0
+
+.L_bn_mul_add_words_tail:
+ lduw [%o0],%g1
+ mulx %o3,%g2,%g2
+ add %g1,%o5,%o4
+ dec %o2
+ add %o4,%g2,%o4
+ srlx %o4,32,%o5
+ brz,pt %o2,.L_bn_mul_add_words_return
+ stuw %o4,[%o0]
+
+ lduw [%o1+4],%g2
+ lduw [%o0+4],%g1
+ mulx %o3,%g2,%g2
+ add %g1,%o5,%o4
+ dec %o2
+ add %o4,%g2,%o4
+ srlx %o4,32,%o5
+ brz,pt %o2,.L_bn_mul_add_words_return
+ stuw %o4,[%o0+4]
+
+ lduw [%o1+8],%g2
+ lduw [%o0+8],%g1
+ mulx %o3,%g2,%g2
+ add %g1,%o5,%o4
+ add %o4,%g2,%o4
+ stuw %o4,[%o0+8]
+ retl
+ srlx %o4,32,%o0
+
+.type bn_mul_add_words,#function
+.size bn_mul_add_words,(.-bn_mul_add_words)
+
+.align 32
+
+.global bn_mul_words
+/*
+ * BN_ULONG bn_mul_words(rp,ap,num,w)
+ * BN_ULONG *rp,*ap;
+ * int num;
+ * BN_ULONG w;
+ */
+bn_mul_words:
+ sra %o2,%g0,%o2 ! signx %o2
+ brgz,a %o2,.L_bn_mul_words_proceed
+ lduw [%o1],%g2
+ retl
+ clr %o0
+ nop
+ nop
+ nop
+
+.L_bn_mul_words_proceed:
+ srl %o3,%g0,%o3 ! clruw %o3
+ andcc %o2,-4,%g0
+ bz,pn %icc,.L_bn_mul_words_tail
+ clr %o5
+
+.L_bn_mul_words_loop: ! wow! 32 aligned!
+ lduw [%o1+4],%g3
+ mulx %o3,%g2,%g2
+ add %g2,%o5,%o4
+ nop
+ stuw %o4,[%o0]
+ srlx %o4,32,%o5
+
+ lduw [%o1+8],%g2
+ mulx %o3,%g3,%g3
+ add %g3,%o5,%o4
+ dec 4,%o2
+ stuw %o4,[%o0+4]
+ srlx %o4,32,%o5
+
+ lduw [%o1+12],%g3
+ mulx %o3,%g2,%g2
+ add %g2,%o5,%o4
+ inc 16,%o1
+ stuw %o4,[%o0+8]
+ srlx %o4,32,%o5
+
+ mulx %o3,%g3,%g3
+ add %g3,%o5,%o4
+ inc 16,%o0
+ stuw %o4,[%o0-4]
+ srlx %o4,32,%o5
+ andcc %o2,-4,%g0
+ bnz,a,pt %icc,.L_bn_mul_words_loop
+ lduw [%o1],%g2
+ nop
+ nop
+
+ brnz,a,pn %o2,.L_bn_mul_words_tail
+ lduw [%o1],%g2
+.L_bn_mul_words_return:
+ retl
+ mov %o5,%o0
+
+.L_bn_mul_words_tail:
+ mulx %o3,%g2,%g2
+ add %g2,%o5,%o4
+ dec %o2
+ srlx %o4,32,%o5
+ brz,pt %o2,.L_bn_mul_words_return
+ stuw %o4,[%o0]
+
+ lduw [%o1+4],%g2
+ mulx %o3,%g2,%g2
+ add %g2,%o5,%o4
+ dec %o2
+ srlx %o4,32,%o5
+ brz,pt %o2,.L_bn_mul_words_return
+ stuw %o4,[%o0+4]
+
+ lduw [%o1+8],%g2
+ mulx %o3,%g2,%g2
+ add %g2,%o5,%o4
+ stuw %o4,[%o0+8]
+ retl
+ srlx %o4,32,%o0
+
+.type bn_mul_words,#function
+.size bn_mul_words,(.-bn_mul_words)
+
+.align 32
+.global bn_sqr_words
+/*
+ * void bn_sqr_words(r,a,n)
+ * BN_ULONG *r,*a;
+ * int n;
+ */
+bn_sqr_words:
+ sra %o2,%g0,%o2 ! signx %o2
+ brgz,a %o2,.L_bn_sqr_words_proceed
+ lduw [%o1],%g2
+ retl
+ clr %o0
+ nop
+ nop
+ nop
+
+.L_bn_sqr_words_proceed:
+ andcc %o2,-4,%g0
+ nop
+ bz,pn %icc,.L_bn_sqr_words_tail
+ nop
+
+.L_bn_sqr_words_loop: ! wow! 32 aligned!
+ lduw [%o1+4],%g3
+ mulx %g2,%g2,%o4
+ stuw %o4,[%o0]
+ srlx %o4,32,%o5
+ stuw %o5,[%o0+4]
+ nop
+
+ lduw [%o1+8],%g2
+ mulx %g3,%g3,%o4
+ dec 4,%o2
+ stuw %o4,[%o0+8]
+ srlx %o4,32,%o5
+ stuw %o5,[%o0+12]
+
+ lduw [%o1+12],%g3
+ mulx %g2,%g2,%o4
+ srlx %o4,32,%o5
+ stuw %o4,[%o0+16]
+ inc 16,%o1
+ stuw %o5,[%o0+20]
+
+ mulx %g3,%g3,%o4
+ inc 32,%o0
+ stuw %o4,[%o0-8]
+ srlx %o4,32,%o5
+ andcc %o2,-4,%g2
+ stuw %o5,[%o0-4]
+ bnz,a,pt %icc,.L_bn_sqr_words_loop
+ lduw [%o1],%g2
+ nop
+
+ brnz,a,pn %o2,.L_bn_sqr_words_tail
+ lduw [%o1],%g2
+.L_bn_sqr_words_return:
+ retl
+ clr %o0
+
+.L_bn_sqr_words_tail:
+ mulx %g2,%g2,%o4
+ dec %o2
+ stuw %o4,[%o0]
+ srlx %o4,32,%o5
+ brz,pt %o2,.L_bn_sqr_words_return
+ stuw %o5,[%o0+4]
+
+ lduw [%o1+4],%g2
+ mulx %g2,%g2,%o4
+ dec %o2
+ stuw %o4,[%o0+8]
+ srlx %o4,32,%o5
+ brz,pt %o2,.L_bn_sqr_words_return
+ stuw %o5,[%o0+12]
+
+ lduw [%o1+8],%g2
+ mulx %g2,%g2,%o4
+ srlx %o4,32,%o5
+ stuw %o4,[%o0+16]
+ stuw %o5,[%o0+20]
+ retl
+ clr %o0
+
+.type bn_sqr_words,#function
+.size bn_sqr_words,(.-bn_sqr_words)
+
+.align 32
+.global bn_div_words
+/*
+ * BN_ULONG bn_div_words(h,l,d)
+ * BN_ULONG h,l,d;
+ */
+bn_div_words:
+ sllx %o0,32,%o0
+ or %o0,%o1,%o0
+ udivx %o0,%o2,%o0
+ retl
+ srl %o0,%g0,%o0 ! clruw %o0
+
+.type bn_div_words,#function
+.size bn_div_words,(.-bn_div_words)
+
+.align 32
+
+.global bn_add_words
+/*
+ * BN_ULONG bn_add_words(rp,ap,bp,n)
+ * BN_ULONG *rp,*ap,*bp;
+ * int n;
+ */
+bn_add_words:
+ sra %o3,%g0,%o3 ! signx %o3
+ brgz,a %o3,.L_bn_add_words_proceed
+ lduw [%o1],%o4
+ retl
+ clr %o0
+
+.L_bn_add_words_proceed:
+ andcc %o3,-4,%g0
+ bz,pn %icc,.L_bn_add_words_tail
+ addcc %g0,0,%g0 ! clear carry flag
+
+.L_bn_add_words_loop: ! wow! 32 aligned!
+ dec 4,%o3
+ lduw [%o2],%o5
+ lduw [%o1+4],%g1
+ lduw [%o2+4],%g2
+ lduw [%o1+8],%g3
+ lduw [%o2+8],%g4
+ addccc %o5,%o4,%o5
+ stuw %o5,[%o0]
+
+ lduw [%o1+12],%o4
+ lduw [%o2+12],%o5
+ inc 16,%o1
+ addccc %g1,%g2,%g1
+ stuw %g1,[%o0+4]
+
+ inc 16,%o2
+ addccc %g3,%g4,%g3
+ stuw %g3,[%o0+8]
+
+ inc 16,%o0
+ addccc %o5,%o4,%o5
+ stuw %o5,[%o0-4]
+ and %o3,-4,%g1
+ brnz,a,pt %g1,.L_bn_add_words_loop
+ lduw [%o1],%o4
+
+ brnz,a,pn %o3,.L_bn_add_words_tail
+ lduw [%o1],%o4
+.L_bn_add_words_return:
+ clr %o0
+ retl
+ movcs %icc,1,%o0
+ nop
+
+.L_bn_add_words_tail:
+ lduw [%o2],%o5
+ dec %o3
+ addccc %o5,%o4,%o5
+ brz,pt %o3,.L_bn_add_words_return
+ stuw %o5,[%o0]
+
+ lduw [%o1+4],%o4
+ lduw [%o2+4],%o5
+ dec %o3
+ addccc %o5,%o4,%o5
+ brz,pt %o3,.L_bn_add_words_return
+ stuw %o5,[%o0+4]
+
+ lduw [%o1+8],%o4
+ lduw [%o2+8],%o5
+ addccc %o5,%o4,%o5
+ stuw %o5,[%o0+8]
+ clr %o0
+ retl
+ movcs %icc,1,%o0
+
+.type bn_add_words,#function
+.size bn_add_words,(.-bn_add_words)
+
+.global bn_sub_words
+/*
+ * BN_ULONG bn_sub_words(rp,ap,bp,n)
+ * BN_ULONG *rp,*ap,*bp;
+ * int n;
+ */
+bn_sub_words:
+ sra %o3,%g0,%o3 ! signx %o3
+ brgz,a %o3,.L_bn_sub_words_proceed
+ lduw [%o1],%o4
+ retl
+ clr %o0
+
+.L_bn_sub_words_proceed:
+ andcc %o3,-4,%g0
+ bz,pn %icc,.L_bn_sub_words_tail
+ addcc %g0,0,%g0 ! clear carry flag
+
+.L_bn_sub_words_loop: ! wow! 32 aligned!
+ dec 4,%o3
+ lduw [%o2],%o5
+ lduw [%o1+4],%g1
+ lduw [%o2+4],%g2
+ lduw [%o1+8],%g3
+ lduw [%o2+8],%g4
+ subccc %o4,%o5,%o5
+ stuw %o5,[%o0]
+
+ lduw [%o1+12],%o4
+ lduw [%o2+12],%o5
+ inc 16,%o1
+ subccc %g1,%g2,%g2
+ stuw %g2,[%o0+4]
+
+ inc 16,%o2
+ subccc %g3,%g4,%g4
+ stuw %g4,[%o0+8]
+
+ inc 16,%o0
+ subccc %o4,%o5,%o5
+ stuw %o5,[%o0-4]
+ and %o3,-4,%g1
+ brnz,a,pt %g1,.L_bn_sub_words_loop
+ lduw [%o1],%o4
+
+ brnz,a,pn %o3,.L_bn_sub_words_tail
+ lduw [%o1],%o4
+.L_bn_sub_words_return:
+ clr %o0
+ retl
+ movcs %icc,1,%o0
+ nop
+
+.L_bn_sub_words_tail: ! wow! 32 aligned!
+ lduw [%o2],%o5
+ dec %o3
+ subccc %o4,%o5,%o5
+ brz,pt %o3,.L_bn_sub_words_return
+ stuw %o5,[%o0]
+
+ lduw [%o1+4],%o4
+ lduw [%o2+4],%o5
+ dec %o3
+ subccc %o4,%o5,%o5
+ brz,pt %o3,.L_bn_sub_words_return
+ stuw %o5,[%o0+4]
+
+ lduw [%o1+8],%o4
+ lduw [%o2+8],%o5
+ subccc %o4,%o5,%o5
+ stuw %o5,[%o0+8]
+ clr %o0
+ retl
+ movcs %icc,1,%o0
+
+.type bn_sub_words,#function
+.size bn_sub_words,(.-bn_sub_words)
+
+/*
+ * Code below depends on the fact that upper parts of the %l0-%l7
+ * and %i0-%i7 are zeroed by kernel after context switch. In
+ * previous versions this comment stated that "the trouble is that
+ * it's not feasible to implement the mumbo-jumbo in less V9
+ * instructions:-(" which apparently isn't true thanks to
+ * 'bcs,a %xcc,.+8; inc %rd' pair. But the performance improvement
+ * results not from the shorter code, but from elimination of
+ * multicycle none-pairable 'rd %y,%rd' instructions.
+ *
+ * Andy.
+ */
+
+/*
+ * Here is register usage map for *all* routines below.
+ */
+#define t_1 %o0
+#define t_2 %o1
+#define c_12 %o2
+#define c_3 %o3
+
+#define ap(I) [%i1+4*I]
+#define bp(I) [%i2+4*I]
+#define rp(I) [%i0+4*I]
+
+#define a_0 %l0
+#define a_1 %l1
+#define a_2 %l2
+#define a_3 %l3
+#define a_4 %l4
+#define a_5 %l5
+#define a_6 %l6
+#define a_7 %l7
+
+#define b_0 %i3
+#define b_1 %i4
+#define b_2 %i5
+#define b_3 %o4
+#define b_4 %o5
+#define b_5 %o7
+#define b_6 %g1
+#define b_7 %g4
+
+.align 32
+.global bn_mul_comba8
+/*
+ * void bn_mul_comba8(r,a,b)
+ * BN_ULONG *r,*a,*b;
+ */
+bn_mul_comba8:
+ save %sp,FRAME_SIZE,%sp
+ mov 1,t_2
+ lduw ap(0),a_0
+ sllx t_2,32,t_2
+ lduw bp(0),b_0 !=
+ lduw bp(1),b_1
+ mulx a_0,b_0,t_1 !mul_add_c(a[0],b[0],c1,c2,c3);
+ srlx t_1,32,c_12
+ stuw t_1,rp(0) !=!r[0]=c1;
+
+ lduw ap(1),a_1
+ mulx a_0,b_1,t_1 !mul_add_c(a[0],b[1],c2,c3,c1);
+ addcc c_12,t_1,c_12
+ clr c_3 !=
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ lduw ap(2),a_2
+ mulx a_1,b_0,t_1 !=!mul_add_c(a[1],b[0],c2,c3,c1);
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12 !=
+ stuw t_1,rp(1) !r[1]=c2;
+ or c_12,c_3,c_12
+
+ mulx a_2,b_0,t_1 !mul_add_c(a[2],b[0],c3,c1,c2);
+ addcc c_12,t_1,c_12 !=
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ lduw bp(2),b_2 !=
+ mulx a_1,b_1,t_1 !mul_add_c(a[1],b[1],c3,c1,c2);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3 !=
+ lduw bp(3),b_3
+ mulx a_0,b_2,t_1 !mul_add_c(a[0],b[2],c3,c1,c2);
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(2) !r[2]=c3;
+ or c_12,c_3,c_12 !=
+
+ mulx a_0,b_3,t_1 !mul_add_c(a[0],b[3],c1,c2,c3);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_1,b_2,t_1 !=!mul_add_c(a[1],b[2],c1,c2,c3);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ lduw ap(3),a_3
+ mulx a_2,b_1,t_1 !mul_add_c(a[2],b[1],c1,c2,c3);
+ addcc c_12,t_1,c_12 !=
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ lduw ap(4),a_4
+ mulx a_3,b_0,t_1 !=!mul_add_c(a[3],b[0],c1,c2,c3);!=
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12 !=
+ stuw t_1,rp(3) !r[3]=c1;
+ or c_12,c_3,c_12
+
+ mulx a_4,b_0,t_1 !mul_add_c(a[4],b[0],c2,c3,c1);
+ addcc c_12,t_1,c_12 !=
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_3,b_1,t_1 !=!mul_add_c(a[3],b[1],c2,c3,c1);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_2,b_2,t_1 !=!mul_add_c(a[2],b[2],c2,c3,c1);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ lduw bp(4),b_4 !=
+ mulx a_1,b_3,t_1 !mul_add_c(a[1],b[3],c2,c3,c1);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3 !=
+ lduw bp(5),b_5
+ mulx a_0,b_4,t_1 !mul_add_c(a[0],b[4],c2,c3,c1);
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(4) !r[4]=c2;
+ or c_12,c_3,c_12 !=
+
+ mulx a_0,b_5,t_1 !mul_add_c(a[0],b[5],c3,c1,c2);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_1,b_4,t_1 !mul_add_c(a[1],b[4],c3,c1,c2);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_2,b_3,t_1 !mul_add_c(a[2],b[3],c3,c1,c2);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_3,b_2,t_1 !mul_add_c(a[3],b[2],c3,c1,c2);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ lduw ap(5),a_5
+ mulx a_4,b_1,t_1 !mul_add_c(a[4],b[1],c3,c1,c2);
+ addcc c_12,t_1,c_12 !=
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ lduw ap(6),a_6
+ mulx a_5,b_0,t_1 !=!mul_add_c(a[5],b[0],c3,c1,c2);
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12 !=
+ stuw t_1,rp(5) !r[5]=c3;
+ or c_12,c_3,c_12
+
+ mulx a_6,b_0,t_1 !mul_add_c(a[6],b[0],c1,c2,c3);
+ addcc c_12,t_1,c_12 !=
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_5,b_1,t_1 !=!mul_add_c(a[5],b[1],c1,c2,c3);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_4,b_2,t_1 !=!mul_add_c(a[4],b[2],c1,c2,c3);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_3,b_3,t_1 !=!mul_add_c(a[3],b[3],c1,c2,c3);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_2,b_4,t_1 !=!mul_add_c(a[2],b[4],c1,c2,c3);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ lduw bp(6),b_6 !=
+ mulx a_1,b_5,t_1 !mul_add_c(a[1],b[5],c1,c2,c3);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3 !=
+ lduw bp(7),b_7
+ mulx a_0,b_6,t_1 !mul_add_c(a[0],b[6],c1,c2,c3);
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(6) !r[6]=c1;
+ or c_12,c_3,c_12 !=
+
+ mulx a_0,b_7,t_1 !mul_add_c(a[0],b[7],c2,c3,c1);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_1,b_6,t_1 !mul_add_c(a[1],b[6],c2,c3,c1);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_2,b_5,t_1 !mul_add_c(a[2],b[5],c2,c3,c1);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_3,b_4,t_1 !mul_add_c(a[3],b[4],c2,c3,c1);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_4,b_3,t_1 !mul_add_c(a[4],b[3],c2,c3,c1);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_5,b_2,t_1 !mul_add_c(a[5],b[2],c2,c3,c1);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ lduw ap(7),a_7
+ mulx a_6,b_1,t_1 !=!mul_add_c(a[6],b[1],c2,c3,c1);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_7,b_0,t_1 !=!mul_add_c(a[7],b[0],c2,c3,c1);
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12 !=
+ stuw t_1,rp(7) !r[7]=c2;
+ or c_12,c_3,c_12
+
+ mulx a_7,b_1,t_1 !=!mul_add_c(a[7],b[1],c3,c1,c2);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3 !=
+ mulx a_6,b_2,t_1 !mul_add_c(a[6],b[2],c3,c1,c2);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3 !=
+ mulx a_5,b_3,t_1 !mul_add_c(a[5],b[3],c3,c1,c2);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3 !=
+ mulx a_4,b_4,t_1 !mul_add_c(a[4],b[4],c3,c1,c2);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3 !=
+ mulx a_3,b_5,t_1 !mul_add_c(a[3],b[5],c3,c1,c2);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3 !=
+ mulx a_2,b_6,t_1 !mul_add_c(a[2],b[6],c3,c1,c2);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3 !=
+ mulx a_1,b_7,t_1 !mul_add_c(a[1],b[7],c3,c1,c2);
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3 !=
+ srlx t_1,32,c_12
+ stuw t_1,rp(8) !r[8]=c3;
+ or c_12,c_3,c_12
+
+ mulx a_2,b_7,t_1 !=!mul_add_c(a[2],b[7],c1,c2,c3);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3 !=
+ mulx a_3,b_6,t_1 !mul_add_c(a[3],b[6],c1,c2,c3);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_4,b_5,t_1 !mul_add_c(a[4],b[5],c1,c2,c3);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_5,b_4,t_1 !mul_add_c(a[5],b[4],c1,c2,c3);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_6,b_3,t_1 !mul_add_c(a[6],b[3],c1,c2,c3);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_7,b_2,t_1 !mul_add_c(a[7],b[2],c1,c2,c3);
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(9) !r[9]=c1;
+ or c_12,c_3,c_12 !=
+
+ mulx a_7,b_3,t_1 !mul_add_c(a[7],b[3],c2,c3,c1);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_6,b_4,t_1 !mul_add_c(a[6],b[4],c2,c3,c1);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_5,b_5,t_1 !mul_add_c(a[5],b[5],c2,c3,c1);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_4,b_6,t_1 !mul_add_c(a[4],b[6],c2,c3,c1);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_3,b_7,t_1 !mul_add_c(a[3],b[7],c2,c3,c1);
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(10) !r[10]=c2;
+ or c_12,c_3,c_12 !=
+
+ mulx a_4,b_7,t_1 !mul_add_c(a[4],b[7],c3,c1,c2);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_5,b_6,t_1 !mul_add_c(a[5],b[6],c3,c1,c2);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_6,b_5,t_1 !mul_add_c(a[6],b[5],c3,c1,c2);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_7,b_4,t_1 !mul_add_c(a[7],b[4],c3,c1,c2);
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(11) !r[11]=c3;
+ or c_12,c_3,c_12 !=
+
+ mulx a_7,b_5,t_1 !mul_add_c(a[7],b[5],c1,c2,c3);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_6,b_6,t_1 !mul_add_c(a[6],b[6],c1,c2,c3);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_5,b_7,t_1 !mul_add_c(a[5],b[7],c1,c2,c3);
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(12) !r[12]=c1;
+ or c_12,c_3,c_12 !=
+
+ mulx a_6,b_7,t_1 !mul_add_c(a[6],b[7],c2,c3,c1);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_7,b_6,t_1 !mul_add_c(a[7],b[6],c2,c3,c1);
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ st t_1,rp(13) !r[13]=c2;
+ or c_12,c_3,c_12 !=
+
+ mulx a_7,b_7,t_1 !mul_add_c(a[7],b[7],c3,c1,c2);
+ addcc c_12,t_1,t_1
+ srlx t_1,32,c_12 !=
+ stuw t_1,rp(14) !r[14]=c3;
+ stuw c_12,rp(15) !r[15]=c1;
+
+ ret
+ restore %g0,%g0,%o0 !=
+
+.type bn_mul_comba8,#function
+.size bn_mul_comba8,(.-bn_mul_comba8)
+
+.align 32
+
+.global bn_mul_comba4
+/*
+ * void bn_mul_comba4(r,a,b)
+ * BN_ULONG *r,*a,*b;
+ */
+bn_mul_comba4:
+ save %sp,FRAME_SIZE,%sp
+ lduw ap(0),a_0
+ mov 1,t_2
+ lduw bp(0),b_0
+ sllx t_2,32,t_2 !=
+ lduw bp(1),b_1
+ mulx a_0,b_0,t_1 !mul_add_c(a[0],b[0],c1,c2,c3);
+ srlx t_1,32,c_12
+ stuw t_1,rp(0) !=!r[0]=c1;
+
+ lduw ap(1),a_1
+ mulx a_0,b_1,t_1 !mul_add_c(a[0],b[1],c2,c3,c1);
+ addcc c_12,t_1,c_12
+ clr c_3 !=
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ lduw ap(2),a_2
+ mulx a_1,b_0,t_1 !=!mul_add_c(a[1],b[0],c2,c3,c1);
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12 !=
+ stuw t_1,rp(1) !r[1]=c2;
+ or c_12,c_3,c_12
+
+ mulx a_2,b_0,t_1 !mul_add_c(a[2],b[0],c3,c1,c2);
+ addcc c_12,t_1,c_12 !=
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ lduw bp(2),b_2 !=
+ mulx a_1,b_1,t_1 !mul_add_c(a[1],b[1],c3,c1,c2);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3 !=
+ lduw bp(3),b_3
+ mulx a_0,b_2,t_1 !mul_add_c(a[0],b[2],c3,c1,c2);
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(2) !r[2]=c3;
+ or c_12,c_3,c_12 !=
+
+ mulx a_0,b_3,t_1 !mul_add_c(a[0],b[3],c1,c2,c3);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ mulx a_1,b_2,t_1 !mul_add_c(a[1],b[2],c1,c2,c3);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8 !=
+ add c_3,t_2,c_3
+ lduw ap(3),a_3
+ mulx a_2,b_1,t_1 !mul_add_c(a[2],b[1],c1,c2,c3);
+ addcc c_12,t_1,c_12 !=
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_3,b_0,t_1 !mul_add_c(a[3],b[0],c1,c2,c3);!=
+ addcc c_12,t_1,t_1 !=
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(3) !=!r[3]=c1;
+ or c_12,c_3,c_12
+
+ mulx a_3,b_1,t_1 !mul_add_c(a[3],b[1],c2,c3,c1);
+ addcc c_12,t_1,c_12
+ clr c_3 !=
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_2,b_2,t_1 !mul_add_c(a[2],b[2],c2,c3,c1);
+ addcc c_12,t_1,c_12 !=
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_1,b_3,t_1 !mul_add_c(a[1],b[3],c2,c3,c1);
+ addcc c_12,t_1,t_1 !=
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(4) !=!r[4]=c2;
+ or c_12,c_3,c_12
+
+ mulx a_2,b_3,t_1 !mul_add_c(a[2],b[3],c3,c1,c2);
+ addcc c_12,t_1,c_12
+ clr c_3 !=
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_3,b_2,t_1 !mul_add_c(a[3],b[2],c3,c1,c2);
+ addcc c_12,t_1,t_1 !=
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(5) !=!r[5]=c3;
+ or c_12,c_3,c_12
+
+ mulx a_3,b_3,t_1 !mul_add_c(a[3],b[3],c1,c2,c3);
+ addcc c_12,t_1,t_1
+ srlx t_1,32,c_12 !=
+ stuw t_1,rp(6) !r[6]=c1;
+ stuw c_12,rp(7) !r[7]=c2;
+
+ ret
+ restore %g0,%g0,%o0
+
+.type bn_mul_comba4,#function
+.size bn_mul_comba4,(.-bn_mul_comba4)
+
+.align 32
+
+.global bn_sqr_comba8
+bn_sqr_comba8:
+ save %sp,FRAME_SIZE,%sp
+ mov 1,t_2
+ lduw ap(0),a_0
+ sllx t_2,32,t_2
+ lduw ap(1),a_1
+ mulx a_0,a_0,t_1 !sqr_add_c(a,0,c1,c2,c3);
+ srlx t_1,32,c_12
+ stuw t_1,rp(0) !r[0]=c1;
+
+ lduw ap(2),a_2
+ mulx a_0,a_1,t_1 !=!sqr_add_c2(a,1,0,c2,c3,c1);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(1) !r[1]=c2;
+ or c_12,c_3,c_12
+
+ mulx a_2,a_0,t_1 !sqr_add_c2(a,2,0,c3,c1,c2);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ lduw ap(3),a_3
+ mulx a_1,a_1,t_1 !sqr_add_c(a,1,c3,c1,c2);
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(2) !r[2]=c3;
+ or c_12,c_3,c_12
+
+ mulx a_0,a_3,t_1 !sqr_add_c2(a,3,0,c1,c2,c3);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ lduw ap(4),a_4
+ mulx a_1,a_2,t_1 !sqr_add_c2(a,2,1,c1,c2,c3);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ st t_1,rp(3) !r[3]=c1;
+ or c_12,c_3,c_12
+
+ mulx a_4,a_0,t_1 !sqr_add_c2(a,4,0,c2,c3,c1);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_3,a_1,t_1 !sqr_add_c2(a,3,1,c2,c3,c1);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ lduw ap(5),a_5
+ mulx a_2,a_2,t_1 !sqr_add_c(a,2,c2,c3,c1);
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(4) !r[4]=c2;
+ or c_12,c_3,c_12
+
+ mulx a_0,a_5,t_1 !sqr_add_c2(a,5,0,c3,c1,c2);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_1,a_4,t_1 !sqr_add_c2(a,4,1,c3,c1,c2);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ lduw ap(6),a_6
+ mulx a_2,a_3,t_1 !sqr_add_c2(a,3,2,c3,c1,c2);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(5) !r[5]=c3;
+ or c_12,c_3,c_12
+
+ mulx a_6,a_0,t_1 !sqr_add_c2(a,6,0,c1,c2,c3);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_5,a_1,t_1 !sqr_add_c2(a,5,1,c1,c2,c3);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_4,a_2,t_1 !sqr_add_c2(a,4,2,c1,c2,c3);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ lduw ap(7),a_7
+ mulx a_3,a_3,t_1 !=!sqr_add_c(a,3,c1,c2,c3);
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(6) !r[6]=c1;
+ or c_12,c_3,c_12
+
+ mulx a_0,a_7,t_1 !sqr_add_c2(a,7,0,c2,c3,c1);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_1,a_6,t_1 !sqr_add_c2(a,6,1,c2,c3,c1);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_2,a_5,t_1 !sqr_add_c2(a,5,2,c2,c3,c1);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_3,a_4,t_1 !sqr_add_c2(a,4,3,c2,c3,c1);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(7) !r[7]=c2;
+ or c_12,c_3,c_12
+
+ mulx a_7,a_1,t_1 !sqr_add_c2(a,7,1,c3,c1,c2);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_6,a_2,t_1 !sqr_add_c2(a,6,2,c3,c1,c2);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_5,a_3,t_1 !sqr_add_c2(a,5,3,c3,c1,c2);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_4,a_4,t_1 !sqr_add_c(a,4,c3,c1,c2);
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(8) !r[8]=c3;
+ or c_12,c_3,c_12
+
+ mulx a_2,a_7,t_1 !sqr_add_c2(a,7,2,c1,c2,c3);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_3,a_6,t_1 !sqr_add_c2(a,6,3,c1,c2,c3);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_4,a_5,t_1 !sqr_add_c2(a,5,4,c1,c2,c3);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(9) !r[9]=c1;
+ or c_12,c_3,c_12
+
+ mulx a_7,a_3,t_1 !sqr_add_c2(a,7,3,c2,c3,c1);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_6,a_4,t_1 !sqr_add_c2(a,6,4,c2,c3,c1);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_5,a_5,t_1 !sqr_add_c(a,5,c2,c3,c1);
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(10) !r[10]=c2;
+ or c_12,c_3,c_12
+
+ mulx a_4,a_7,t_1 !sqr_add_c2(a,7,4,c3,c1,c2);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_5,a_6,t_1 !sqr_add_c2(a,6,5,c3,c1,c2);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(11) !r[11]=c3;
+ or c_12,c_3,c_12
+
+ mulx a_7,a_5,t_1 !sqr_add_c2(a,7,5,c1,c2,c3);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_6,a_6,t_1 !sqr_add_c(a,6,c1,c2,c3);
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(12) !r[12]=c1;
+ or c_12,c_3,c_12
+
+ mulx a_6,a_7,t_1 !sqr_add_c2(a,7,6,c2,c3,c1);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(13) !r[13]=c2;
+ or c_12,c_3,c_12
+
+ mulx a_7,a_7,t_1 !sqr_add_c(a,7,c3,c1,c2);
+ addcc c_12,t_1,t_1
+ srlx t_1,32,c_12
+ stuw t_1,rp(14) !r[14]=c3;
+ stuw c_12,rp(15) !r[15]=c1;
+
+ ret
+ restore %g0,%g0,%o0
+
+.type bn_sqr_comba8,#function
+.size bn_sqr_comba8,(.-bn_sqr_comba8)
+
+.align 32
+
+.global bn_sqr_comba4
+/*
+ * void bn_sqr_comba4(r,a)
+ * BN_ULONG *r,*a;
+ */
+bn_sqr_comba4:
+ save %sp,FRAME_SIZE,%sp
+ mov 1,t_2
+ lduw ap(0),a_0
+ sllx t_2,32,t_2
+ lduw ap(1),a_1
+ mulx a_0,a_0,t_1 !sqr_add_c(a,0,c1,c2,c3);
+ srlx t_1,32,c_12
+ stuw t_1,rp(0) !r[0]=c1;
+
+ lduw ap(2),a_2
+ mulx a_0,a_1,t_1 !sqr_add_c2(a,1,0,c2,c3,c1);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(1) !r[1]=c2;
+ or c_12,c_3,c_12
+
+ mulx a_2,a_0,t_1 !sqr_add_c2(a,2,0,c3,c1,c2);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ lduw ap(3),a_3
+ mulx a_1,a_1,t_1 !sqr_add_c(a,1,c3,c1,c2);
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(2) !r[2]=c3;
+ or c_12,c_3,c_12
+
+ mulx a_0,a_3,t_1 !sqr_add_c2(a,3,0,c1,c2,c3);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_1,a_2,t_1 !sqr_add_c2(a,2,1,c1,c2,c3);
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(3) !r[3]=c1;
+ or c_12,c_3,c_12
+
+ mulx a_3,a_1,t_1 !sqr_add_c2(a,3,1,c2,c3,c1);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,c_12
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ mulx a_2,a_2,t_1 !sqr_add_c(a,2,c2,c3,c1);
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(4) !r[4]=c2;
+ or c_12,c_3,c_12
+
+ mulx a_2,a_3,t_1 !sqr_add_c2(a,3,2,c3,c1,c2);
+ addcc c_12,t_1,c_12
+ clr c_3
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ addcc c_12,t_1,t_1
+ bcs,a %xcc,.+8
+ add c_3,t_2,c_3
+ srlx t_1,32,c_12
+ stuw t_1,rp(5) !r[5]=c3;
+ or c_12,c_3,c_12
+
+ mulx a_3,a_3,t_1 !sqr_add_c(a,3,c1,c2,c3);
+ addcc c_12,t_1,t_1
+ srlx t_1,32,c_12
+ stuw t_1,rp(6) !r[6]=c1;
+ stuw c_12,rp(7) !r[7]=c2;
+
+ ret
+ restore %g0,%g0,%o0
+
+.type bn_sqr_comba4,#function
+.size bn_sqr_comba4,(.-bn_sqr_comba4)
+
+.align 32
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/sparcv9-gf2m.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/sparcv9-gf2m.pl
new file mode 100644
index 0000000..238a93d
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/sparcv9-gf2m.pl
@@ -0,0 +1,200 @@
+#! /usr/bin/env perl
+# Copyright 2012-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+#
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+#
+# October 2012
+#
+# The module implements bn_GF2m_mul_2x2 polynomial multiplication used
+# in bn_gf2m.c. It's kind of low-hanging mechanical port from C for
+# the time being... Except that it has two code paths: one suitable
+# for all SPARCv9 processors and one for VIS3-capable ones. Former
+# delivers ~25-45% more, more for longer keys, heaviest DH and DSA
+# verify operations on venerable UltraSPARC II. On T4 VIS3 code is
+# ~100-230% faster than gcc-generated code and ~35-90% faster than
+# the pure SPARCv9 code path.
+
+$output = pop;
+open STDOUT,">$output";
+
+$locals=16*8;
+
+$tab="%l0";
+
+@T=("%g2","%g3");
+@i=("%g4","%g5");
+
+($a1,$a2,$a4,$a8,$a12,$a48)=map("%o$_",(0..5));
+($lo,$hi,$b)=("%g1",$a8,"%o7"); $a=$lo;
+
+$code.=<<___;
+#include <sparc_arch.h>
+
+#ifdef __arch64__
+.register %g2,#scratch
+.register %g3,#scratch
+#endif
+
+#ifdef __PIC__
+SPARC_PIC_THUNK(%g1)
+#endif
+
+.globl bn_GF2m_mul_2x2
+.align 16
+bn_GF2m_mul_2x2:
+ SPARC_LOAD_ADDRESS_LEAF(OPENSSL_sparcv9cap_P,%g1,%g5)
+ ld [%g1+0],%g1 ! OPENSSL_sparcv9cap_P[0]
+
+ andcc %g1, SPARCV9_VIS3, %g0
+ bz,pn %icc,.Lsoftware
+ nop
+
+ sllx %o1, 32, %o1
+ sllx %o3, 32, %o3
+ or %o2, %o1, %o1
+ or %o4, %o3, %o3
+ .word 0x95b262ab ! xmulx %o1, %o3, %o2
+ .word 0x99b262cb ! xmulxhi %o1, %o3, %o4
+ srlx %o2, 32, %o1 ! 13 cycles later
+ st %o2, [%o0+0]
+ st %o1, [%o0+4]
+ srlx %o4, 32, %o3
+ st %o4, [%o0+8]
+ retl
+ st %o3, [%o0+12]
+
+.align 16
+.Lsoftware:
+ save %sp,-STACK_FRAME-$locals,%sp
+
+ sllx %i1,32,$a
+ mov -1,$a12
+ sllx %i3,32,$b
+ or %i2,$a,$a
+ srlx $a12,1,$a48 ! 0x7fff...
+ or %i4,$b,$b
+ srlx $a12,2,$a12 ! 0x3fff...
+ add %sp,STACK_BIAS+STACK_FRAME,$tab
+
+ sllx $a,2,$a4
+ mov $a,$a1
+ sllx $a,1,$a2
+
+ srax $a4,63,@i[1] ! broadcast 61st bit
+ and $a48,$a4,$a4 ! (a<<2)&0x7fff...
+ srlx $a48,2,$a48
+ srax $a2,63,@i[0] ! broadcast 62nd bit
+ and $a12,$a2,$a2 ! (a<<1)&0x3fff...
+ srax $a1,63,$lo ! broadcast 63rd bit
+ and $a48,$a1,$a1 ! (a<<0)&0x1fff...
+
+ sllx $a1,3,$a8
+ and $b,$lo,$lo
+ and $b,@i[0],@i[0]
+ and $b,@i[1],@i[1]
+
+ stx %g0,[$tab+0*8] ! tab[0]=0
+ xor $a1,$a2,$a12
+ stx $a1,[$tab+1*8] ! tab[1]=a1
+ stx $a2,[$tab+2*8] ! tab[2]=a2
+ xor $a4,$a8,$a48
+ stx $a12,[$tab+3*8] ! tab[3]=a1^a2
+ xor $a4,$a1,$a1
+
+ stx $a4,[$tab+4*8] ! tab[4]=a4
+ xor $a4,$a2,$a2
+ stx $a1,[$tab+5*8] ! tab[5]=a1^a4
+ xor $a4,$a12,$a12
+ stx $a2,[$tab+6*8] ! tab[6]=a2^a4
+ xor $a48,$a1,$a1
+ stx $a12,[$tab+7*8] ! tab[7]=a1^a2^a4
+ xor $a48,$a2,$a2
+
+ stx $a8,[$tab+8*8] ! tab[8]=a8
+ xor $a48,$a12,$a12
+ stx $a1,[$tab+9*8] ! tab[9]=a1^a8
+ xor $a4,$a1,$a1
+ stx $a2,[$tab+10*8] ! tab[10]=a2^a8
+ xor $a4,$a2,$a2
+ stx $a12,[$tab+11*8] ! tab[11]=a1^a2^a8
+
+ xor $a4,$a12,$a12
+ stx $a48,[$tab+12*8] ! tab[12]=a4^a8
+ srlx $lo,1,$hi
+ stx $a1,[$tab+13*8] ! tab[13]=a1^a4^a8
+ sllx $lo,63,$lo
+ stx $a2,[$tab+14*8] ! tab[14]=a2^a4^a8
+ srlx @i[0],2,@T[0]
+ stx $a12,[$tab+15*8] ! tab[15]=a1^a2^a4^a8
+
+ sllx @i[0],62,$a1
+ sllx $b,3,@i[0]
+ srlx @i[1],3,@T[1]
+ and @i[0],`0xf<<3`,@i[0]
+ sllx @i[1],61,$a2
+ ldx [$tab+@i[0]],@i[0]
+ srlx $b,4-3,@i[1]
+ xor @T[0],$hi,$hi
+ and @i[1],`0xf<<3`,@i[1]
+ xor $a1,$lo,$lo
+ ldx [$tab+@i[1]],@i[1]
+ xor @T[1],$hi,$hi
+
+ xor @i[0],$lo,$lo
+ srlx $b,8-3,@i[0]
+ xor $a2,$lo,$lo
+ and @i[0],`0xf<<3`,@i[0]
+___
+for($n=1;$n<14;$n++) {
+$code.=<<___;
+ sllx @i[1],`$n*4`,@T[0]
+ ldx [$tab+@i[0]],@i[0]
+ srlx @i[1],`64-$n*4`,@T[1]
+ xor @T[0],$lo,$lo
+ srlx $b,`($n+2)*4`-3,@i[1]
+ xor @T[1],$hi,$hi
+ and @i[1],`0xf<<3`,@i[1]
+___
+ push(@i,shift(@i)); push(@T,shift(@T));
+}
+$code.=<<___;
+ sllx @i[1],`$n*4`,@T[0]
+ ldx [$tab+@i[0]],@i[0]
+ srlx @i[1],`64-$n*4`,@T[1]
+ xor @T[0],$lo,$lo
+
+ sllx @i[0],`($n+1)*4`,@T[0]
+ xor @T[1],$hi,$hi
+ srlx @i[0],`64-($n+1)*4`,@T[1]
+ xor @T[0],$lo,$lo
+ xor @T[1],$hi,$hi
+
+ srlx $lo,32,%i1
+ st $lo,[%i0+0]
+ st %i1,[%i0+4]
+ srlx $hi,32,%i2
+ st $hi,[%i0+8]
+ st %i2,[%i0+12]
+
+ ret
+ restore
+.type bn_GF2m_mul_2x2,#function
+.size bn_GF2m_mul_2x2,.-bn_GF2m_mul_2x2
+.asciz "GF(2^m) Multiplication for SPARCv9, CRYPTOGAMS by <appro\@openssl.org>"
+.align 4
+___
+
+$code =~ s/\`([^\`]*)\`/eval($1)/gem;
+print $code;
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/sparcv9-mont.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/sparcv9-mont.pl
new file mode 100644
index 0000000..d1a3c2b
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/sparcv9-mont.pl
@@ -0,0 +1,620 @@
+#! /usr/bin/env perl
+# Copyright 2005-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+
+# December 2005
+#
+# Pure SPARCv9/8+ and IALU-only bn_mul_mont implementation. The reasons
+# for undertaken effort are multiple. First of all, UltraSPARC is not
+# the whole SPARCv9 universe and other VIS-free implementations deserve
+# optimized code as much. Secondly, newly introduced UltraSPARC T1,
+# a.k.a. Niagara, has shared FPU and concurrent FPU-intensive paths,
+# such as sparcv9a-mont, will simply sink it. Yes, T1 is equipped with
+# several integrated RSA/DSA accelerator circuits accessible through
+# kernel driver [only(*)], but having decent user-land software
+# implementation is important too. Finally, reasons like desire to
+# experiment with dedicated squaring procedure. Yes, this module
+# implements one, because it was easiest to draft it in SPARCv9
+# instructions...
+
+# (*) Engine accessing the driver in question is on my TODO list.
+# For reference, accelerator is estimated to give 6 to 10 times
+# improvement on single-threaded RSA sign. It should be noted
+# that 6-10x improvement coefficient does not actually mean
+# something extraordinary in terms of absolute [single-threaded]
+# performance, as SPARCv9 instruction set is by all means least
+# suitable for high performance crypto among other 64 bit
+# platforms. 6-10x factor simply places T1 in same performance
+# domain as say AMD64 and IA-64. Improvement of RSA verify don't
+# appear impressive at all, but it's the sign operation which is
+# far more critical/interesting.
+
+# You might notice that inner loops are modulo-scheduled:-) This has
+# essentially negligible impact on UltraSPARC performance, it's
+# Fujitsu SPARC64 V users who should notice and hopefully appreciate
+# the advantage... Currently this module surpasses sparcv9a-mont.pl
+# by ~20% on UltraSPARC-III and later cores, but recall that sparcv9a
+# module still have hidden potential [see TODO list there], which is
+# estimated to be larger than 20%...
+
+$output = pop;
+open STDOUT,">$output";
+
+# int bn_mul_mont(
+$rp="%i0"; # BN_ULONG *rp,
+$ap="%i1"; # const BN_ULONG *ap,
+$bp="%i2"; # const BN_ULONG *bp,
+$np="%i3"; # const BN_ULONG *np,
+$n0="%i4"; # const BN_ULONG *n0,
+$num="%i5"; # int num);
+
+$frame="STACK_FRAME";
+$bias="STACK_BIAS";
+
+$car0="%o0";
+$car1="%o1";
+$car2="%o2"; # 1 bit
+$acc0="%o3";
+$acc1="%o4";
+$mask="%g1"; # 32 bits, what a waste...
+$tmp0="%g4";
+$tmp1="%g5";
+
+$i="%l0";
+$j="%l1";
+$mul0="%l2";
+$mul1="%l3";
+$tp="%l4";
+$apj="%l5";
+$npj="%l6";
+$tpj="%l7";
+
+$fname="bn_mul_mont_int";
+
+$code=<<___;
+#include "sparc_arch.h"
+
+.section ".text",#alloc,#execinstr
+
+.global $fname
+.align 32
+$fname:
+ cmp %o5,4 ! 128 bits minimum
+ bge,pt %icc,.Lenter
+ sethi %hi(0xffffffff),$mask
+ retl
+ clr %o0
+.align 32
+.Lenter:
+ save %sp,-$frame,%sp
+ sll $num,2,$num ! num*=4
+ or $mask,%lo(0xffffffff),$mask
+ ld [$n0],$n0
+ cmp $ap,$bp
+ and $num,$mask,$num
+ ld [$bp],$mul0 ! bp[0]
+ nop
+
+ add %sp,$bias,%o7 ! real top of stack
+ ld [$ap],$car0 ! ap[0] ! redundant in squaring context
+ sub %o7,$num,%o7
+ ld [$ap+4],$apj ! ap[1]
+ and %o7,-1024,%o7
+ ld [$np],$car1 ! np[0]
+ sub %o7,$bias,%sp ! alloca
+ ld [$np+4],$npj ! np[1]
+ be,pt SIZE_T_CC,.Lbn_sqr_mont
+ mov 12,$j
+
+ mulx $car0,$mul0,$car0 ! ap[0]*bp[0]
+ mulx $apj,$mul0,$tmp0 !prologue! ap[1]*bp[0]
+ and $car0,$mask,$acc0
+ add %sp,$bias+$frame,$tp
+ ld [$ap+8],$apj !prologue!
+
+ mulx $n0,$acc0,$mul1 ! "t[0]"*n0
+ and $mul1,$mask,$mul1
+
+ mulx $car1,$mul1,$car1 ! np[0]*"t[0]"*n0
+ mulx $npj,$mul1,$acc1 !prologue! np[1]*"t[0]"*n0
+ srlx $car0,32,$car0
+ add $acc0,$car1,$car1
+ ld [$np+8],$npj !prologue!
+ srlx $car1,32,$car1
+ mov $tmp0,$acc0 !prologue!
+
+.L1st:
+ mulx $apj,$mul0,$tmp0
+ mulx $npj,$mul1,$tmp1
+ add $acc0,$car0,$car0
+ ld [$ap+$j],$apj ! ap[j]
+ and $car0,$mask,$acc0
+ add $acc1,$car1,$car1
+ ld [$np+$j],$npj ! np[j]
+ srlx $car0,32,$car0
+ add $acc0,$car1,$car1
+ add $j,4,$j ! j++
+ mov $tmp0,$acc0
+ st $car1,[$tp]
+ cmp $j,$num
+ mov $tmp1,$acc1
+ srlx $car1,32,$car1
+ bl %icc,.L1st
+ add $tp,4,$tp ! tp++
+!.L1st
+
+ mulx $apj,$mul0,$tmp0 !epilogue!
+ mulx $npj,$mul1,$tmp1
+ add $acc0,$car0,$car0
+ and $car0,$mask,$acc0
+ add $acc1,$car1,$car1
+ srlx $car0,32,$car0
+ add $acc0,$car1,$car1
+ st $car1,[$tp]
+ srlx $car1,32,$car1
+
+ add $tmp0,$car0,$car0
+ and $car0,$mask,$acc0
+ add $tmp1,$car1,$car1
+ srlx $car0,32,$car0
+ add $acc0,$car1,$car1
+ st $car1,[$tp+4]
+ srlx $car1,32,$car1
+
+ add $car0,$car1,$car1
+ st $car1,[$tp+8]
+ srlx $car1,32,$car2
+
+ mov 4,$i ! i++
+ ld [$bp+4],$mul0 ! bp[1]
+.Louter:
+ add %sp,$bias+$frame,$tp
+ ld [$ap],$car0 ! ap[0]
+ ld [$ap+4],$apj ! ap[1]
+ ld [$np],$car1 ! np[0]
+ ld [$np+4],$npj ! np[1]
+ ld [$tp],$tmp1 ! tp[0]
+ ld [$tp+4],$tpj ! tp[1]
+ mov 12,$j
+
+ mulx $car0,$mul0,$car0
+ mulx $apj,$mul0,$tmp0 !prologue!
+ add $tmp1,$car0,$car0
+ ld [$ap+8],$apj !prologue!
+ and $car0,$mask,$acc0
+
+ mulx $n0,$acc0,$mul1
+ and $mul1,$mask,$mul1
+
+ mulx $car1,$mul1,$car1
+ mulx $npj,$mul1,$acc1 !prologue!
+ srlx $car0,32,$car0
+ add $acc0,$car1,$car1
+ ld [$np+8],$npj !prologue!
+ srlx $car1,32,$car1
+ mov $tmp0,$acc0 !prologue!
+
+.Linner:
+ mulx $apj,$mul0,$tmp0
+ mulx $npj,$mul1,$tmp1
+ add $tpj,$car0,$car0
+ ld [$ap+$j],$apj ! ap[j]
+ add $acc0,$car0,$car0
+ add $acc1,$car1,$car1
+ ld [$np+$j],$npj ! np[j]
+ and $car0,$mask,$acc0
+ ld [$tp+8],$tpj ! tp[j]
+ srlx $car0,32,$car0
+ add $acc0,$car1,$car1
+ add $j,4,$j ! j++
+ mov $tmp0,$acc0
+ st $car1,[$tp] ! tp[j-1]
+ srlx $car1,32,$car1
+ mov $tmp1,$acc1
+ cmp $j,$num
+ bl %icc,.Linner
+ add $tp,4,$tp ! tp++
+!.Linner
+
+ mulx $apj,$mul0,$tmp0 !epilogue!
+ mulx $npj,$mul1,$tmp1
+ add $tpj,$car0,$car0
+ add $acc0,$car0,$car0
+ ld [$tp+8],$tpj ! tp[j]
+ and $car0,$mask,$acc0
+ add $acc1,$car1,$car1
+ srlx $car0,32,$car0
+ add $acc0,$car1,$car1
+ st $car1,[$tp] ! tp[j-1]
+ srlx $car1,32,$car1
+
+ add $tpj,$car0,$car0
+ add $tmp0,$car0,$car0
+ and $car0,$mask,$acc0
+ add $tmp1,$car1,$car1
+ add $acc0,$car1,$car1
+ st $car1,[$tp+4] ! tp[j-1]
+ srlx $car0,32,$car0
+ add $i,4,$i ! i++
+ srlx $car1,32,$car1
+
+ add $car0,$car1,$car1
+ cmp $i,$num
+ add $car2,$car1,$car1
+ st $car1,[$tp+8]
+
+ srlx $car1,32,$car2
+ bl,a %icc,.Louter
+ ld [$bp+$i],$mul0 ! bp[i]
+!.Louter
+
+ add $tp,12,$tp
+
+.Ltail:
+ add $np,$num,$np
+ add $rp,$num,$rp
+ sub %g0,$num,%o7 ! k=-num
+ ba .Lsub
+ subcc %g0,%g0,%g0 ! clear %icc.c
+.align 16
+.Lsub:
+ ld [$tp+%o7],%o0
+ ld [$np+%o7],%o1
+ subccc %o0,%o1,%o1 ! tp[j]-np[j]
+ add $rp,%o7,$i
+ add %o7,4,%o7
+ brnz %o7,.Lsub
+ st %o1,[$i]
+ subccc $car2,0,$car2 ! handle upmost overflow bit
+ sub %g0,$num,%o7
+
+.Lcopy:
+ ld [$tp+%o7],%o1 ! conditional copy
+ ld [$rp+%o7],%o0
+ st %g0,[$tp+%o7] ! zap tp
+ movcs %icc,%o1,%o0
+ st %o0,[$rp+%o7]
+ add %o7,4,%o7
+ brnz %o7,.Lcopy
+ nop
+ mov 1,%i0
+ ret
+ restore
+___
+
+########
+######## .Lbn_sqr_mont gives up to 20% *overall* improvement over
+######## code without following dedicated squaring procedure.
+########
+$sbit="%o5";
+
+$code.=<<___;
+.align 32
+.Lbn_sqr_mont:
+ mulx $mul0,$mul0,$car0 ! ap[0]*ap[0]
+ mulx $apj,$mul0,$tmp0 !prologue!
+ and $car0,$mask,$acc0
+ add %sp,$bias+$frame,$tp
+ ld [$ap+8],$apj !prologue!
+
+ mulx $n0,$acc0,$mul1 ! "t[0]"*n0
+ srlx $car0,32,$car0
+ and $mul1,$mask,$mul1
+
+ mulx $car1,$mul1,$car1 ! np[0]*"t[0]"*n0
+ mulx $npj,$mul1,$acc1 !prologue!
+ and $car0,1,$sbit
+ ld [$np+8],$npj !prologue!
+ srlx $car0,1,$car0
+ add $acc0,$car1,$car1
+ srlx $car1,32,$car1
+ mov $tmp0,$acc0 !prologue!
+
+.Lsqr_1st:
+ mulx $apj,$mul0,$tmp0
+ mulx $npj,$mul1,$tmp1
+ add $acc0,$car0,$car0 ! ap[j]*a0+c0
+ add $acc1,$car1,$car1
+ ld [$ap+$j],$apj ! ap[j]
+ and $car0,$mask,$acc0
+ ld [$np+$j],$npj ! np[j]
+ srlx $car0,32,$car0
+ add $acc0,$acc0,$acc0
+ or $sbit,$acc0,$acc0
+ mov $tmp1,$acc1
+ srlx $acc0,32,$sbit
+ add $j,4,$j ! j++
+ and $acc0,$mask,$acc0
+ cmp $j,$num
+ add $acc0,$car1,$car1
+ st $car1,[$tp]
+ mov $tmp0,$acc0
+ srlx $car1,32,$car1
+ bl %icc,.Lsqr_1st
+ add $tp,4,$tp ! tp++
+!.Lsqr_1st
+
+ mulx $apj,$mul0,$tmp0 ! epilogue
+ mulx $npj,$mul1,$tmp1
+ add $acc0,$car0,$car0 ! ap[j]*a0+c0
+ add $acc1,$car1,$car1
+ and $car0,$mask,$acc0
+ srlx $car0,32,$car0
+ add $acc0,$acc0,$acc0
+ or $sbit,$acc0,$acc0
+ srlx $acc0,32,$sbit
+ and $acc0,$mask,$acc0
+ add $acc0,$car1,$car1
+ st $car1,[$tp]
+ srlx $car1,32,$car1
+
+ add $tmp0,$car0,$car0 ! ap[j]*a0+c0
+ add $tmp1,$car1,$car1
+ and $car0,$mask,$acc0
+ srlx $car0,32,$car0
+ add $acc0,$acc0,$acc0
+ or $sbit,$acc0,$acc0
+ srlx $acc0,32,$sbit
+ and $acc0,$mask,$acc0
+ add $acc0,$car1,$car1
+ st $car1,[$tp+4]
+ srlx $car1,32,$car1
+
+ add $car0,$car0,$car0
+ or $sbit,$car0,$car0
+ add $car0,$car1,$car1
+ st $car1,[$tp+8]
+ srlx $car1,32,$car2
+
+ ld [%sp+$bias+$frame],$tmp0 ! tp[0]
+ ld [%sp+$bias+$frame+4],$tmp1 ! tp[1]
+ ld [%sp+$bias+$frame+8],$tpj ! tp[2]
+ ld [$ap+4],$mul0 ! ap[1]
+ ld [$ap+8],$apj ! ap[2]
+ ld [$np],$car1 ! np[0]
+ ld [$np+4],$npj ! np[1]
+ mulx $n0,$tmp0,$mul1
+
+ mulx $mul0,$mul0,$car0
+ and $mul1,$mask,$mul1
+
+ mulx $car1,$mul1,$car1
+ mulx $npj,$mul1,$acc1
+ add $tmp0,$car1,$car1
+ and $car0,$mask,$acc0
+ ld [$np+8],$npj ! np[2]
+ srlx $car1,32,$car1
+ add $tmp1,$car1,$car1
+ srlx $car0,32,$car0
+ add $acc0,$car1,$car1
+ and $car0,1,$sbit
+ add $acc1,$car1,$car1
+ srlx $car0,1,$car0
+ mov 12,$j
+ st $car1,[%sp+$bias+$frame] ! tp[0]=
+ srlx $car1,32,$car1
+ add %sp,$bias+$frame+4,$tp
+
+.Lsqr_2nd:
+ mulx $apj,$mul0,$acc0
+ mulx $npj,$mul1,$acc1
+ add $acc0,$car0,$car0
+ add $tpj,$sbit,$sbit
+ ld [$ap+$j],$apj ! ap[j]
+ and $car0,$mask,$acc0
+ ld [$np+$j],$npj ! np[j]
+ srlx $car0,32,$car0
+ add $acc1,$car1,$car1
+ ld [$tp+8],$tpj ! tp[j]
+ add $acc0,$acc0,$acc0
+ add $j,4,$j ! j++
+ add $sbit,$acc0,$acc0
+ srlx $acc0,32,$sbit
+ and $acc0,$mask,$acc0
+ cmp $j,$num
+ add $acc0,$car1,$car1
+ st $car1,[$tp] ! tp[j-1]
+ srlx $car1,32,$car1
+ bl %icc,.Lsqr_2nd
+ add $tp,4,$tp ! tp++
+!.Lsqr_2nd
+
+ mulx $apj,$mul0,$acc0
+ mulx $npj,$mul1,$acc1
+ add $acc0,$car0,$car0
+ add $tpj,$sbit,$sbit
+ and $car0,$mask,$acc0
+ srlx $car0,32,$car0
+ add $acc1,$car1,$car1
+ add $acc0,$acc0,$acc0
+ add $sbit,$acc0,$acc0
+ srlx $acc0,32,$sbit
+ and $acc0,$mask,$acc0
+ add $acc0,$car1,$car1
+ st $car1,[$tp] ! tp[j-1]
+ srlx $car1,32,$car1
+
+ add $car0,$car0,$car0
+ add $sbit,$car0,$car0
+ add $car0,$car1,$car1
+ add $car2,$car1,$car1
+ st $car1,[$tp+4]
+ srlx $car1,32,$car2
+
+ ld [%sp+$bias+$frame],$tmp1 ! tp[0]
+ ld [%sp+$bias+$frame+4],$tpj ! tp[1]
+ ld [$ap+8],$mul0 ! ap[2]
+ ld [$np],$car1 ! np[0]
+ ld [$np+4],$npj ! np[1]
+ mulx $n0,$tmp1,$mul1
+ and $mul1,$mask,$mul1
+ mov 8,$i
+
+ mulx $mul0,$mul0,$car0
+ mulx $car1,$mul1,$car1
+ and $car0,$mask,$acc0
+ add $tmp1,$car1,$car1
+ srlx $car0,32,$car0
+ add %sp,$bias+$frame,$tp
+ srlx $car1,32,$car1
+ and $car0,1,$sbit
+ srlx $car0,1,$car0
+ mov 4,$j
+
+.Lsqr_outer:
+.Lsqr_inner1:
+ mulx $npj,$mul1,$acc1
+ add $tpj,$car1,$car1
+ add $j,4,$j
+ ld [$tp+8],$tpj
+ cmp $j,$i
+ add $acc1,$car1,$car1
+ ld [$np+$j],$npj
+ st $car1,[$tp]
+ srlx $car1,32,$car1
+ bl %icc,.Lsqr_inner1
+ add $tp,4,$tp
+!.Lsqr_inner1
+
+ add $j,4,$j
+ ld [$ap+$j],$apj ! ap[j]
+ mulx $npj,$mul1,$acc1
+ add $tpj,$car1,$car1
+ ld [$np+$j],$npj ! np[j]
+ srlx $car1,32,$tmp0
+ and $car1,$mask,$car1
+ add $tmp0,$sbit,$sbit
+ add $acc0,$car1,$car1
+ ld [$tp+8],$tpj ! tp[j]
+ add $acc1,$car1,$car1
+ st $car1,[$tp]
+ srlx $car1,32,$car1
+
+ add $j,4,$j
+ cmp $j,$num
+ be,pn %icc,.Lsqr_no_inner2
+ add $tp,4,$tp
+
+.Lsqr_inner2:
+ mulx $apj,$mul0,$acc0
+ mulx $npj,$mul1,$acc1
+ add $tpj,$sbit,$sbit
+ add $acc0,$car0,$car0
+ ld [$ap+$j],$apj ! ap[j]
+ and $car0,$mask,$acc0
+ ld [$np+$j],$npj ! np[j]
+ srlx $car0,32,$car0
+ add $acc0,$acc0,$acc0
+ ld [$tp+8],$tpj ! tp[j]
+ add $sbit,$acc0,$acc0
+ add $j,4,$j ! j++
+ srlx $acc0,32,$sbit
+ and $acc0,$mask,$acc0
+ cmp $j,$num
+ add $acc0,$car1,$car1
+ add $acc1,$car1,$car1
+ st $car1,[$tp] ! tp[j-1]
+ srlx $car1,32,$car1
+ bl %icc,.Lsqr_inner2
+ add $tp,4,$tp ! tp++
+
+.Lsqr_no_inner2:
+ mulx $apj,$mul0,$acc0
+ mulx $npj,$mul1,$acc1
+ add $tpj,$sbit,$sbit
+ add $acc0,$car0,$car0
+ and $car0,$mask,$acc0
+ srlx $car0,32,$car0
+ add $acc0,$acc0,$acc0
+ add $sbit,$acc0,$acc0
+ srlx $acc0,32,$sbit
+ and $acc0,$mask,$acc0
+ add $acc0,$car1,$car1
+ add $acc1,$car1,$car1
+ st $car1,[$tp] ! tp[j-1]
+ srlx $car1,32,$car1
+
+ add $car0,$car0,$car0
+ add $sbit,$car0,$car0
+ add $car0,$car1,$car1
+ add $car2,$car1,$car1
+ st $car1,[$tp+4]
+ srlx $car1,32,$car2
+
+ add $i,4,$i ! i++
+ ld [%sp+$bias+$frame],$tmp1 ! tp[0]
+ ld [%sp+$bias+$frame+4],$tpj ! tp[1]
+ ld [$ap+$i],$mul0 ! ap[j]
+ ld [$np],$car1 ! np[0]
+ ld [$np+4],$npj ! np[1]
+ mulx $n0,$tmp1,$mul1
+ and $mul1,$mask,$mul1
+ add $i,4,$tmp0
+
+ mulx $mul0,$mul0,$car0
+ mulx $car1,$mul1,$car1
+ and $car0,$mask,$acc0
+ add $tmp1,$car1,$car1
+ srlx $car0,32,$car0
+ add %sp,$bias+$frame,$tp
+ srlx $car1,32,$car1
+ and $car0,1,$sbit
+ srlx $car0,1,$car0
+
+ cmp $tmp0,$num ! i<num-1
+ bl %icc,.Lsqr_outer
+ mov 4,$j
+
+.Lsqr_last:
+ mulx $npj,$mul1,$acc1
+ add $tpj,$car1,$car1
+ add $j,4,$j
+ ld [$tp+8],$tpj
+ cmp $j,$i
+ add $acc1,$car1,$car1
+ ld [$np+$j],$npj
+ st $car1,[$tp]
+ srlx $car1,32,$car1
+ bl %icc,.Lsqr_last
+ add $tp,4,$tp
+!.Lsqr_last
+
+ mulx $npj,$mul1,$acc1
+ add $tpj,$acc0,$acc0
+ srlx $acc0,32,$tmp0
+ and $acc0,$mask,$acc0
+ add $tmp0,$sbit,$sbit
+ add $acc0,$car1,$car1
+ add $acc1,$car1,$car1
+ st $car1,[$tp]
+ srlx $car1,32,$car1
+
+ add $car0,$car0,$car0 ! recover $car0
+ add $sbit,$car0,$car0
+ add $car0,$car1,$car1
+ add $car2,$car1,$car1
+ st $car1,[$tp+4]
+ srlx $car1,32,$car2
+
+ ba .Ltail
+ add $tp,8,$tp
+.type $fname,#function
+.size $fname,(.-$fname)
+.asciz "Montgomery Multiplication for SPARCv9, CRYPTOGAMS by <appro\@openssl.org>"
+.align 32
+___
+$code =~ s/\`([^\`]*)\`/eval($1)/gem;
+print $code;
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/sparcv9a-mont.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/sparcv9a-mont.pl
new file mode 100755
index 0000000..7a1fca1
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/sparcv9a-mont.pl
@@ -0,0 +1,887 @@
+#! /usr/bin/env perl
+# Copyright 2005-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+
+# October 2005
+#
+# "Teaser" Montgomery multiplication module for UltraSPARC. Why FPU?
+# Because unlike integer multiplier, which simply stalls whole CPU,
+# FPU is fully pipelined and can effectively emit 48 bit partial
+# product every cycle. Why not blended SPARC v9? One can argue that
+# making this module dependent on UltraSPARC VIS extension limits its
+# binary compatibility. Well yes, it does exclude SPARC64 prior-V(!)
+# implementations from compatibility matrix. But the rest, whole Sun
+# UltraSPARC family and brand new Fujitsu's SPARC64 V, all support
+# VIS extension instructions used in this module. This is considered
+# good enough to not care about HAL SPARC64 users [if any] who have
+# integer-only pure SPARCv9 module to "fall down" to.
+
+# USI&II cores currently exhibit uniform 2x improvement [over pre-
+# bn_mul_mont codebase] for all key lengths and benchmarks. On USIII
+# performance improves few percents for shorter keys and worsens few
+# percents for longer keys. This is because USIII integer multiplier
+# is >3x faster than USI&II one, which is harder to match [but see
+# TODO list below]. It should also be noted that SPARC64 V features
+# out-of-order execution, which *might* mean that integer multiplier
+# is pipelined, which in turn *might* be impossible to match... On
+# additional note, SPARC64 V implements FP Multiply-Add instruction,
+# which is perfectly usable in this context... In other words, as far
+# as Fujitsu SPARC64 V goes, talk to the author:-)
+
+# The implementation implies following "non-natural" limitations on
+# input arguments:
+# - num may not be less than 4;
+# - num has to be even;
+# Failure to meet either condition has no fatal effects, simply
+# doesn't give any performance gain.
+
+# TODO:
+# - modulo-schedule inner loop for better performance (on in-order
+# execution core such as UltraSPARC this shall result in further
+# noticeable(!) improvement);
+# - dedicated squaring procedure[?];
+
+######################################################################
+# November 2006
+#
+# Modulo-scheduled inner loops allow to interleave floating point and
+# integer instructions and minimize Read-After-Write penalties. This
+# results in *further* 20-50% performance improvement [depending on
+# key length, more for longer keys] on USI&II cores and 30-80% - on
+# USIII&IV.
+
+$output = pop;
+open STDOUT,">$output";
+
+$fname="bn_mul_mont_fpu";
+
+$frame="STACK_FRAME";
+$bias="STACK_BIAS";
+$locals=64;
+
+# In order to provide for 32-/64-bit ABI duality, I keep integers wider
+# than 32 bit in %g1-%g4 and %o0-%o5. %l0-%l7 and %i0-%i5 are used
+# exclusively for pointers, indexes and other small values...
+# int bn_mul_mont(
+$rp="%i0"; # BN_ULONG *rp,
+$ap="%i1"; # const BN_ULONG *ap,
+$bp="%i2"; # const BN_ULONG *bp,
+$np="%i3"; # const BN_ULONG *np,
+$n0="%i4"; # const BN_ULONG *n0,
+$num="%i5"; # int num);
+
+$tp="%l0"; # t[num]
+$ap_l="%l1"; # a[num],n[num] are smashed to 32-bit words and saved
+$ap_h="%l2"; # to these four vectors as double-precision FP values.
+$np_l="%l3"; # This way a bunch of fxtods are eliminated in second
+$np_h="%l4"; # loop and L1-cache aliasing is minimized...
+$i="%l5";
+$j="%l6";
+$mask="%l7"; # 16-bit mask, 0xffff
+
+$n0="%g4"; # reassigned(!) to "64-bit" register
+$carry="%i4"; # %i4 reused(!) for a carry bit
+
+# FP register naming chart
+#
+# ..HILO
+# dcba
+# --------
+# LOa
+# LOb
+# LOc
+# LOd
+# HIa
+# HIb
+# HIc
+# HId
+# ..a
+# ..b
+$ba="%f0"; $bb="%f2"; $bc="%f4"; $bd="%f6";
+$na="%f8"; $nb="%f10"; $nc="%f12"; $nd="%f14";
+$alo="%f16"; $alo_="%f17"; $ahi="%f18"; $ahi_="%f19";
+$nlo="%f20"; $nlo_="%f21"; $nhi="%f22"; $nhi_="%f23";
+
+$dota="%f24"; $dotb="%f26";
+
+$aloa="%f32"; $alob="%f34"; $aloc="%f36"; $alod="%f38";
+$ahia="%f40"; $ahib="%f42"; $ahic="%f44"; $ahid="%f46";
+$nloa="%f48"; $nlob="%f50"; $nloc="%f52"; $nlod="%f54";
+$nhia="%f56"; $nhib="%f58"; $nhic="%f60"; $nhid="%f62";
+
+$ASI_FL16_P=0xD2; # magic ASI value to engage 16-bit FP load
+
+$code=<<___;
+#include "sparc_arch.h"
+
+.section ".text",#alloc,#execinstr
+
+.global $fname
+.align 32
+$fname:
+ save %sp,-$frame-$locals,%sp
+
+ cmp $num,4
+ bl,a,pn %icc,.Lret
+ clr %i0
+ andcc $num,1,%g0 ! $num has to be even...
+ bnz,a,pn %icc,.Lret
+ clr %i0 ! signal "unsupported input value"
+
+ srl $num,1,$num
+ sethi %hi(0xffff),$mask
+ ld [%i4+0],$n0 ! $n0 reassigned, remember?
+ or $mask,%lo(0xffff),$mask
+ ld [%i4+4],%o0
+ sllx %o0,32,%o0
+ or %o0,$n0,$n0 ! $n0=n0[1].n0[0]
+
+ sll $num,3,$num ! num*=8
+
+ add %sp,$bias,%o0 ! real top of stack
+ sll $num,2,%o1
+ add %o1,$num,%o1 ! %o1=num*5
+ sub %o0,%o1,%o0
+ and %o0,-2048,%o0 ! optimize TLB utilization
+ sub %o0,$bias,%sp ! alloca(5*num*8)
+
+ rd %asi,%o7 ! save %asi
+ add %sp,$bias+$frame+$locals,$tp
+ add $tp,$num,$ap_l
+ add $ap_l,$num,$ap_l ! [an]p_[lh] point at the vectors' ends !
+ add $ap_l,$num,$ap_h
+ add $ap_h,$num,$np_l
+ add $np_l,$num,$np_h
+
+ wr %g0,$ASI_FL16_P,%asi ! setup %asi for 16-bit FP loads
+
+ add $rp,$num,$rp ! readjust input pointers to point
+ add $ap,$num,$ap ! at the ends too...
+ add $bp,$num,$bp
+ add $np,$num,$np
+
+ stx %o7,[%sp+$bias+$frame+48] ! save %asi
+
+ sub %g0,$num,$i ! i=-num
+ sub %g0,$num,$j ! j=-num
+
+ add $ap,$j,%o3
+ add $bp,$i,%o4
+
+ ld [%o3+4],%g1 ! bp[0]
+ ld [%o3+0],%o0
+ ld [%o4+4],%g5 ! ap[0]
+ sllx %g1,32,%g1
+ ld [%o4+0],%o1
+ sllx %g5,32,%g5
+ or %g1,%o0,%o0
+ or %g5,%o1,%o1
+
+ add $np,$j,%o5
+
+ mulx %o1,%o0,%o0 ! ap[0]*bp[0]
+ mulx $n0,%o0,%o0 ! ap[0]*bp[0]*n0
+ stx %o0,[%sp+$bias+$frame+0]
+
+ ld [%o3+0],$alo_ ! load a[j] as pair of 32-bit words
+ fzeros $alo
+ ld [%o3+4],$ahi_
+ fzeros $ahi
+ ld [%o5+0],$nlo_ ! load n[j] as pair of 32-bit words
+ fzeros $nlo
+ ld [%o5+4],$nhi_
+ fzeros $nhi
+
+ ! transfer b[i] to FPU as 4x16-bit values
+ ldda [%o4+2]%asi,$ba
+ fxtod $alo,$alo
+ ldda [%o4+0]%asi,$bb
+ fxtod $ahi,$ahi
+ ldda [%o4+6]%asi,$bc
+ fxtod $nlo,$nlo
+ ldda [%o4+4]%asi,$bd
+ fxtod $nhi,$nhi
+
+ ! transfer ap[0]*b[0]*n0 to FPU as 4x16-bit values
+ ldda [%sp+$bias+$frame+6]%asi,$na
+ fxtod $ba,$ba
+ ldda [%sp+$bias+$frame+4]%asi,$nb
+ fxtod $bb,$bb
+ ldda [%sp+$bias+$frame+2]%asi,$nc
+ fxtod $bc,$bc
+ ldda [%sp+$bias+$frame+0]%asi,$nd
+ fxtod $bd,$bd
+
+ std $alo,[$ap_l+$j] ! save smashed ap[j] in double format
+ fxtod $na,$na
+ std $ahi,[$ap_h+$j]
+ fxtod $nb,$nb
+ std $nlo,[$np_l+$j] ! save smashed np[j] in double format
+ fxtod $nc,$nc
+ std $nhi,[$np_h+$j]
+ fxtod $nd,$nd
+
+ fmuld $alo,$ba,$aloa
+ fmuld $nlo,$na,$nloa
+ fmuld $alo,$bb,$alob
+ fmuld $nlo,$nb,$nlob
+ fmuld $alo,$bc,$aloc
+ faddd $aloa,$nloa,$nloa
+ fmuld $nlo,$nc,$nloc
+ fmuld $alo,$bd,$alod
+ faddd $alob,$nlob,$nlob
+ fmuld $nlo,$nd,$nlod
+ fmuld $ahi,$ba,$ahia
+ faddd $aloc,$nloc,$nloc
+ fmuld $nhi,$na,$nhia
+ fmuld $ahi,$bb,$ahib
+ faddd $alod,$nlod,$nlod
+ fmuld $nhi,$nb,$nhib
+ fmuld $ahi,$bc,$ahic
+ faddd $ahia,$nhia,$nhia
+ fmuld $nhi,$nc,$nhic
+ fmuld $ahi,$bd,$ahid
+ faddd $ahib,$nhib,$nhib
+ fmuld $nhi,$nd,$nhid
+
+ faddd $ahic,$nhic,$dota ! $nhic
+ faddd $ahid,$nhid,$dotb ! $nhid
+
+ faddd $nloc,$nhia,$nloc
+ faddd $nlod,$nhib,$nlod
+
+ fdtox $nloa,$nloa
+ fdtox $nlob,$nlob
+ fdtox $nloc,$nloc
+ fdtox $nlod,$nlod
+
+ std $nloa,[%sp+$bias+$frame+0]
+ add $j,8,$j
+ std $nlob,[%sp+$bias+$frame+8]
+ add $ap,$j,%o4
+ std $nloc,[%sp+$bias+$frame+16]
+ add $np,$j,%o5
+ std $nlod,[%sp+$bias+$frame+24]
+
+ ld [%o4+0],$alo_ ! load a[j] as pair of 32-bit words
+ fzeros $alo
+ ld [%o4+4],$ahi_
+ fzeros $ahi
+ ld [%o5+0],$nlo_ ! load n[j] as pair of 32-bit words
+ fzeros $nlo
+ ld [%o5+4],$nhi_
+ fzeros $nhi
+
+ fxtod $alo,$alo
+ fxtod $ahi,$ahi
+ fxtod $nlo,$nlo
+ fxtod $nhi,$nhi
+
+ ldx [%sp+$bias+$frame+0],%o0
+ fmuld $alo,$ba,$aloa
+ ldx [%sp+$bias+$frame+8],%o1
+ fmuld $nlo,$na,$nloa
+ ldx [%sp+$bias+$frame+16],%o2
+ fmuld $alo,$bb,$alob
+ ldx [%sp+$bias+$frame+24],%o3
+ fmuld $nlo,$nb,$nlob
+
+ srlx %o0,16,%o7
+ std $alo,[$ap_l+$j] ! save smashed ap[j] in double format
+ fmuld $alo,$bc,$aloc
+ add %o7,%o1,%o1
+ std $ahi,[$ap_h+$j]
+ faddd $aloa,$nloa,$nloa
+ fmuld $nlo,$nc,$nloc
+ srlx %o1,16,%o7
+ std $nlo,[$np_l+$j] ! save smashed np[j] in double format
+ fmuld $alo,$bd,$alod
+ add %o7,%o2,%o2
+ std $nhi,[$np_h+$j]
+ faddd $alob,$nlob,$nlob
+ fmuld $nlo,$nd,$nlod
+ srlx %o2,16,%o7
+ fmuld $ahi,$ba,$ahia
+ add %o7,%o3,%o3 ! %o3.%o2[0..15].%o1[0..15].%o0[0..15]
+ faddd $aloc,$nloc,$nloc
+ fmuld $nhi,$na,$nhia
+ !and %o0,$mask,%o0
+ !and %o1,$mask,%o1
+ !and %o2,$mask,%o2
+ !sllx %o1,16,%o1
+ !sllx %o2,32,%o2
+ !sllx %o3,48,%o7
+ !or %o1,%o0,%o0
+ !or %o2,%o0,%o0
+ !or %o7,%o0,%o0 ! 64-bit result
+ srlx %o3,16,%g1 ! 34-bit carry
+ fmuld $ahi,$bb,$ahib
+
+ faddd $alod,$nlod,$nlod
+ fmuld $nhi,$nb,$nhib
+ fmuld $ahi,$bc,$ahic
+ faddd $ahia,$nhia,$nhia
+ fmuld $nhi,$nc,$nhic
+ fmuld $ahi,$bd,$ahid
+ faddd $ahib,$nhib,$nhib
+ fmuld $nhi,$nd,$nhid
+
+ faddd $dota,$nloa,$nloa
+ faddd $dotb,$nlob,$nlob
+ faddd $ahic,$nhic,$dota ! $nhic
+ faddd $ahid,$nhid,$dotb ! $nhid
+
+ faddd $nloc,$nhia,$nloc
+ faddd $nlod,$nhib,$nlod
+
+ fdtox $nloa,$nloa
+ fdtox $nlob,$nlob
+ fdtox $nloc,$nloc
+ fdtox $nlod,$nlod
+
+ std $nloa,[%sp+$bias+$frame+0]
+ std $nlob,[%sp+$bias+$frame+8]
+ addcc $j,8,$j
+ std $nloc,[%sp+$bias+$frame+16]
+ bz,pn %icc,.L1stskip
+ std $nlod,[%sp+$bias+$frame+24]
+
+.align 32 ! incidentally already aligned !
+.L1st:
+ add $ap,$j,%o4
+ add $np,$j,%o5
+ ld [%o4+0],$alo_ ! load a[j] as pair of 32-bit words
+ fzeros $alo
+ ld [%o4+4],$ahi_
+ fzeros $ahi
+ ld [%o5+0],$nlo_ ! load n[j] as pair of 32-bit words
+ fzeros $nlo
+ ld [%o5+4],$nhi_
+ fzeros $nhi
+
+ fxtod $alo,$alo
+ fxtod $ahi,$ahi
+ fxtod $nlo,$nlo
+ fxtod $nhi,$nhi
+
+ ldx [%sp+$bias+$frame+0],%o0
+ fmuld $alo,$ba,$aloa
+ ldx [%sp+$bias+$frame+8],%o1
+ fmuld $nlo,$na,$nloa
+ ldx [%sp+$bias+$frame+16],%o2
+ fmuld $alo,$bb,$alob
+ ldx [%sp+$bias+$frame+24],%o3
+ fmuld $nlo,$nb,$nlob
+
+ srlx %o0,16,%o7
+ std $alo,[$ap_l+$j] ! save smashed ap[j] in double format
+ fmuld $alo,$bc,$aloc
+ add %o7,%o1,%o1
+ std $ahi,[$ap_h+$j]
+ faddd $aloa,$nloa,$nloa
+ fmuld $nlo,$nc,$nloc
+ srlx %o1,16,%o7
+ std $nlo,[$np_l+$j] ! save smashed np[j] in double format
+ fmuld $alo,$bd,$alod
+ add %o7,%o2,%o2
+ std $nhi,[$np_h+$j]
+ faddd $alob,$nlob,$nlob
+ fmuld $nlo,$nd,$nlod
+ srlx %o2,16,%o7
+ fmuld $ahi,$ba,$ahia
+ add %o7,%o3,%o3 ! %o3.%o2[0..15].%o1[0..15].%o0[0..15]
+ and %o0,$mask,%o0
+ faddd $aloc,$nloc,$nloc
+ fmuld $nhi,$na,$nhia
+ and %o1,$mask,%o1
+ and %o2,$mask,%o2
+ fmuld $ahi,$bb,$ahib
+ sllx %o1,16,%o1
+ faddd $alod,$nlod,$nlod
+ fmuld $nhi,$nb,$nhib
+ sllx %o2,32,%o2
+ fmuld $ahi,$bc,$ahic
+ sllx %o3,48,%o7
+ or %o1,%o0,%o0
+ faddd $ahia,$nhia,$nhia
+ fmuld $nhi,$nc,$nhic
+ or %o2,%o0,%o0
+ fmuld $ahi,$bd,$ahid
+ or %o7,%o0,%o0 ! 64-bit result
+ faddd $ahib,$nhib,$nhib
+ fmuld $nhi,$nd,$nhid
+ addcc %g1,%o0,%o0
+ faddd $dota,$nloa,$nloa
+ srlx %o3,16,%g1 ! 34-bit carry
+ faddd $dotb,$nlob,$nlob
+ bcs,a %xcc,.+8
+ add %g1,1,%g1
+
+ stx %o0,[$tp] ! tp[j-1]=
+
+ faddd $ahic,$nhic,$dota ! $nhic
+ faddd $ahid,$nhid,$dotb ! $nhid
+
+ faddd $nloc,$nhia,$nloc
+ faddd $nlod,$nhib,$nlod
+
+ fdtox $nloa,$nloa
+ fdtox $nlob,$nlob
+ fdtox $nloc,$nloc
+ fdtox $nlod,$nlod
+
+ std $nloa,[%sp+$bias+$frame+0]
+ std $nlob,[%sp+$bias+$frame+8]
+ std $nloc,[%sp+$bias+$frame+16]
+ std $nlod,[%sp+$bias+$frame+24]
+
+ addcc $j,8,$j
+ bnz,pt %icc,.L1st
+ add $tp,8,$tp
+
+.L1stskip:
+ fdtox $dota,$dota
+ fdtox $dotb,$dotb
+
+ ldx [%sp+$bias+$frame+0],%o0
+ ldx [%sp+$bias+$frame+8],%o1
+ ldx [%sp+$bias+$frame+16],%o2
+ ldx [%sp+$bias+$frame+24],%o3
+
+ srlx %o0,16,%o7
+ std $dota,[%sp+$bias+$frame+32]
+ add %o7,%o1,%o1
+ std $dotb,[%sp+$bias+$frame+40]
+ srlx %o1,16,%o7
+ add %o7,%o2,%o2
+ srlx %o2,16,%o7
+ add %o7,%o3,%o3 ! %o3.%o2[0..15].%o1[0..15].%o0[0..15]
+ and %o0,$mask,%o0
+ and %o1,$mask,%o1
+ and %o2,$mask,%o2
+ sllx %o1,16,%o1
+ sllx %o2,32,%o2
+ sllx %o3,48,%o7
+ or %o1,%o0,%o0
+ or %o2,%o0,%o0
+ or %o7,%o0,%o0 ! 64-bit result
+ ldx [%sp+$bias+$frame+32],%o4
+ addcc %g1,%o0,%o0
+ ldx [%sp+$bias+$frame+40],%o5
+ srlx %o3,16,%g1 ! 34-bit carry
+ bcs,a %xcc,.+8
+ add %g1,1,%g1
+
+ stx %o0,[$tp] ! tp[j-1]=
+ add $tp,8,$tp
+
+ srlx %o4,16,%o7
+ add %o7,%o5,%o5
+ and %o4,$mask,%o4
+ sllx %o5,16,%o7
+ or %o7,%o4,%o4
+ addcc %g1,%o4,%o4
+ srlx %o5,48,%g1
+ bcs,a %xcc,.+8
+ add %g1,1,%g1
+
+ mov %g1,$carry
+ stx %o4,[$tp] ! tp[num-1]=
+
+ ba .Louter
+ add $i,8,$i
+.align 32
+.Louter:
+ sub %g0,$num,$j ! j=-num
+ add %sp,$bias+$frame+$locals,$tp
+
+ add $ap,$j,%o3
+ add $bp,$i,%o4
+
+ ld [%o3+4],%g1 ! bp[i]
+ ld [%o3+0],%o0
+ ld [%o4+4],%g5 ! ap[0]
+ sllx %g1,32,%g1
+ ld [%o4+0],%o1
+ sllx %g5,32,%g5
+ or %g1,%o0,%o0
+ or %g5,%o1,%o1
+
+ ldx [$tp],%o2 ! tp[0]
+ mulx %o1,%o0,%o0
+ addcc %o2,%o0,%o0
+ mulx $n0,%o0,%o0 ! (ap[0]*bp[i]+t[0])*n0
+ stx %o0,[%sp+$bias+$frame+0]
+
+ ! transfer b[i] to FPU as 4x16-bit values
+ ldda [%o4+2]%asi,$ba
+ ldda [%o4+0]%asi,$bb
+ ldda [%o4+6]%asi,$bc
+ ldda [%o4+4]%asi,$bd
+
+ ! transfer (ap[0]*b[i]+t[0])*n0 to FPU as 4x16-bit values
+ ldda [%sp+$bias+$frame+6]%asi,$na
+ fxtod $ba,$ba
+ ldda [%sp+$bias+$frame+4]%asi,$nb
+ fxtod $bb,$bb
+ ldda [%sp+$bias+$frame+2]%asi,$nc
+ fxtod $bc,$bc
+ ldda [%sp+$bias+$frame+0]%asi,$nd
+ fxtod $bd,$bd
+ ldd [$ap_l+$j],$alo ! load a[j] in double format
+ fxtod $na,$na
+ ldd [$ap_h+$j],$ahi
+ fxtod $nb,$nb
+ ldd [$np_l+$j],$nlo ! load n[j] in double format
+ fxtod $nc,$nc
+ ldd [$np_h+$j],$nhi
+ fxtod $nd,$nd
+
+ fmuld $alo,$ba,$aloa
+ fmuld $nlo,$na,$nloa
+ fmuld $alo,$bb,$alob
+ fmuld $nlo,$nb,$nlob
+ fmuld $alo,$bc,$aloc
+ faddd $aloa,$nloa,$nloa
+ fmuld $nlo,$nc,$nloc
+ fmuld $alo,$bd,$alod
+ faddd $alob,$nlob,$nlob
+ fmuld $nlo,$nd,$nlod
+ fmuld $ahi,$ba,$ahia
+ faddd $aloc,$nloc,$nloc
+ fmuld $nhi,$na,$nhia
+ fmuld $ahi,$bb,$ahib
+ faddd $alod,$nlod,$nlod
+ fmuld $nhi,$nb,$nhib
+ fmuld $ahi,$bc,$ahic
+ faddd $ahia,$nhia,$nhia
+ fmuld $nhi,$nc,$nhic
+ fmuld $ahi,$bd,$ahid
+ faddd $ahib,$nhib,$nhib
+ fmuld $nhi,$nd,$nhid
+
+ faddd $ahic,$nhic,$dota ! $nhic
+ faddd $ahid,$nhid,$dotb ! $nhid
+
+ faddd $nloc,$nhia,$nloc
+ faddd $nlod,$nhib,$nlod
+
+ fdtox $nloa,$nloa
+ fdtox $nlob,$nlob
+ fdtox $nloc,$nloc
+ fdtox $nlod,$nlod
+
+ std $nloa,[%sp+$bias+$frame+0]
+ std $nlob,[%sp+$bias+$frame+8]
+ std $nloc,[%sp+$bias+$frame+16]
+ add $j,8,$j
+ std $nlod,[%sp+$bias+$frame+24]
+
+ ldd [$ap_l+$j],$alo ! load a[j] in double format
+ ldd [$ap_h+$j],$ahi
+ ldd [$np_l+$j],$nlo ! load n[j] in double format
+ ldd [$np_h+$j],$nhi
+
+ fmuld $alo,$ba,$aloa
+ fmuld $nlo,$na,$nloa
+ fmuld $alo,$bb,$alob
+ fmuld $nlo,$nb,$nlob
+ fmuld $alo,$bc,$aloc
+ ldx [%sp+$bias+$frame+0],%o0
+ faddd $aloa,$nloa,$nloa
+ fmuld $nlo,$nc,$nloc
+ ldx [%sp+$bias+$frame+8],%o1
+ fmuld $alo,$bd,$alod
+ ldx [%sp+$bias+$frame+16],%o2
+ faddd $alob,$nlob,$nlob
+ fmuld $nlo,$nd,$nlod
+ ldx [%sp+$bias+$frame+24],%o3
+ fmuld $ahi,$ba,$ahia
+
+ srlx %o0,16,%o7
+ faddd $aloc,$nloc,$nloc
+ fmuld $nhi,$na,$nhia
+ add %o7,%o1,%o1
+ fmuld $ahi,$bb,$ahib
+ srlx %o1,16,%o7
+ faddd $alod,$nlod,$nlod
+ fmuld $nhi,$nb,$nhib
+ add %o7,%o2,%o2
+ fmuld $ahi,$bc,$ahic
+ srlx %o2,16,%o7
+ faddd $ahia,$nhia,$nhia
+ fmuld $nhi,$nc,$nhic
+ add %o7,%o3,%o3 ! %o3.%o2[0..15].%o1[0..15].%o0[0..15]
+ ! why?
+ and %o0,$mask,%o0
+ fmuld $ahi,$bd,$ahid
+ and %o1,$mask,%o1
+ and %o2,$mask,%o2
+ faddd $ahib,$nhib,$nhib
+ fmuld $nhi,$nd,$nhid
+ sllx %o1,16,%o1
+ faddd $dota,$nloa,$nloa
+ sllx %o2,32,%o2
+ faddd $dotb,$nlob,$nlob
+ sllx %o3,48,%o7
+ or %o1,%o0,%o0
+ faddd $ahic,$nhic,$dota ! $nhic
+ or %o2,%o0,%o0
+ faddd $ahid,$nhid,$dotb ! $nhid
+ or %o7,%o0,%o0 ! 64-bit result
+ ldx [$tp],%o7
+ faddd $nloc,$nhia,$nloc
+ addcc %o7,%o0,%o0
+ ! end-of-why?
+ faddd $nlod,$nhib,$nlod
+ srlx %o3,16,%g1 ! 34-bit carry
+ fdtox $nloa,$nloa
+ bcs,a %xcc,.+8
+ add %g1,1,%g1
+
+ fdtox $nlob,$nlob
+ fdtox $nloc,$nloc
+ fdtox $nlod,$nlod
+
+ std $nloa,[%sp+$bias+$frame+0]
+ std $nlob,[%sp+$bias+$frame+8]
+ addcc $j,8,$j
+ std $nloc,[%sp+$bias+$frame+16]
+ bz,pn %icc,.Linnerskip
+ std $nlod,[%sp+$bias+$frame+24]
+
+ ba .Linner
+ nop
+.align 32
+.Linner:
+ ldd [$ap_l+$j],$alo ! load a[j] in double format
+ ldd [$ap_h+$j],$ahi
+ ldd [$np_l+$j],$nlo ! load n[j] in double format
+ ldd [$np_h+$j],$nhi
+
+ fmuld $alo,$ba,$aloa
+ fmuld $nlo,$na,$nloa
+ fmuld $alo,$bb,$alob
+ fmuld $nlo,$nb,$nlob
+ fmuld $alo,$bc,$aloc
+ ldx [%sp+$bias+$frame+0],%o0
+ faddd $aloa,$nloa,$nloa
+ fmuld $nlo,$nc,$nloc
+ ldx [%sp+$bias+$frame+8],%o1
+ fmuld $alo,$bd,$alod
+ ldx [%sp+$bias+$frame+16],%o2
+ faddd $alob,$nlob,$nlob
+ fmuld $nlo,$nd,$nlod
+ ldx [%sp+$bias+$frame+24],%o3
+ fmuld $ahi,$ba,$ahia
+
+ srlx %o0,16,%o7
+ faddd $aloc,$nloc,$nloc
+ fmuld $nhi,$na,$nhia
+ add %o7,%o1,%o1
+ fmuld $ahi,$bb,$ahib
+ srlx %o1,16,%o7
+ faddd $alod,$nlod,$nlod
+ fmuld $nhi,$nb,$nhib
+ add %o7,%o2,%o2
+ fmuld $ahi,$bc,$ahic
+ srlx %o2,16,%o7
+ faddd $ahia,$nhia,$nhia
+ fmuld $nhi,$nc,$nhic
+ add %o7,%o3,%o3 ! %o3.%o2[0..15].%o1[0..15].%o0[0..15]
+ and %o0,$mask,%o0
+ fmuld $ahi,$bd,$ahid
+ and %o1,$mask,%o1
+ and %o2,$mask,%o2
+ faddd $ahib,$nhib,$nhib
+ fmuld $nhi,$nd,$nhid
+ sllx %o1,16,%o1
+ faddd $dota,$nloa,$nloa
+ sllx %o2,32,%o2
+ faddd $dotb,$nlob,$nlob
+ sllx %o3,48,%o7
+ or %o1,%o0,%o0
+ faddd $ahic,$nhic,$dota ! $nhic
+ or %o2,%o0,%o0
+ faddd $ahid,$nhid,$dotb ! $nhid
+ or %o7,%o0,%o0 ! 64-bit result
+ faddd $nloc,$nhia,$nloc
+ addcc %g1,%o0,%o0
+ ldx [$tp+8],%o7 ! tp[j]
+ faddd $nlod,$nhib,$nlod
+ srlx %o3,16,%g1 ! 34-bit carry
+ fdtox $nloa,$nloa
+ bcs,a %xcc,.+8
+ add %g1,1,%g1
+ fdtox $nlob,$nlob
+ addcc %o7,%o0,%o0
+ fdtox $nloc,$nloc
+ bcs,a %xcc,.+8
+ add %g1,1,%g1
+
+ stx %o0,[$tp] ! tp[j-1]
+ fdtox $nlod,$nlod
+
+ std $nloa,[%sp+$bias+$frame+0]
+ std $nlob,[%sp+$bias+$frame+8]
+ std $nloc,[%sp+$bias+$frame+16]
+ addcc $j,8,$j
+ std $nlod,[%sp+$bias+$frame+24]
+ bnz,pt %icc,.Linner
+ add $tp,8,$tp
+
+.Linnerskip:
+ fdtox $dota,$dota
+ fdtox $dotb,$dotb
+
+ ldx [%sp+$bias+$frame+0],%o0
+ ldx [%sp+$bias+$frame+8],%o1
+ ldx [%sp+$bias+$frame+16],%o2
+ ldx [%sp+$bias+$frame+24],%o3
+
+ srlx %o0,16,%o7
+ std $dota,[%sp+$bias+$frame+32]
+ add %o7,%o1,%o1
+ std $dotb,[%sp+$bias+$frame+40]
+ srlx %o1,16,%o7
+ add %o7,%o2,%o2
+ srlx %o2,16,%o7
+ add %o7,%o3,%o3 ! %o3.%o2[0..15].%o1[0..15].%o0[0..15]
+ and %o0,$mask,%o0
+ and %o1,$mask,%o1
+ and %o2,$mask,%o2
+ sllx %o1,16,%o1
+ sllx %o2,32,%o2
+ sllx %o3,48,%o7
+ or %o1,%o0,%o0
+ or %o2,%o0,%o0
+ ldx [%sp+$bias+$frame+32],%o4
+ or %o7,%o0,%o0 ! 64-bit result
+ ldx [%sp+$bias+$frame+40],%o5
+ addcc %g1,%o0,%o0
+ ldx [$tp+8],%o7 ! tp[j]
+ srlx %o3,16,%g1 ! 34-bit carry
+ bcs,a %xcc,.+8
+ add %g1,1,%g1
+
+ addcc %o7,%o0,%o0
+ bcs,a %xcc,.+8
+ add %g1,1,%g1
+
+ stx %o0,[$tp] ! tp[j-1]
+ add $tp,8,$tp
+
+ srlx %o4,16,%o7
+ add %o7,%o5,%o5
+ and %o4,$mask,%o4
+ sllx %o5,16,%o7
+ or %o7,%o4,%o4
+ addcc %g1,%o4,%o4
+ srlx %o5,48,%g1
+ bcs,a %xcc,.+8
+ add %g1,1,%g1
+
+ addcc $carry,%o4,%o4
+ stx %o4,[$tp] ! tp[num-1]
+ mov %g1,$carry
+ bcs,a %xcc,.+8
+ add $carry,1,$carry
+
+ addcc $i,8,$i
+ bnz %icc,.Louter
+ nop
+
+ add $tp,8,$tp ! adjust tp to point at the end
+ orn %g0,%g0,%g4
+ sub %g0,$num,%o7 ! n=-num
+ ba .Lsub
+ subcc %g0,%g0,%g0 ! clear %icc.c
+
+.align 32
+.Lsub:
+ ldx [$tp+%o7],%o0
+ add $np,%o7,%g1
+ ld [%g1+0],%o2
+ ld [%g1+4],%o3
+ srlx %o0,32,%o1
+ subccc %o0,%o2,%o2
+ add $rp,%o7,%g1
+ subccc %o1,%o3,%o3
+ st %o2,[%g1+0]
+ add %o7,8,%o7
+ brnz,pt %o7,.Lsub
+ st %o3,[%g1+4]
+ subc $carry,0,%g4
+ sub %g0,$num,%o7 ! n=-num
+ ba .Lcopy
+ nop
+
+.align 32
+.Lcopy:
+ ldx [$tp+%o7],%o0
+ add $rp,%o7,%g1
+ ld [%g1+0],%o2
+ ld [%g1+4],%o3
+ stx %g0,[$tp+%o7]
+ and %o0,%g4,%o0
+ srlx %o0,32,%o1
+ andn %o2,%g4,%o2
+ andn %o3,%g4,%o3
+ or %o2,%o0,%o0
+ or %o3,%o1,%o1
+ st %o0,[%g1+0]
+ add %o7,8,%o7
+ brnz,pt %o7,.Lcopy
+ st %o1,[%g1+4]
+ sub %g0,$num,%o7 ! n=-num
+
+.Lzap:
+ stx %g0,[$ap_l+%o7]
+ stx %g0,[$ap_h+%o7]
+ stx %g0,[$np_l+%o7]
+ stx %g0,[$np_h+%o7]
+ add %o7,8,%o7
+ brnz,pt %o7,.Lzap
+ nop
+
+ ldx [%sp+$bias+$frame+48],%o7
+ wr %g0,%o7,%asi ! restore %asi
+
+ mov 1,%i0
+.Lret:
+ ret
+ restore
+.type $fname,#function
+.size $fname,(.-$fname)
+.asciz "Montgomery Multiplication for UltraSPARC, CRYPTOGAMS by <appro\@openssl.org>"
+.align 32
+___
+
+$code =~ s/\`([^\`]*)\`/eval($1)/gem;
+
+# Below substitution makes it possible to compile without demanding
+# VIS extensions on command line, e.g. -xarch=v9 vs. -xarch=v9a. I
+# dare to do this, because VIS capability is detected at run-time now
+# and this routine is not called on CPU not capable to execute it. Do
+# note that fzeros is not the only VIS dependency! Another dependency
+# is implicit and is just _a_ numerical value loaded to %asi register,
+# which assembler can't recognize as VIS specific...
+$code =~ s/fzeros\s+%f([0-9]+)/
+ sprintf(".word\t0x%x\t! fzeros %%f%d",0x81b00c20|($1<<25),$1)
+ /gem;
+
+print $code;
+# flush
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/via-mont.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/via-mont.pl
new file mode 100644
index 0000000..9dbc8d4
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/via-mont.pl
@@ -0,0 +1,251 @@
+#! /usr/bin/env perl
+# Copyright 2006-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+#
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+#
+# Wrapper around 'rep montmul', VIA-specific instruction accessing
+# PadLock Montgomery Multiplier. The wrapper is designed as drop-in
+# replacement for OpenSSL bn_mul_mont [first implemented in 0.9.9].
+#
+# Below are interleaved outputs from 'openssl speed rsa dsa' for 4
+# different software configurations on 1.5GHz VIA Esther processor.
+# Lines marked with "software integer" denote performance of hand-
+# coded integer-only assembler found in OpenSSL 0.9.7. "Software SSE2"
+# refers to hand-coded SSE2 Montgomery multiplication procedure found
+# OpenSSL 0.9.9. "Hardware VIA SDK" refers to padlock_pmm routine from
+# Padlock SDK 2.0.1 available for download from VIA, which naturally
+# utilizes the magic 'repz montmul' instruction. And finally "hardware
+# this" refers to *this* implementation which also uses 'repz montmul'
+#
+# sign verify sign/s verify/s
+# rsa 512 bits 0.001720s 0.000140s 581.4 7149.7 software integer
+# rsa 512 bits 0.000690s 0.000086s 1450.3 11606.0 software SSE2
+# rsa 512 bits 0.006136s 0.000201s 163.0 4974.5 hardware VIA SDK
+# rsa 512 bits 0.000712s 0.000050s 1404.9 19858.5 hardware this
+#
+# rsa 1024 bits 0.008518s 0.000413s 117.4 2420.8 software integer
+# rsa 1024 bits 0.004275s 0.000277s 233.9 3609.7 software SSE2
+# rsa 1024 bits 0.012136s 0.000260s 82.4 3844.5 hardware VIA SDK
+# rsa 1024 bits 0.002522s 0.000116s 396.5 8650.9 hardware this
+#
+# rsa 2048 bits 0.050101s 0.001371s 20.0 729.6 software integer
+# rsa 2048 bits 0.030273s 0.001008s 33.0 991.9 software SSE2
+# rsa 2048 bits 0.030833s 0.000976s 32.4 1025.1 hardware VIA SDK
+# rsa 2048 bits 0.011879s 0.000342s 84.2 2921.7 hardware this
+#
+# rsa 4096 bits 0.327097s 0.004859s 3.1 205.8 software integer
+# rsa 4096 bits 0.229318s 0.003859s 4.4 259.2 software SSE2
+# rsa 4096 bits 0.233953s 0.003274s 4.3 305.4 hardware VIA SDK
+# rsa 4096 bits 0.070493s 0.001166s 14.2 857.6 hardware this
+#
+# dsa 512 bits 0.001342s 0.001651s 745.2 605.7 software integer
+# dsa 512 bits 0.000844s 0.000987s 1185.3 1013.1 software SSE2
+# dsa 512 bits 0.001902s 0.002247s 525.6 444.9 hardware VIA SDK
+# dsa 512 bits 0.000458s 0.000524s 2182.2 1909.1 hardware this
+#
+# dsa 1024 bits 0.003964s 0.004926s 252.3 203.0 software integer
+# dsa 1024 bits 0.002686s 0.003166s 372.3 315.8 software SSE2
+# dsa 1024 bits 0.002397s 0.002823s 417.1 354.3 hardware VIA SDK
+# dsa 1024 bits 0.000978s 0.001170s 1022.2 855.0 hardware this
+#
+# dsa 2048 bits 0.013280s 0.016518s 75.3 60.5 software integer
+# dsa 2048 bits 0.009911s 0.011522s 100.9 86.8 software SSE2
+# dsa 2048 bits 0.009542s 0.011763s 104.8 85.0 hardware VIA SDK
+# dsa 2048 bits 0.002884s 0.003352s 346.8 298.3 hardware this
+#
+# To give you some other reference point here is output for 2.4GHz P4
+# running hand-coded SSE2 bn_mul_mont found in 0.9.9, i.e. "software
+# SSE2" in above terms.
+#
+# rsa 512 bits 0.000407s 0.000047s 2454.2 21137.0
+# rsa 1024 bits 0.002426s 0.000141s 412.1 7100.0
+# rsa 2048 bits 0.015046s 0.000491s 66.5 2034.9
+# rsa 4096 bits 0.109770s 0.002379s 9.1 420.3
+# dsa 512 bits 0.000438s 0.000525s 2281.1 1904.1
+# dsa 1024 bits 0.001346s 0.001595s 742.7 627.0
+# dsa 2048 bits 0.004745s 0.005582s 210.7 179.1
+#
+# Conclusions:
+# - VIA SDK leaves a *lot* of room for improvement (which this
+# implementation successfully fills:-);
+# - 'rep montmul' gives up to >3x performance improvement depending on
+# key length;
+# - in terms of absolute performance it delivers approximately as much
+# as modern out-of-order 32-bit cores [again, for longer keys].
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+push(@INC,"${dir}","${dir}../../perlasm");
+require "x86asm.pl";
+
+$output = pop;
+open STDOUT,">$output";
+
+&asm_init($ARGV[0]);
+
+# int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp, const BN_ULONG *np,const BN_ULONG *n0, int num);
+$func="bn_mul_mont_padlock";
+
+$pad=16*1; # amount of reserved bytes on top of every vector
+
+# stack layout
+$mZeroPrime=&DWP(0,"esp"); # these are specified by VIA
+$A=&DWP(4,"esp");
+$B=&DWP(8,"esp");
+$T=&DWP(12,"esp");
+$M=&DWP(16,"esp");
+$scratch=&DWP(20,"esp");
+$rp=&DWP(24,"esp"); # these are mine
+$sp=&DWP(28,"esp");
+# &DWP(32,"esp") # 32 byte scratch area
+# &DWP(64+(4*$num+$pad)*0,"esp") # padded tp[num]
+# &DWP(64+(4*$num+$pad)*1,"esp") # padded copy of ap[num]
+# &DWP(64+(4*$num+$pad)*2,"esp") # padded copy of bp[num]
+# &DWP(64+(4*$num+$pad)*3,"esp") # padded copy of np[num]
+# Note that SDK suggests to unconditionally allocate 2K per vector. This
+# has quite an impact on performance. It naturally depends on key length,
+# but to give an example 1024 bit private RSA key operations suffer >30%
+# penalty. I allocate only as much as actually required...
+
+&function_begin($func);
+ &xor ("eax","eax");
+ &mov ("ecx",&wparam(5)); # num
+ # meet VIA's limitations for num [note that the specification
+ # expresses them in bits, while we work with amount of 32-bit words]
+ &test ("ecx",3);
+ &jnz (&label("leave")); # num % 4 != 0
+ &cmp ("ecx",8);
+ &jb (&label("leave")); # num < 8
+ &cmp ("ecx",1024);
+ &ja (&label("leave")); # num > 1024
+
+ &pushf ();
+ &cld ();
+
+ &mov ("edi",&wparam(0)); # rp
+ &mov ("eax",&wparam(1)); # ap
+ &mov ("ebx",&wparam(2)); # bp
+ &mov ("edx",&wparam(3)); # np
+ &mov ("esi",&wparam(4)); # n0
+ &mov ("esi",&DWP(0,"esi")); # *n0
+
+ &lea ("ecx",&DWP($pad,"","ecx",4)); # ecx becomes vector size in bytes
+ &lea ("ebp",&DWP(64,"","ecx",4)); # allocate 4 vectors + 64 bytes
+ &neg ("ebp");
+ &add ("ebp","esp");
+ &and ("ebp",-64); # align to cache-line
+ &xchg ("ebp","esp"); # alloca
+
+ &mov ($rp,"edi"); # save rp
+ &mov ($sp,"ebp"); # save esp
+
+ &mov ($mZeroPrime,"esi");
+ &lea ("esi",&DWP(64,"esp")); # tp
+ &mov ($T,"esi");
+ &lea ("edi",&DWP(32,"esp")); # scratch area
+ &mov ($scratch,"edi");
+ &mov ("esi","eax");
+
+ &lea ("ebp",&DWP(-$pad,"ecx"));
+ &shr ("ebp",2); # restore original num value in ebp
+
+ &xor ("eax","eax");
+
+ &mov ("ecx","ebp");
+ &lea ("ecx",&DWP((32+$pad)/4,"ecx"));# padded tp + scratch
+ &data_byte(0xf3,0xab); # rep stosl, bzero
+
+ &mov ("ecx","ebp");
+ &lea ("edi",&DWP(64+$pad,"esp","ecx",4));# pointer to ap copy
+ &mov ($A,"edi");
+ &data_byte(0xf3,0xa5); # rep movsl, memcpy
+ &mov ("ecx",$pad/4);
+ &data_byte(0xf3,0xab); # rep stosl, bzero pad
+ # edi points at the end of padded ap copy...
+
+ &mov ("ecx","ebp");
+ &mov ("esi","ebx");
+ &mov ($B,"edi");
+ &data_byte(0xf3,0xa5); # rep movsl, memcpy
+ &mov ("ecx",$pad/4);
+ &data_byte(0xf3,0xab); # rep stosl, bzero pad
+ # edi points at the end of padded bp copy...
+
+ &mov ("ecx","ebp");
+ &mov ("esi","edx");
+ &mov ($M,"edi");
+ &data_byte(0xf3,0xa5); # rep movsl, memcpy
+ &mov ("ecx",$pad/4);
+ &data_byte(0xf3,0xab); # rep stosl, bzero pad
+ # edi points at the end of padded np copy...
+
+ # let magic happen...
+ &mov ("ecx","ebp");
+ &mov ("esi","esp");
+ &shl ("ecx",5); # convert word counter to bit counter
+ &align (4);
+ &data_byte(0xf3,0x0f,0xa6,0xc0);# rep montmul
+
+ &mov ("ecx","ebp");
+ &lea ("esi",&DWP(64,"esp")); # tp
+ # edi still points at the end of padded np copy...
+ &neg ("ebp");
+ &lea ("ebp",&DWP(-$pad,"edi","ebp",4)); # so just "rewind"
+ &mov ("edi",$rp); # restore rp
+ &xor ("edx","edx"); # i=0 and clear CF
+
+&set_label("sub",8);
+ &mov ("eax",&DWP(0,"esi","edx",4));
+ &sbb ("eax",&DWP(0,"ebp","edx",4));
+ &mov (&DWP(0,"edi","edx",4),"eax"); # rp[i]=tp[i]-np[i]
+ &lea ("edx",&DWP(1,"edx")); # i++
+ &loop (&label("sub")); # doesn't affect CF!
+
+ &mov ("eax",&DWP(0,"esi","edx",4)); # upmost overflow bit
+ &sbb ("eax",0);
+
+ &mov ("ecx","edx"); # num
+ &mov ("edx",0); # i=0
+
+&set_label("copy",8);
+ &mov ("ebx",&DWP(0,"esi","edx",4));
+ &mov ("eax",&DWP(0,"edi","edx",4));
+ &mov (&DWP(0,"esi","edx",4),"ecx"); # zap tp
+ &cmovc ("eax","ebx");
+ &mov (&DWP(0,"edi","edx",4),"eax");
+ &lea ("edx",&DWP(1,"edx")); # i++
+ &loop (&label("copy"));
+
+ &mov ("ebp",$sp);
+ &xor ("eax","eax");
+
+ &mov ("ecx",64/4);
+ &mov ("edi","esp"); # zap frame including scratch area
+ &data_byte(0xf3,0xab); # rep stosl, bzero
+
+ # zap copies of ap, bp and np
+ &lea ("edi",&DWP(64+$pad,"esp","edx",4));# pointer to ap
+ &lea ("ecx",&DWP(3*$pad/4,"edx","edx",2));
+ &data_byte(0xf3,0xab); # rep stosl, bzero
+
+ &mov ("esp","ebp");
+ &inc ("eax"); # signal "done"
+ &popf ();
+&set_label("leave");
+&function_end($func);
+
+&asciz("Padlock Montgomery Multiplication, CRYPTOGAMS by <appro\@openssl.org>");
+
+&asm_finish();
+
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/vis3-mont.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/vis3-mont.pl
new file mode 100644
index 0000000..d797af8
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/vis3-mont.pl
@@ -0,0 +1,384 @@
+#! /usr/bin/env perl
+# Copyright 2012-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+
+# October 2012.
+#
+# SPARCv9 VIS3 Montgomery multiplication procedure suitable for T3 and
+# onward. There are three new instructions used here: umulxhi,
+# addxc[cc] and initializing store. On T3 RSA private key operations
+# are 1.54/1.87/2.11/2.26 times faster for 512/1024/2048/4096-bit key
+# lengths. This is without dedicated squaring procedure. On T4
+# corresponding coefficients are 1.47/2.10/2.80/2.90x, which is mostly
+# for reference purposes, because T4 has dedicated Montgomery
+# multiplication and squaring *instructions* that deliver even more.
+
+$output = pop;
+open STDOUT,">$output";
+
+$frame = "STACK_FRAME";
+$bias = "STACK_BIAS";
+
+$code.=<<___;
+#include "sparc_arch.h"
+
+#ifdef __arch64__
+.register %g2,#scratch
+.register %g3,#scratch
+#endif
+
+.section ".text",#alloc,#execinstr
+___
+
+($n0,$m0,$m1,$lo0,$hi0, $lo1,$hi1,$aj,$alo,$nj,$nlo,$tj)=
+ (map("%g$_",(1..5)),map("%o$_",(0..5,7)));
+
+# int bn_mul_mont(
+$rp="%o0"; # BN_ULONG *rp,
+$ap="%o1"; # const BN_ULONG *ap,
+$bp="%o2"; # const BN_ULONG *bp,
+$np="%o3"; # const BN_ULONG *np,
+$n0p="%o4"; # const BN_ULONG *n0,
+$num="%o5"; # int num); # caller ensures that num is even
+ # and >=6
+$code.=<<___;
+.globl bn_mul_mont_vis3
+.align 32
+bn_mul_mont_vis3:
+ add %sp, $bias, %g4 ! real top of stack
+ sll $num, 2, $num ! size in bytes
+ add $num, 63, %g5
+ andn %g5, 63, %g5 ! buffer size rounded up to 64 bytes
+ add %g5, %g5, %g1
+ add %g5, %g1, %g1 ! 3*buffer size
+ sub %g4, %g1, %g1
+ andn %g1, 63, %g1 ! align at 64 byte
+ sub %g1, $frame, %g1 ! new top of stack
+ sub %g1, %g4, %g1
+
+ save %sp, %g1, %sp
+___
+
+# +-------------------------------+<----- %sp
+# . .
+# +-------------------------------+<----- aligned at 64 bytes
+# | __int64 tmp[0] |
+# +-------------------------------+
+# . .
+# . .
+# +-------------------------------+<----- aligned at 64 bytes
+# | __int64 ap[1..0] | converted ap[]
+# +-------------------------------+
+# | __int64 np[1..0] | converted np[]
+# +-------------------------------+
+# | __int64 ap[3..2] |
+# . .
+# . .
+# +-------------------------------+
+($rp,$ap,$bp,$np,$n0p,$num)=map("%i$_",(0..5));
+($t0,$t1,$t2,$t3,$cnt,$tp,$bufsz,$anp)=map("%l$_",(0..7));
+($ovf,$i)=($t0,$t1);
+$code.=<<___;
+ ld [$n0p+0], $t0 ! pull n0[0..1] value
+ add %sp, $bias+$frame, $tp
+ ld [$n0p+4], $t1
+ add $tp, %g5, $anp
+ ld [$bp+0], $t2 ! m0=bp[0]
+ sllx $t1, 32, $n0
+ ld [$bp+4], $t3
+ or $t0, $n0, $n0
+ add $bp, 8, $bp
+
+ ld [$ap+0], $t0 ! ap[0]
+ sllx $t3, 32, $m0
+ ld [$ap+4], $t1
+ or $t2, $m0, $m0
+
+ ld [$ap+8], $t2 ! ap[1]
+ sllx $t1, 32, $aj
+ ld [$ap+12], $t3
+ or $t0, $aj, $aj
+ add $ap, 16, $ap
+ stx $aj, [$anp] ! converted ap[0]
+
+ mulx $aj, $m0, $lo0 ! ap[0]*bp[0]
+ umulxhi $aj, $m0, $hi0
+
+ ld [$np+0], $t0 ! np[0]
+ sllx $t3, 32, $aj
+ ld [$np+4], $t1
+ or $t2, $aj, $aj
+
+ ld [$np+8], $t2 ! np[1]
+ sllx $t1, 32, $nj
+ ld [$np+12], $t3
+ or $t0, $nj, $nj
+ add $np, 16, $np
+ stx $nj, [$anp+8] ! converted np[0]
+
+ mulx $lo0, $n0, $m1 ! "tp[0]"*n0
+ stx $aj, [$anp+16] ! converted ap[1]
+
+ mulx $aj, $m0, $alo ! ap[1]*bp[0]
+ umulxhi $aj, $m0, $aj ! ahi=aj
+
+ mulx $nj, $m1, $lo1 ! np[0]*m1
+ umulxhi $nj, $m1, $hi1
+
+ sllx $t3, 32, $nj
+ or $t2, $nj, $nj
+ stx $nj, [$anp+24] ! converted np[1]
+ add $anp, 32, $anp
+
+ addcc $lo0, $lo1, $lo1
+ addxc %g0, $hi1, $hi1
+
+ mulx $nj, $m1, $nlo ! np[1]*m1
+ umulxhi $nj, $m1, $nj ! nhi=nj
+
+ ba .L1st
+ sub $num, 24, $cnt ! cnt=num-3
+
+.align 16
+.L1st:
+ ld [$ap+0], $t0 ! ap[j]
+ addcc $alo, $hi0, $lo0
+ ld [$ap+4], $t1
+ addxc $aj, %g0, $hi0
+
+ sllx $t1, 32, $aj
+ add $ap, 8, $ap
+ or $t0, $aj, $aj
+ stx $aj, [$anp] ! converted ap[j]
+
+ ld [$np+0], $t2 ! np[j]
+ addcc $nlo, $hi1, $lo1
+ ld [$np+4], $t3
+ addxc $nj, %g0, $hi1 ! nhi=nj
+
+ sllx $t3, 32, $nj
+ add $np, 8, $np
+ mulx $aj, $m0, $alo ! ap[j]*bp[0]
+ or $t2, $nj, $nj
+ umulxhi $aj, $m0, $aj ! ahi=aj
+ stx $nj, [$anp+8] ! converted np[j]
+ add $anp, 16, $anp ! anp++
+
+ mulx $nj, $m1, $nlo ! np[j]*m1
+ addcc $lo0, $lo1, $lo1 ! np[j]*m1+ap[j]*bp[0]
+ umulxhi $nj, $m1, $nj ! nhi=nj
+ addxc %g0, $hi1, $hi1
+ stx $lo1, [$tp] ! tp[j-1]
+ add $tp, 8, $tp ! tp++
+
+ brnz,pt $cnt, .L1st
+ sub $cnt, 8, $cnt ! j--
+!.L1st
+ addcc $alo, $hi0, $lo0
+ addxc $aj, %g0, $hi0 ! ahi=aj
+
+ addcc $nlo, $hi1, $lo1
+ addxc $nj, %g0, $hi1
+ addcc $lo0, $lo1, $lo1 ! np[j]*m1+ap[j]*bp[0]
+ addxc %g0, $hi1, $hi1
+ stx $lo1, [$tp] ! tp[j-1]
+ add $tp, 8, $tp
+
+ addcc $hi0, $hi1, $hi1
+ addxc %g0, %g0, $ovf ! upmost overflow bit
+ stx $hi1, [$tp]
+ add $tp, 8, $tp
+
+ ba .Louter
+ sub $num, 16, $i ! i=num-2
+
+.align 16
+.Louter:
+ ld [$bp+0], $t2 ! m0=bp[i]
+ ld [$bp+4], $t3
+
+ sub $anp, $num, $anp ! rewind
+ sub $tp, $num, $tp
+ sub $anp, $num, $anp
+
+ add $bp, 8, $bp
+ sllx $t3, 32, $m0
+ ldx [$anp+0], $aj ! ap[0]
+ or $t2, $m0, $m0
+ ldx [$anp+8], $nj ! np[0]
+
+ mulx $aj, $m0, $lo0 ! ap[0]*bp[i]
+ ldx [$tp], $tj ! tp[0]
+ umulxhi $aj, $m0, $hi0
+ ldx [$anp+16], $aj ! ap[1]
+ addcc $lo0, $tj, $lo0 ! ap[0]*bp[i]+tp[0]
+ mulx $aj, $m0, $alo ! ap[1]*bp[i]
+ addxc %g0, $hi0, $hi0
+ mulx $lo0, $n0, $m1 ! tp[0]*n0
+ umulxhi $aj, $m0, $aj ! ahi=aj
+ mulx $nj, $m1, $lo1 ! np[0]*m1
+ umulxhi $nj, $m1, $hi1
+ ldx [$anp+24], $nj ! np[1]
+ add $anp, 32, $anp
+ addcc $lo1, $lo0, $lo1
+ mulx $nj, $m1, $nlo ! np[1]*m1
+ addxc %g0, $hi1, $hi1
+ umulxhi $nj, $m1, $nj ! nhi=nj
+
+ ba .Linner
+ sub $num, 24, $cnt ! cnt=num-3
+.align 16
+.Linner:
+ addcc $alo, $hi0, $lo0
+ ldx [$tp+8], $tj ! tp[j]
+ addxc $aj, %g0, $hi0 ! ahi=aj
+ ldx [$anp+0], $aj ! ap[j]
+ addcc $nlo, $hi1, $lo1
+ mulx $aj, $m0, $alo ! ap[j]*bp[i]
+ addxc $nj, %g0, $hi1 ! nhi=nj
+ ldx [$anp+8], $nj ! np[j]
+ add $anp, 16, $anp
+ umulxhi $aj, $m0, $aj ! ahi=aj
+ addcc $lo0, $tj, $lo0 ! ap[j]*bp[i]+tp[j]
+ mulx $nj, $m1, $nlo ! np[j]*m1
+ addxc %g0, $hi0, $hi0
+ umulxhi $nj, $m1, $nj ! nhi=nj
+ addcc $lo1, $lo0, $lo1 ! np[j]*m1+ap[j]*bp[i]+tp[j]
+ addxc %g0, $hi1, $hi1
+ stx $lo1, [$tp] ! tp[j-1]
+ add $tp, 8, $tp
+ brnz,pt $cnt, .Linner
+ sub $cnt, 8, $cnt
+!.Linner
+ ldx [$tp+8], $tj ! tp[j]
+ addcc $alo, $hi0, $lo0
+ addxc $aj, %g0, $hi0 ! ahi=aj
+ addcc $lo0, $tj, $lo0 ! ap[j]*bp[i]+tp[j]
+ addxc %g0, $hi0, $hi0
+
+ addcc $nlo, $hi1, $lo1
+ addxc $nj, %g0, $hi1 ! nhi=nj
+ addcc $lo1, $lo0, $lo1 ! np[j]*m1+ap[j]*bp[i]+tp[j]
+ addxc %g0, $hi1, $hi1
+ stx $lo1, [$tp] ! tp[j-1]
+
+ subcc %g0, $ovf, %g0 ! move upmost overflow to CCR.xcc
+ addxccc $hi1, $hi0, $hi1
+ addxc %g0, %g0, $ovf
+ stx $hi1, [$tp+8]
+ add $tp, 16, $tp
+
+ brnz,pt $i, .Louter
+ sub $i, 8, $i
+
+ sub $anp, $num, $anp ! rewind
+ sub $tp, $num, $tp
+ sub $anp, $num, $anp
+ ba .Lsub
+ subcc $num, 8, $cnt ! cnt=num-1 and clear CCR.xcc
+
+.align 16
+.Lsub:
+ ldx [$tp], $tj
+ add $tp, 8, $tp
+ ldx [$anp+8], $nj
+ add $anp, 16, $anp
+ subccc $tj, $nj, $t2 ! tp[j]-np[j]
+ srlx $tj, 32, $tj
+ srlx $nj, 32, $nj
+ subccc $tj, $nj, $t3
+ add $rp, 8, $rp
+ st $t2, [$rp-4] ! reverse order
+ st $t3, [$rp-8]
+ brnz,pt $cnt, .Lsub
+ sub $cnt, 8, $cnt
+
+ sub $anp, $num, $anp ! rewind
+ sub $tp, $num, $tp
+ sub $anp, $num, $anp
+ sub $rp, $num, $rp
+
+ subccc $ovf, %g0, $ovf ! handle upmost overflow bit
+ ba .Lcopy
+ sub $num, 8, $cnt
+
+.align 16
+.Lcopy: ! conditional copy
+ ld [$tp+0], $t0
+ ld [$tp+4], $t1
+ ld [$rp+0], $t2
+ ld [$rp+4], $t3
+ stx %g0, [$tp] ! zap
+ add $tp, 8, $tp
+ stx %g0, [$anp] ! zap
+ stx %g0, [$anp+8]
+ add $anp, 16, $anp
+ movcs %icc, $t0, $t2
+ movcs %icc, $t1, $t3
+ st $t3, [$rp+0] ! flip order
+ st $t2, [$rp+4]
+ add $rp, 8, $rp
+ brnz $cnt, .Lcopy
+ sub $cnt, 8, $cnt
+
+ mov 1, %o0
+ ret
+ restore
+.type bn_mul_mont_vis3, #function
+.size bn_mul_mont_vis3, .-bn_mul_mont_vis3
+.asciz "Montgomery Multiplication for SPARCv9 VIS3, CRYPTOGAMS by <appro\@openssl.org>"
+.align 4
+___
+
+# Purpose of these subroutines is to explicitly encode VIS instructions,
+# so that one can compile the module without having to specify VIS
+# extensions on compiler command line, e.g. -xarch=v9 vs. -xarch=v9a.
+# Idea is to reserve for option to produce "universal" binary and let
+# programmer detect if current CPU is VIS capable at run-time.
+sub unvis3 {
+my ($mnemonic,$rs1,$rs2,$rd)=@_;
+my %bias = ( "g" => 0, "o" => 8, "l" => 16, "i" => 24 );
+my ($ref,$opf);
+my %visopf = ( "addxc" => 0x011,
+ "addxccc" => 0x013,
+ "umulxhi" => 0x016 );
+
+ $ref = "$mnemonic\t$rs1,$rs2,$rd";
+
+ if ($opf=$visopf{$mnemonic}) {
+ foreach ($rs1,$rs2,$rd) {
+ return $ref if (!/%([goli])([0-9])/);
+ $_=$bias{$1}+$2;
+ }
+
+ return sprintf ".word\t0x%08x !%s",
+ 0x81b00000|$rd<<25|$rs1<<14|$opf<<5|$rs2,
+ $ref;
+ } else {
+ return $ref;
+ }
+}
+
+foreach (split("\n",$code)) {
+ s/\`([^\`]*)\`/eval $1/ge;
+
+ s/\b(umulxhi|addxc[c]{0,2})\s+(%[goli][0-7]),\s*(%[goli][0-7]),\s*(%[goli][0-7])/
+ &unvis3($1,$2,$3,$4)
+ /ge;
+
+ print $_,"\n";
+}
+
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/x86-gf2m.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/x86-gf2m.pl
new file mode 100644
index 0000000..436d90b
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/x86-gf2m.pl
@@ -0,0 +1,325 @@
+#! /usr/bin/env perl
+# Copyright 2011-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+#
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+#
+# May 2011
+#
+# The module implements bn_GF2m_mul_2x2 polynomial multiplication used
+# in bn_gf2m.c. It's kind of low-hanging mechanical port from C for
+# the time being... Except that it has three code paths: pure integer
+# code suitable for any x86 CPU, MMX code suitable for PIII and later
+# and PCLMULQDQ suitable for Westmere and later. Improvement varies
+# from one benchmark and µ-arch to another. Below are interval values
+# for 163- and 571-bit ECDH benchmarks relative to compiler-generated
+# code:
+#
+# PIII 16%-30%
+# P4 12%-12%
+# Opteron 18%-40%
+# Core2 19%-44%
+# Atom 38%-64%
+# Westmere 53%-121%(PCLMULQDQ)/20%-32%(MMX)
+# Sandy Bridge 72%-127%(PCLMULQDQ)/27%-23%(MMX)
+#
+# Note that above improvement coefficients are not coefficients for
+# bn_GF2m_mul_2x2 itself. For example 120% ECDH improvement is result
+# of bn_GF2m_mul_2x2 being >4x faster. As it gets faster, benchmark
+# is more and more dominated by other subroutines, most notably by
+# BN_GF2m_mod[_mul]_arr...
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+push(@INC,"${dir}","${dir}../../perlasm");
+require "x86asm.pl";
+
+$output = pop;
+open STDOUT,">$output";
+
+&asm_init($ARGV[0],$x86only = $ARGV[$#ARGV] eq "386");
+
+$sse2=0;
+for (@ARGV) { $sse2=1 if (/-DOPENSSL_IA32_SSE2/); }
+
+&external_label("OPENSSL_ia32cap_P") if ($sse2);
+
+$a="eax";
+$b="ebx";
+($a1,$a2,$a4)=("ecx","edx","ebp");
+
+$R="mm0";
+@T=("mm1","mm2");
+($A,$B,$B30,$B31)=("mm2","mm3","mm4","mm5");
+@i=("esi","edi");
+
+ if (!$x86only) {
+&function_begin_B("_mul_1x1_mmx");
+ &sub ("esp",32+4);
+ &mov ($a1,$a);
+ &lea ($a2,&DWP(0,$a,$a));
+ &and ($a1,0x3fffffff);
+ &lea ($a4,&DWP(0,$a2,$a2));
+ &mov (&DWP(0*4,"esp"),0);
+ &and ($a2,0x7fffffff);
+ &movd ($A,$a);
+ &movd ($B,$b);
+ &mov (&DWP(1*4,"esp"),$a1); # a1
+ &xor ($a1,$a2); # a1^a2
+ &pxor ($B31,$B31);
+ &pxor ($B30,$B30);
+ &mov (&DWP(2*4,"esp"),$a2); # a2
+ &xor ($a2,$a4); # a2^a4
+ &mov (&DWP(3*4,"esp"),$a1); # a1^a2
+ &pcmpgtd($B31,$A); # broadcast 31st bit
+ &paddd ($A,$A); # $A<<=1
+ &xor ($a1,$a2); # a1^a4=a1^a2^a2^a4
+ &mov (&DWP(4*4,"esp"),$a4); # a4
+ &xor ($a4,$a2); # a2=a4^a2^a4
+ &pand ($B31,$B);
+ &pcmpgtd($B30,$A); # broadcast 30th bit
+ &mov (&DWP(5*4,"esp"),$a1); # a1^a4
+ &xor ($a4,$a1); # a1^a2^a4
+ &psllq ($B31,31);
+ &pand ($B30,$B);
+ &mov (&DWP(6*4,"esp"),$a2); # a2^a4
+ &mov (@i[0],0x7);
+ &mov (&DWP(7*4,"esp"),$a4); # a1^a2^a4
+ &mov ($a4,@i[0]);
+ &and (@i[0],$b);
+ &shr ($b,3);
+ &mov (@i[1],$a4);
+ &psllq ($B30,30);
+ &and (@i[1],$b);
+ &shr ($b,3);
+ &movd ($R,&DWP(0,"esp",@i[0],4));
+ &mov (@i[0],$a4);
+ &and (@i[0],$b);
+ &shr ($b,3);
+ for($n=1;$n<9;$n++) {
+ &movd (@T[1],&DWP(0,"esp",@i[1],4));
+ &mov (@i[1],$a4);
+ &psllq (@T[1],3*$n);
+ &and (@i[1],$b);
+ &shr ($b,3);
+ &pxor ($R,@T[1]);
+
+ push(@i,shift(@i)); push(@T,shift(@T));
+ }
+ &movd (@T[1],&DWP(0,"esp",@i[1],4));
+ &pxor ($R,$B30);
+ &psllq (@T[1],3*$n++);
+ &pxor ($R,@T[1]);
+
+ &movd (@T[0],&DWP(0,"esp",@i[0],4));
+ &pxor ($R,$B31);
+ &psllq (@T[0],3*$n);
+ &add ("esp",32+4);
+ &pxor ($R,@T[0]);
+ &ret ();
+&function_end_B("_mul_1x1_mmx");
+ }
+
+($lo,$hi)=("eax","edx");
+@T=("ecx","ebp");
+
+&function_begin_B("_mul_1x1_ialu");
+ &sub ("esp",32+4);
+ &mov ($a1,$a);
+ &lea ($a2,&DWP(0,$a,$a));
+ &lea ($a4,&DWP(0,"",$a,4));
+ &and ($a1,0x3fffffff);
+ &lea (@i[1],&DWP(0,$lo,$lo));
+ &sar ($lo,31); # broadcast 31st bit
+ &mov (&DWP(0*4,"esp"),0);
+ &and ($a2,0x7fffffff);
+ &mov (&DWP(1*4,"esp"),$a1); # a1
+ &xor ($a1,$a2); # a1^a2
+ &mov (&DWP(2*4,"esp"),$a2); # a2
+ &xor ($a2,$a4); # a2^a4
+ &mov (&DWP(3*4,"esp"),$a1); # a1^a2
+ &xor ($a1,$a2); # a1^a4=a1^a2^a2^a4
+ &mov (&DWP(4*4,"esp"),$a4); # a4
+ &xor ($a4,$a2); # a2=a4^a2^a4
+ &mov (&DWP(5*4,"esp"),$a1); # a1^a4
+ &xor ($a4,$a1); # a1^a2^a4
+ &sar (@i[1],31); # broadcast 30th bit
+ &and ($lo,$b);
+ &mov (&DWP(6*4,"esp"),$a2); # a2^a4
+ &and (@i[1],$b);
+ &mov (&DWP(7*4,"esp"),$a4); # a1^a2^a4
+ &mov ($hi,$lo);
+ &shl ($lo,31);
+ &mov (@T[0],@i[1]);
+ &shr ($hi,1);
+
+ &mov (@i[0],0x7);
+ &shl (@i[1],30);
+ &and (@i[0],$b);
+ &shr (@T[0],2);
+ &xor ($lo,@i[1]);
+
+ &shr ($b,3);
+ &mov (@i[1],0x7); # 5-byte instruction!?
+ &and (@i[1],$b);
+ &shr ($b,3);
+ &xor ($hi,@T[0]);
+ &xor ($lo,&DWP(0,"esp",@i[0],4));
+ &mov (@i[0],0x7);
+ &and (@i[0],$b);
+ &shr ($b,3);
+ for($n=1;$n<9;$n++) {
+ &mov (@T[1],&DWP(0,"esp",@i[1],4));
+ &mov (@i[1],0x7);
+ &mov (@T[0],@T[1]);
+ &shl (@T[1],3*$n);
+ &and (@i[1],$b);
+ &shr (@T[0],32-3*$n);
+ &xor ($lo,@T[1]);
+ &shr ($b,3);
+ &xor ($hi,@T[0]);
+
+ push(@i,shift(@i)); push(@T,shift(@T));
+ }
+ &mov (@T[1],&DWP(0,"esp",@i[1],4));
+ &mov (@T[0],@T[1]);
+ &shl (@T[1],3*$n);
+ &mov (@i[1],&DWP(0,"esp",@i[0],4));
+ &shr (@T[0],32-3*$n); $n++;
+ &mov (@i[0],@i[1]);
+ &xor ($lo,@T[1]);
+ &shl (@i[1],3*$n);
+ &xor ($hi,@T[0]);
+ &shr (@i[0],32-3*$n);
+ &xor ($lo,@i[1]);
+ &xor ($hi,@i[0]);
+
+ &add ("esp",32+4);
+ &ret ();
+&function_end_B("_mul_1x1_ialu");
+
+# void bn_GF2m_mul_2x2(BN_ULONG *r, BN_ULONG a1, BN_ULONG a0, BN_ULONG b1, BN_ULONG b0);
+&function_begin_B("bn_GF2m_mul_2x2");
+if (!$x86only) {
+ &picmeup("edx","OPENSSL_ia32cap_P");
+ &mov ("eax",&DWP(0,"edx"));
+ &mov ("edx",&DWP(4,"edx"));
+ &test ("eax",1<<23); # check MMX bit
+ &jz (&label("ialu"));
+if ($sse2) {
+ &test ("eax",1<<24); # check FXSR bit
+ &jz (&label("mmx"));
+ &test ("edx",1<<1); # check PCLMULQDQ bit
+ &jz (&label("mmx"));
+
+ &movups ("xmm0",&QWP(8,"esp"));
+ &shufps ("xmm0","xmm0",0b10110001);
+ &pclmulqdq ("xmm0","xmm0",1);
+ &mov ("eax",&DWP(4,"esp"));
+ &movups (&QWP(0,"eax"),"xmm0");
+ &ret ();
+
+&set_label("mmx",16);
+}
+ &push ("ebp");
+ &push ("ebx");
+ &push ("esi");
+ &push ("edi");
+ &mov ($a,&wparam(1));
+ &mov ($b,&wparam(3));
+ &call ("_mul_1x1_mmx"); # a1·b1
+ &movq ("mm7",$R);
+
+ &mov ($a,&wparam(2));
+ &mov ($b,&wparam(4));
+ &call ("_mul_1x1_mmx"); # a0·b0
+ &movq ("mm6",$R);
+
+ &mov ($a,&wparam(1));
+ &mov ($b,&wparam(3));
+ &xor ($a,&wparam(2));
+ &xor ($b,&wparam(4));
+ &call ("_mul_1x1_mmx"); # (a0+a1)·(b0+b1)
+ &pxor ($R,"mm7");
+ &mov ($a,&wparam(0));
+ &pxor ($R,"mm6"); # (a0+a1)·(b0+b1)-a1·b1-a0·b0
+
+ &movq ($A,$R);
+ &psllq ($R,32);
+ &pop ("edi");
+ &psrlq ($A,32);
+ &pop ("esi");
+ &pxor ($R,"mm6");
+ &pop ("ebx");
+ &pxor ($A,"mm7");
+ &movq (&QWP(0,$a),$R);
+ &pop ("ebp");
+ &movq (&QWP(8,$a),$A);
+ &emms ();
+ &ret ();
+&set_label("ialu",16);
+}
+ &push ("ebp");
+ &push ("ebx");
+ &push ("esi");
+ &push ("edi");
+ &stack_push(4+1);
+
+ &mov ($a,&wparam(1));
+ &mov ($b,&wparam(3));
+ &call ("_mul_1x1_ialu"); # a1·b1
+ &mov (&DWP(8,"esp"),$lo);
+ &mov (&DWP(12,"esp"),$hi);
+
+ &mov ($a,&wparam(2));
+ &mov ($b,&wparam(4));
+ &call ("_mul_1x1_ialu"); # a0·b0
+ &mov (&DWP(0,"esp"),$lo);
+ &mov (&DWP(4,"esp"),$hi);
+
+ &mov ($a,&wparam(1));
+ &mov ($b,&wparam(3));
+ &xor ($a,&wparam(2));
+ &xor ($b,&wparam(4));
+ &call ("_mul_1x1_ialu"); # (a0+a1)·(b0+b1)
+
+ &mov ("ebp",&wparam(0));
+ @r=("ebx","ecx","edi","esi");
+ &mov (@r[0],&DWP(0,"esp"));
+ &mov (@r[1],&DWP(4,"esp"));
+ &mov (@r[2],&DWP(8,"esp"));
+ &mov (@r[3],&DWP(12,"esp"));
+
+ &xor ($lo,$hi);
+ &xor ($hi,@r[1]);
+ &xor ($lo,@r[0]);
+ &mov (&DWP(0,"ebp"),@r[0]);
+ &xor ($hi,@r[2]);
+ &mov (&DWP(12,"ebp"),@r[3]);
+ &xor ($lo,@r[3]);
+ &stack_pop(4+1);
+ &xor ($hi,@r[3]);
+ &pop ("edi");
+ &xor ($lo,$hi);
+ &pop ("esi");
+ &mov (&DWP(8,"ebp"),$hi);
+ &pop ("ebx");
+ &mov (&DWP(4,"ebp"),$lo);
+ &pop ("ebp");
+ &ret ();
+&function_end_B("bn_GF2m_mul_2x2");
+
+&asciz ("GF(2^m) Multiplication for x86, CRYPTOGAMS by <appro\@openssl.org>");
+
+&asm_finish();
+
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/x86-mont.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/x86-mont.pl
new file mode 100755
index 0000000..2103f80
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/x86-mont.pl
@@ -0,0 +1,631 @@
+#! /usr/bin/env perl
+# Copyright 2005-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+
+# October 2005
+#
+# This is a "teaser" code, as it can be improved in several ways...
+# First of all non-SSE2 path should be implemented (yes, for now it
+# performs Montgomery multiplication/convolution only on SSE2-capable
+# CPUs such as P4, others fall down to original code). Then inner loop
+# can be unrolled and modulo-scheduled to improve ILP and possibly
+# moved to 128-bit XMM register bank (though it would require input
+# rearrangement and/or increase bus bandwidth utilization). Dedicated
+# squaring procedure should give further performance improvement...
+# Yet, for being draft, the code improves rsa512 *sign* benchmark by
+# 110%(!), rsa1024 one - by 70% and rsa4096 - by 20%:-)
+
+# December 2006
+#
+# Modulo-scheduling SSE2 loops results in further 15-20% improvement.
+# Integer-only code [being equipped with dedicated squaring procedure]
+# gives ~40% on rsa512 sign benchmark...
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+push(@INC,"${dir}","${dir}../../perlasm");
+require "x86asm.pl";
+
+$output = pop;
+open STDOUT,">$output";
+
+&asm_init($ARGV[0]);
+
+$sse2=0;
+for (@ARGV) { $sse2=1 if (/-DOPENSSL_IA32_SSE2/); }
+
+&external_label("OPENSSL_ia32cap_P") if ($sse2);
+
+&function_begin("bn_mul_mont");
+
+$i="edx";
+$j="ecx";
+$ap="esi"; $tp="esi"; # overlapping variables!!!
+$rp="edi"; $bp="edi"; # overlapping variables!!!
+$np="ebp";
+$num="ebx";
+
+$_num=&DWP(4*0,"esp"); # stack top layout
+$_rp=&DWP(4*1,"esp");
+$_ap=&DWP(4*2,"esp");
+$_bp=&DWP(4*3,"esp");
+$_np=&DWP(4*4,"esp");
+$_n0=&DWP(4*5,"esp"); $_n0q=&QWP(4*5,"esp");
+$_sp=&DWP(4*6,"esp");
+$_bpend=&DWP(4*7,"esp");
+$frame=32; # size of above frame rounded up to 16n
+
+ &xor ("eax","eax");
+ &mov ("edi",&wparam(5)); # int num
+ &cmp ("edi",4);
+ &jl (&label("just_leave"));
+
+ &lea ("esi",&wparam(0)); # put aside pointer to argument block
+ &lea ("edx",&wparam(1)); # load ap
+ &add ("edi",2); # extra two words on top of tp
+ &neg ("edi");
+ &lea ("ebp",&DWP(-$frame,"esp","edi",4)); # future alloca($frame+4*(num+2))
+ &neg ("edi");
+
+ # minimize cache contention by arranging 2K window between stack
+ # pointer and ap argument [np is also position sensitive vector,
+ # but it's assumed to be near ap, as it's allocated at ~same
+ # time].
+ &mov ("eax","ebp");
+ &sub ("eax","edx");
+ &and ("eax",2047);
+ &sub ("ebp","eax"); # this aligns sp and ap modulo 2048
+
+ &xor ("edx","ebp");
+ &and ("edx",2048);
+ &xor ("edx",2048);
+ &sub ("ebp","edx"); # this splits them apart modulo 4096
+
+ &and ("ebp",-64); # align to cache line
+
+ # An OS-agnostic version of __chkstk.
+ #
+ # Some OSes (Windows) insist on stack being "wired" to
+ # physical memory in strictly sequential manner, i.e. if stack
+ # allocation spans two pages, then reference to farmost one can
+ # be punishable by SEGV. But page walking can do good even on
+ # other OSes, because it guarantees that villain thread hits
+ # the guard page before it can make damage to innocent one...
+ &mov ("eax","esp");
+ &sub ("eax","ebp");
+ &and ("eax",-4096);
+ &mov ("edx","esp"); # saved stack pointer!
+ &lea ("esp",&DWP(0,"ebp","eax"));
+ &mov ("eax",&DWP(0,"esp"));
+ &cmp ("esp","ebp");
+ &ja (&label("page_walk"));
+ &jmp (&label("page_walk_done"));
+
+&set_label("page_walk",16);
+ &lea ("esp",&DWP(-4096,"esp"));
+ &mov ("eax",&DWP(0,"esp"));
+ &cmp ("esp","ebp");
+ &ja (&label("page_walk"));
+&set_label("page_walk_done");
+
+ ################################# load argument block...
+ &mov ("eax",&DWP(0*4,"esi"));# BN_ULONG *rp
+ &mov ("ebx",&DWP(1*4,"esi"));# const BN_ULONG *ap
+ &mov ("ecx",&DWP(2*4,"esi"));# const BN_ULONG *bp
+ &mov ("ebp",&DWP(3*4,"esi"));# const BN_ULONG *np
+ &mov ("esi",&DWP(4*4,"esi"));# const BN_ULONG *n0
+ #&mov ("edi",&DWP(5*4,"esi"));# int num
+
+ &mov ("esi",&DWP(0,"esi")); # pull n0[0]
+ &mov ($_rp,"eax"); # ... save a copy of argument block
+ &mov ($_ap,"ebx");
+ &mov ($_bp,"ecx");
+ &mov ($_np,"ebp");
+ &mov ($_n0,"esi");
+ &lea ($num,&DWP(-3,"edi")); # num=num-1 to assist modulo-scheduling
+ #&mov ($_num,$num); # redundant as $num is not reused
+ &mov ($_sp,"edx"); # saved stack pointer!
+
+if($sse2) {
+$acc0="mm0"; # mmx register bank layout
+$acc1="mm1";
+$car0="mm2";
+$car1="mm3";
+$mul0="mm4";
+$mul1="mm5";
+$temp="mm6";
+$mask="mm7";
+
+ &picmeup("eax","OPENSSL_ia32cap_P");
+ &bt (&DWP(0,"eax"),26);
+ &jnc (&label("non_sse2"));
+
+ &mov ("eax",-1);
+ &movd ($mask,"eax"); # mask 32 lower bits
+
+ &mov ($ap,$_ap); # load input pointers
+ &mov ($bp,$_bp);
+ &mov ($np,$_np);
+
+ &xor ($i,$i); # i=0
+ &xor ($j,$j); # j=0
+
+ &movd ($mul0,&DWP(0,$bp)); # bp[0]
+ &movd ($mul1,&DWP(0,$ap)); # ap[0]
+ &movd ($car1,&DWP(0,$np)); # np[0]
+
+ &pmuludq($mul1,$mul0); # ap[0]*bp[0]
+ &movq ($car0,$mul1);
+ &movq ($acc0,$mul1); # I wish movd worked for
+ &pand ($acc0,$mask); # inter-register transfers
+
+ &pmuludq($mul1,$_n0q); # *=n0
+
+ &pmuludq($car1,$mul1); # "t[0]"*np[0]*n0
+ &paddq ($car1,$acc0);
+
+ &movd ($acc1,&DWP(4,$np)); # np[1]
+ &movd ($acc0,&DWP(4,$ap)); # ap[1]
+
+ &psrlq ($car0,32);
+ &psrlq ($car1,32);
+
+ &inc ($j); # j++
+&set_label("1st",16);
+ &pmuludq($acc0,$mul0); # ap[j]*bp[0]
+ &pmuludq($acc1,$mul1); # np[j]*m1
+ &paddq ($car0,$acc0); # +=c0
+ &paddq ($car1,$acc1); # +=c1
+
+ &movq ($acc0,$car0);
+ &pand ($acc0,$mask);
+ &movd ($acc1,&DWP(4,$np,$j,4)); # np[j+1]
+ &paddq ($car1,$acc0); # +=ap[j]*bp[0];
+ &movd ($acc0,&DWP(4,$ap,$j,4)); # ap[j+1]
+ &psrlq ($car0,32);
+ &movd (&DWP($frame-4,"esp",$j,4),$car1); # tp[j-1]=
+ &psrlq ($car1,32);
+
+ &lea ($j,&DWP(1,$j));
+ &cmp ($j,$num);
+ &jl (&label("1st"));
+
+ &pmuludq($acc0,$mul0); # ap[num-1]*bp[0]
+ &pmuludq($acc1,$mul1); # np[num-1]*m1
+ &paddq ($car0,$acc0); # +=c0
+ &paddq ($car1,$acc1); # +=c1
+
+ &movq ($acc0,$car0);
+ &pand ($acc0,$mask);
+ &paddq ($car1,$acc0); # +=ap[num-1]*bp[0];
+ &movd (&DWP($frame-4,"esp",$j,4),$car1); # tp[num-2]=
+
+ &psrlq ($car0,32);
+ &psrlq ($car1,32);
+
+ &paddq ($car1,$car0);
+ &movq (&QWP($frame,"esp",$num,4),$car1); # tp[num].tp[num-1]
+
+ &inc ($i); # i++
+&set_label("outer");
+ &xor ($j,$j); # j=0
+
+ &movd ($mul0,&DWP(0,$bp,$i,4)); # bp[i]
+ &movd ($mul1,&DWP(0,$ap)); # ap[0]
+ &movd ($temp,&DWP($frame,"esp")); # tp[0]
+ &movd ($car1,&DWP(0,$np)); # np[0]
+ &pmuludq($mul1,$mul0); # ap[0]*bp[i]
+
+ &paddq ($mul1,$temp); # +=tp[0]
+ &movq ($acc0,$mul1);
+ &movq ($car0,$mul1);
+ &pand ($acc0,$mask);
+
+ &pmuludq($mul1,$_n0q); # *=n0
+
+ &pmuludq($car1,$mul1);
+ &paddq ($car1,$acc0);
+
+ &movd ($temp,&DWP($frame+4,"esp")); # tp[1]
+ &movd ($acc1,&DWP(4,$np)); # np[1]
+ &movd ($acc0,&DWP(4,$ap)); # ap[1]
+
+ &psrlq ($car0,32);
+ &psrlq ($car1,32);
+ &paddq ($car0,$temp); # +=tp[1]
+
+ &inc ($j); # j++
+ &dec ($num);
+&set_label("inner");
+ &pmuludq($acc0,$mul0); # ap[j]*bp[i]
+ &pmuludq($acc1,$mul1); # np[j]*m1
+ &paddq ($car0,$acc0); # +=c0
+ &paddq ($car1,$acc1); # +=c1
+
+ &movq ($acc0,$car0);
+ &movd ($temp,&DWP($frame+4,"esp",$j,4));# tp[j+1]
+ &pand ($acc0,$mask);
+ &movd ($acc1,&DWP(4,$np,$j,4)); # np[j+1]
+ &paddq ($car1,$acc0); # +=ap[j]*bp[i]+tp[j]
+ &movd ($acc0,&DWP(4,$ap,$j,4)); # ap[j+1]
+ &psrlq ($car0,32);
+ &movd (&DWP($frame-4,"esp",$j,4),$car1);# tp[j-1]=
+ &psrlq ($car1,32);
+ &paddq ($car0,$temp); # +=tp[j+1]
+
+ &dec ($num);
+ &lea ($j,&DWP(1,$j)); # j++
+ &jnz (&label("inner"));
+
+ &mov ($num,$j);
+ &pmuludq($acc0,$mul0); # ap[num-1]*bp[i]
+ &pmuludq($acc1,$mul1); # np[num-1]*m1
+ &paddq ($car0,$acc0); # +=c0
+ &paddq ($car1,$acc1); # +=c1
+
+ &movq ($acc0,$car0);
+ &pand ($acc0,$mask);
+ &paddq ($car1,$acc0); # +=ap[num-1]*bp[i]+tp[num-1]
+ &movd (&DWP($frame-4,"esp",$j,4),$car1); # tp[num-2]=
+ &psrlq ($car0,32);
+ &psrlq ($car1,32);
+
+ &movd ($temp,&DWP($frame+4,"esp",$num,4)); # += tp[num]
+ &paddq ($car1,$car0);
+ &paddq ($car1,$temp);
+ &movq (&QWP($frame,"esp",$num,4),$car1); # tp[num].tp[num-1]
+
+ &lea ($i,&DWP(1,$i)); # i++
+ &cmp ($i,$num);
+ &jle (&label("outer"));
+
+ &emms (); # done with mmx bank
+ &jmp (&label("common_tail"));
+
+&set_label("non_sse2",16);
+}
+
+if (0) {
+ &mov ("esp",$_sp);
+ &xor ("eax","eax"); # signal "not fast enough [yet]"
+ &jmp (&label("just_leave"));
+ # While the below code provides competitive performance for
+ # all key lengths on modern Intel cores, it's still more
+ # than 10% slower for 4096-bit key elsewhere:-( "Competitive"
+ # means compared to the original integer-only assembler.
+ # 512-bit RSA sign is better by ~40%, but that's about all
+ # one can say about all CPUs...
+} else {
+$inp="esi"; # integer path uses these registers differently
+$word="edi";
+$carry="ebp";
+
+ &mov ($inp,$_ap);
+ &lea ($carry,&DWP(1,$num));
+ &mov ($word,$_bp);
+ &xor ($j,$j); # j=0
+ &mov ("edx",$inp);
+ &and ($carry,1); # see if num is even
+ &sub ("edx",$word); # see if ap==bp
+ &lea ("eax",&DWP(4,$word,$num,4)); # &bp[num]
+ &or ($carry,"edx");
+ &mov ($word,&DWP(0,$word)); # bp[0]
+ &jz (&label("bn_sqr_mont"));
+ &mov ($_bpend,"eax");
+ &mov ("eax",&DWP(0,$inp));
+ &xor ("edx","edx");
+
+&set_label("mull",16);
+ &mov ($carry,"edx");
+ &mul ($word); # ap[j]*bp[0]
+ &add ($carry,"eax");
+ &lea ($j,&DWP(1,$j));
+ &adc ("edx",0);
+ &mov ("eax",&DWP(0,$inp,$j,4)); # ap[j+1]
+ &cmp ($j,$num);
+ &mov (&DWP($frame-4,"esp",$j,4),$carry); # tp[j]=
+ &jl (&label("mull"));
+
+ &mov ($carry,"edx");
+ &mul ($word); # ap[num-1]*bp[0]
+ &mov ($word,$_n0);
+ &add ("eax",$carry);
+ &mov ($inp,$_np);
+ &adc ("edx",0);
+ &imul ($word,&DWP($frame,"esp")); # n0*tp[0]
+
+ &mov (&DWP($frame,"esp",$num,4),"eax"); # tp[num-1]=
+ &xor ($j,$j);
+ &mov (&DWP($frame+4,"esp",$num,4),"edx"); # tp[num]=
+ &mov (&DWP($frame+8,"esp",$num,4),$j); # tp[num+1]=
+
+ &mov ("eax",&DWP(0,$inp)); # np[0]
+ &mul ($word); # np[0]*m
+ &add ("eax",&DWP($frame,"esp")); # +=tp[0]
+ &mov ("eax",&DWP(4,$inp)); # np[1]
+ &adc ("edx",0);
+ &inc ($j);
+
+ &jmp (&label("2ndmadd"));
+
+&set_label("1stmadd",16);
+ &mov ($carry,"edx");
+ &mul ($word); # ap[j]*bp[i]
+ &add ($carry,&DWP($frame,"esp",$j,4)); # +=tp[j]
+ &lea ($j,&DWP(1,$j));
+ &adc ("edx",0);
+ &add ($carry,"eax");
+ &mov ("eax",&DWP(0,$inp,$j,4)); # ap[j+1]
+ &adc ("edx",0);
+ &cmp ($j,$num);
+ &mov (&DWP($frame-4,"esp",$j,4),$carry); # tp[j]=
+ &jl (&label("1stmadd"));
+
+ &mov ($carry,"edx");
+ &mul ($word); # ap[num-1]*bp[i]
+ &add ("eax",&DWP($frame,"esp",$num,4)); # +=tp[num-1]
+ &mov ($word,$_n0);
+ &adc ("edx",0);
+ &mov ($inp,$_np);
+ &add ($carry,"eax");
+ &adc ("edx",0);
+ &imul ($word,&DWP($frame,"esp")); # n0*tp[0]
+
+ &xor ($j,$j);
+ &add ("edx",&DWP($frame+4,"esp",$num,4)); # carry+=tp[num]
+ &mov (&DWP($frame,"esp",$num,4),$carry); # tp[num-1]=
+ &adc ($j,0);
+ &mov ("eax",&DWP(0,$inp)); # np[0]
+ &mov (&DWP($frame+4,"esp",$num,4),"edx"); # tp[num]=
+ &mov (&DWP($frame+8,"esp",$num,4),$j); # tp[num+1]=
+
+ &mul ($word); # np[0]*m
+ &add ("eax",&DWP($frame,"esp")); # +=tp[0]
+ &mov ("eax",&DWP(4,$inp)); # np[1]
+ &adc ("edx",0);
+ &mov ($j,1);
+
+&set_label("2ndmadd",16);
+ &mov ($carry,"edx");
+ &mul ($word); # np[j]*m
+ &add ($carry,&DWP($frame,"esp",$j,4)); # +=tp[j]
+ &lea ($j,&DWP(1,$j));
+ &adc ("edx",0);
+ &add ($carry,"eax");
+ &mov ("eax",&DWP(0,$inp,$j,4)); # np[j+1]
+ &adc ("edx",0);
+ &cmp ($j,$num);
+ &mov (&DWP($frame-8,"esp",$j,4),$carry); # tp[j-1]=
+ &jl (&label("2ndmadd"));
+
+ &mov ($carry,"edx");
+ &mul ($word); # np[j]*m
+ &add ($carry,&DWP($frame,"esp",$num,4)); # +=tp[num-1]
+ &adc ("edx",0);
+ &add ($carry,"eax");
+ &adc ("edx",0);
+ &mov (&DWP($frame-4,"esp",$num,4),$carry); # tp[num-2]=
+
+ &xor ("eax","eax");
+ &mov ($j,$_bp); # &bp[i]
+ &add ("edx",&DWP($frame+4,"esp",$num,4)); # carry+=tp[num]
+ &adc ("eax",&DWP($frame+8,"esp",$num,4)); # +=tp[num+1]
+ &lea ($j,&DWP(4,$j));
+ &mov (&DWP($frame,"esp",$num,4),"edx"); # tp[num-1]=
+ &cmp ($j,$_bpend);
+ &mov (&DWP($frame+4,"esp",$num,4),"eax"); # tp[num]=
+ &je (&label("common_tail"));
+
+ &mov ($word,&DWP(0,$j)); # bp[i+1]
+ &mov ($inp,$_ap);
+ &mov ($_bp,$j); # &bp[++i]
+ &xor ($j,$j);
+ &xor ("edx","edx");
+ &mov ("eax",&DWP(0,$inp));
+ &jmp (&label("1stmadd"));
+
+&set_label("bn_sqr_mont",16);
+$sbit=$num;
+ &mov ($_num,$num);
+ &mov ($_bp,$j); # i=0
+
+ &mov ("eax",$word); # ap[0]
+ &mul ($word); # ap[0]*ap[0]
+ &mov (&DWP($frame,"esp"),"eax"); # tp[0]=
+ &mov ($sbit,"edx");
+ &shr ("edx",1);
+ &and ($sbit,1);
+ &inc ($j);
+&set_label("sqr",16);
+ &mov ("eax",&DWP(0,$inp,$j,4)); # ap[j]
+ &mov ($carry,"edx");
+ &mul ($word); # ap[j]*ap[0]
+ &add ("eax",$carry);
+ &lea ($j,&DWP(1,$j));
+ &adc ("edx",0);
+ &lea ($carry,&DWP(0,$sbit,"eax",2));
+ &shr ("eax",31);
+ &cmp ($j,$_num);
+ &mov ($sbit,"eax");
+ &mov (&DWP($frame-4,"esp",$j,4),$carry); # tp[j]=
+ &jl (&label("sqr"));
+
+ &mov ("eax",&DWP(0,$inp,$j,4)); # ap[num-1]
+ &mov ($carry,"edx");
+ &mul ($word); # ap[num-1]*ap[0]
+ &add ("eax",$carry);
+ &mov ($word,$_n0);
+ &adc ("edx",0);
+ &mov ($inp,$_np);
+ &lea ($carry,&DWP(0,$sbit,"eax",2));
+ &imul ($word,&DWP($frame,"esp")); # n0*tp[0]
+ &shr ("eax",31);
+ &mov (&DWP($frame,"esp",$j,4),$carry); # tp[num-1]=
+
+ &lea ($carry,&DWP(0,"eax","edx",2));
+ &mov ("eax",&DWP(0,$inp)); # np[0]
+ &shr ("edx",31);
+ &mov (&DWP($frame+4,"esp",$j,4),$carry); # tp[num]=
+ &mov (&DWP($frame+8,"esp",$j,4),"edx"); # tp[num+1]=
+
+ &mul ($word); # np[0]*m
+ &add ("eax",&DWP($frame,"esp")); # +=tp[0]
+ &mov ($num,$j);
+ &adc ("edx",0);
+ &mov ("eax",&DWP(4,$inp)); # np[1]
+ &mov ($j,1);
+
+&set_label("3rdmadd",16);
+ &mov ($carry,"edx");
+ &mul ($word); # np[j]*m
+ &add ($carry,&DWP($frame,"esp",$j,4)); # +=tp[j]
+ &adc ("edx",0);
+ &add ($carry,"eax");
+ &mov ("eax",&DWP(4,$inp,$j,4)); # np[j+1]
+ &adc ("edx",0);
+ &mov (&DWP($frame-4,"esp",$j,4),$carry); # tp[j-1]=
+
+ &mov ($carry,"edx");
+ &mul ($word); # np[j+1]*m
+ &add ($carry,&DWP($frame+4,"esp",$j,4)); # +=tp[j+1]
+ &lea ($j,&DWP(2,$j));
+ &adc ("edx",0);
+ &add ($carry,"eax");
+ &mov ("eax",&DWP(0,$inp,$j,4)); # np[j+2]
+ &adc ("edx",0);
+ &cmp ($j,$num);
+ &mov (&DWP($frame-8,"esp",$j,4),$carry); # tp[j]=
+ &jl (&label("3rdmadd"));
+
+ &mov ($carry,"edx");
+ &mul ($word); # np[j]*m
+ &add ($carry,&DWP($frame,"esp",$num,4)); # +=tp[num-1]
+ &adc ("edx",0);
+ &add ($carry,"eax");
+ &adc ("edx",0);
+ &mov (&DWP($frame-4,"esp",$num,4),$carry); # tp[num-2]=
+
+ &mov ($j,$_bp); # i
+ &xor ("eax","eax");
+ &mov ($inp,$_ap);
+ &add ("edx",&DWP($frame+4,"esp",$num,4)); # carry+=tp[num]
+ &adc ("eax",&DWP($frame+8,"esp",$num,4)); # +=tp[num+1]
+ &mov (&DWP($frame,"esp",$num,4),"edx"); # tp[num-1]=
+ &cmp ($j,$num);
+ &mov (&DWP($frame+4,"esp",$num,4),"eax"); # tp[num]=
+ &je (&label("common_tail"));
+
+ &mov ($word,&DWP(4,$inp,$j,4)); # ap[i]
+ &lea ($j,&DWP(1,$j));
+ &mov ("eax",$word);
+ &mov ($_bp,$j); # ++i
+ &mul ($word); # ap[i]*ap[i]
+ &add ("eax",&DWP($frame,"esp",$j,4)); # +=tp[i]
+ &adc ("edx",0);
+ &mov (&DWP($frame,"esp",$j,4),"eax"); # tp[i]=
+ &xor ($carry,$carry);
+ &cmp ($j,$num);
+ &lea ($j,&DWP(1,$j));
+ &je (&label("sqrlast"));
+
+ &mov ($sbit,"edx"); # zaps $num
+ &shr ("edx",1);
+ &and ($sbit,1);
+&set_label("sqradd",16);
+ &mov ("eax",&DWP(0,$inp,$j,4)); # ap[j]
+ &mov ($carry,"edx");
+ &mul ($word); # ap[j]*ap[i]
+ &add ("eax",$carry);
+ &lea ($carry,&DWP(0,"eax","eax"));
+ &adc ("edx",0);
+ &shr ("eax",31);
+ &add ($carry,&DWP($frame,"esp",$j,4)); # +=tp[j]
+ &lea ($j,&DWP(1,$j));
+ &adc ("eax",0);
+ &add ($carry,$sbit);
+ &adc ("eax",0);
+ &cmp ($j,$_num);
+ &mov (&DWP($frame-4,"esp",$j,4),$carry); # tp[j]=
+ &mov ($sbit,"eax");
+ &jle (&label("sqradd"));
+
+ &mov ($carry,"edx");
+ &add ("edx","edx");
+ &shr ($carry,31);
+ &add ("edx",$sbit);
+ &adc ($carry,0);
+&set_label("sqrlast");
+ &mov ($word,$_n0);
+ &mov ($inp,$_np);
+ &imul ($word,&DWP($frame,"esp")); # n0*tp[0]
+
+ &add ("edx",&DWP($frame,"esp",$j,4)); # +=tp[num]
+ &mov ("eax",&DWP(0,$inp)); # np[0]
+ &adc ($carry,0);
+ &mov (&DWP($frame,"esp",$j,4),"edx"); # tp[num]=
+ &mov (&DWP($frame+4,"esp",$j,4),$carry); # tp[num+1]=
+
+ &mul ($word); # np[0]*m
+ &add ("eax",&DWP($frame,"esp")); # +=tp[0]
+ &lea ($num,&DWP(-1,$j));
+ &adc ("edx",0);
+ &mov ($j,1);
+ &mov ("eax",&DWP(4,$inp)); # np[1]
+
+ &jmp (&label("3rdmadd"));
+}
+
+&set_label("common_tail",16);
+ &mov ($np,$_np); # load modulus pointer
+ &mov ($rp,$_rp); # load result pointer
+ &lea ($tp,&DWP($frame,"esp")); # [$ap and $bp are zapped]
+
+ &mov ("eax",&DWP(0,$tp)); # tp[0]
+ &mov ($j,$num); # j=num-1
+ &xor ($i,$i); # i=0 and clear CF!
+
+&set_label("sub",16);
+ &sbb ("eax",&DWP(0,$np,$i,4));
+ &mov (&DWP(0,$rp,$i,4),"eax"); # rp[i]=tp[i]-np[i]
+ &dec ($j); # doesn't affect CF!
+ &mov ("eax",&DWP(4,$tp,$i,4)); # tp[i+1]
+ &lea ($i,&DWP(1,$i)); # i++
+ &jge (&label("sub"));
+
+ &sbb ("eax",0); # handle upmost overflow bit
+ &mov ("edx",-1);
+ &xor ("edx","eax");
+ &jmp (&label("copy"));
+
+&set_label("copy",16); # conditional copy
+ &mov ($tp,&DWP($frame,"esp",$num,4));
+ &mov ($np,&DWP(0,$rp,$num,4));
+ &mov (&DWP($frame,"esp",$num,4),$j); # zap temporary vector
+ &and ($tp,"eax");
+ &and ($np,"edx");
+ &or ($np,$tp);
+ &mov (&DWP(0,$rp,$num,4),$np);
+ &dec ($num);
+ &jge (&label("copy"));
+
+ &mov ("esp",$_sp); # pull saved stack pointer
+ &mov ("eax",1);
+&set_label("just_leave");
+&function_end("bn_mul_mont");
+
+&asciz("Montgomery Multiplication for x86, CRYPTOGAMS by <appro\@openssl.org>");
+
+&asm_finish();
+
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/x86_64-gcc.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/x86_64-gcc.c
new file mode 100644
index 0000000..e6fdaad
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/x86_64-gcc.c
@@ -0,0 +1,643 @@
+/*
+ * Copyright 2002-2018 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include "../bn_local.h"
+#if !(defined(__GNUC__) && __GNUC__>=2)
+# include "../bn_asm.c" /* kind of dirty hack for Sun Studio */
+#else
+/*-
+ * x86_64 BIGNUM accelerator version 0.1, December 2002.
+ *
+ * Implemented by Andy Polyakov <appro@openssl.org> for the OpenSSL
+ * project.
+ *
+ * Rights for redistribution and usage in source and binary forms are
+ * granted according to the OpenSSL license. Warranty of any kind is
+ * disclaimed.
+ *
+ * Q. Version 0.1? It doesn't sound like Andy, he used to assign real
+ * versions, like 1.0...
+ * A. Well, that's because this code is basically a quick-n-dirty
+ * proof-of-concept hack. As you can see it's implemented with
+ * inline assembler, which means that you're bound to GCC and that
+ * there might be enough room for further improvement.
+ *
+ * Q. Why inline assembler?
+ * A. x86_64 features own ABI which I'm not familiar with. This is
+ * why I decided to let the compiler take care of subroutine
+ * prologue/epilogue as well as register allocation. For reference.
+ * Win64 implements different ABI for AMD64, different from Linux.
+ *
+ * Q. How much faster does it get?
+ * A. 'apps/openssl speed rsa dsa' output with no-asm:
+ *
+ * sign verify sign/s verify/s
+ * rsa 512 bits 0.0006s 0.0001s 1683.8 18456.2
+ * rsa 1024 bits 0.0028s 0.0002s 356.0 6407.0
+ * rsa 2048 bits 0.0172s 0.0005s 58.0 1957.8
+ * rsa 4096 bits 0.1155s 0.0018s 8.7 555.6
+ * sign verify sign/s verify/s
+ * dsa 512 bits 0.0005s 0.0006s 2100.8 1768.3
+ * dsa 1024 bits 0.0014s 0.0018s 692.3 559.2
+ * dsa 2048 bits 0.0049s 0.0061s 204.7 165.0
+ *
+ * 'apps/openssl speed rsa dsa' output with this module:
+ *
+ * sign verify sign/s verify/s
+ * rsa 512 bits 0.0004s 0.0000s 2767.1 33297.9
+ * rsa 1024 bits 0.0012s 0.0001s 867.4 14674.7
+ * rsa 2048 bits 0.0061s 0.0002s 164.0 5270.0
+ * rsa 4096 bits 0.0384s 0.0006s 26.1 1650.8
+ * sign verify sign/s verify/s
+ * dsa 512 bits 0.0002s 0.0003s 4442.2 3786.3
+ * dsa 1024 bits 0.0005s 0.0007s 1835.1 1497.4
+ * dsa 2048 bits 0.0016s 0.0020s 620.4 504.6
+ *
+ * For the reference. IA-32 assembler implementation performs
+ * very much like 64-bit code compiled with no-asm on the same
+ * machine.
+ */
+
+# undef mul
+# undef mul_add
+
+/*-
+ * "m"(a), "+m"(r) is the way to favor DirectPath µ-code;
+ * "g"(0) let the compiler to decide where does it
+ * want to keep the value of zero;
+ */
+# define mul_add(r,a,word,carry) do { \
+ register BN_ULONG high,low; \
+ asm ("mulq %3" \
+ : "=a"(low),"=d"(high) \
+ : "a"(word),"m"(a) \
+ : "cc"); \
+ asm ("addq %2,%0; adcq %3,%1" \
+ : "+r"(carry),"+d"(high)\
+ : "a"(low),"g"(0) \
+ : "cc"); \
+ asm ("addq %2,%0; adcq %3,%1" \
+ : "+m"(r),"+d"(high) \
+ : "r"(carry),"g"(0) \
+ : "cc"); \
+ carry=high; \
+ } while (0)
+
+# define mul(r,a,word,carry) do { \
+ register BN_ULONG high,low; \
+ asm ("mulq %3" \
+ : "=a"(low),"=d"(high) \
+ : "a"(word),"g"(a) \
+ : "cc"); \
+ asm ("addq %2,%0; adcq %3,%1" \
+ : "+r"(carry),"+d"(high)\
+ : "a"(low),"g"(0) \
+ : "cc"); \
+ (r)=carry, carry=high; \
+ } while (0)
+# undef sqr
+# define sqr(r0,r1,a) \
+ asm ("mulq %2" \
+ : "=a"(r0),"=d"(r1) \
+ : "a"(a) \
+ : "cc");
+
+BN_ULONG bn_mul_add_words(BN_ULONG *rp, const BN_ULONG *ap, int num,
+ BN_ULONG w)
+{
+ BN_ULONG c1 = 0;
+
+ if (num <= 0)
+ return c1;
+
+ while (num & ~3) {
+ mul_add(rp[0], ap[0], w, c1);
+ mul_add(rp[1], ap[1], w, c1);
+ mul_add(rp[2], ap[2], w, c1);
+ mul_add(rp[3], ap[3], w, c1);
+ ap += 4;
+ rp += 4;
+ num -= 4;
+ }
+ if (num) {
+ mul_add(rp[0], ap[0], w, c1);
+ if (--num == 0)
+ return c1;
+ mul_add(rp[1], ap[1], w, c1);
+ if (--num == 0)
+ return c1;
+ mul_add(rp[2], ap[2], w, c1);
+ return c1;
+ }
+
+ return c1;
+}
+
+BN_ULONG bn_mul_words(BN_ULONG *rp, const BN_ULONG *ap, int num, BN_ULONG w)
+{
+ BN_ULONG c1 = 0;
+
+ if (num <= 0)
+ return c1;
+
+ while (num & ~3) {
+ mul(rp[0], ap[0], w, c1);
+ mul(rp[1], ap[1], w, c1);
+ mul(rp[2], ap[2], w, c1);
+ mul(rp[3], ap[3], w, c1);
+ ap += 4;
+ rp += 4;
+ num -= 4;
+ }
+ if (num) {
+ mul(rp[0], ap[0], w, c1);
+ if (--num == 0)
+ return c1;
+ mul(rp[1], ap[1], w, c1);
+ if (--num == 0)
+ return c1;
+ mul(rp[2], ap[2], w, c1);
+ }
+ return c1;
+}
+
+void bn_sqr_words(BN_ULONG *r, const BN_ULONG *a, int n)
+{
+ if (n <= 0)
+ return;
+
+ while (n & ~3) {
+ sqr(r[0], r[1], a[0]);
+ sqr(r[2], r[3], a[1]);
+ sqr(r[4], r[5], a[2]);
+ sqr(r[6], r[7], a[3]);
+ a += 4;
+ r += 8;
+ n -= 4;
+ }
+ if (n) {
+ sqr(r[0], r[1], a[0]);
+ if (--n == 0)
+ return;
+ sqr(r[2], r[3], a[1]);
+ if (--n == 0)
+ return;
+ sqr(r[4], r[5], a[2]);
+ }
+}
+
+BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d)
+{
+ BN_ULONG ret, waste;
+
+ asm("divq %4":"=a"(ret), "=d"(waste)
+ : "a"(l), "d"(h), "r"(d)
+ : "cc");
+
+ return ret;
+}
+
+BN_ULONG bn_add_words(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
+ int n)
+{
+ BN_ULONG ret;
+ size_t i = 0;
+
+ if (n <= 0)
+ return 0;
+
+ asm volatile (" subq %0,%0 \n" /* clear carry */
+ " jmp 1f \n"
+ ".p2align 4 \n"
+ "1: movq (%4,%2,8),%0 \n"
+ " adcq (%5,%2,8),%0 \n"
+ " movq %0,(%3,%2,8) \n"
+ " lea 1(%2),%2 \n"
+ " dec %1 \n"
+ " jnz 1b \n"
+ " sbbq %0,%0 \n"
+ :"=&r" (ret), "+c"(n), "+r"(i)
+ :"r"(rp), "r"(ap), "r"(bp)
+ :"cc", "memory");
+
+ return ret & 1;
+}
+
+# ifndef SIMICS
+BN_ULONG bn_sub_words(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
+ int n)
+{
+ BN_ULONG ret;
+ size_t i = 0;
+
+ if (n <= 0)
+ return 0;
+
+ asm volatile (" subq %0,%0 \n" /* clear borrow */
+ " jmp 1f \n"
+ ".p2align 4 \n"
+ "1: movq (%4,%2,8),%0 \n"
+ " sbbq (%5,%2,8),%0 \n"
+ " movq %0,(%3,%2,8) \n"
+ " lea 1(%2),%2 \n"
+ " dec %1 \n"
+ " jnz 1b \n"
+ " sbbq %0,%0 \n"
+ :"=&r" (ret), "+c"(n), "+r"(i)
+ :"r"(rp), "r"(ap), "r"(bp)
+ :"cc", "memory");
+
+ return ret & 1;
+}
+# else
+/* Simics 1.4<7 has buggy sbbq:-( */
+# define BN_MASK2 0xffffffffffffffffL
+BN_ULONG bn_sub_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
+{
+ BN_ULONG t1, t2;
+ int c = 0;
+
+ if (n <= 0)
+ return (BN_ULONG)0;
+
+ for (;;) {
+ t1 = a[0];
+ t2 = b[0];
+ r[0] = (t1 - t2 - c) & BN_MASK2;
+ if (t1 != t2)
+ c = (t1 < t2);
+ if (--n <= 0)
+ break;
+
+ t1 = a[1];
+ t2 = b[1];
+ r[1] = (t1 - t2 - c) & BN_MASK2;
+ if (t1 != t2)
+ c = (t1 < t2);
+ if (--n <= 0)
+ break;
+
+ t1 = a[2];
+ t2 = b[2];
+ r[2] = (t1 - t2 - c) & BN_MASK2;
+ if (t1 != t2)
+ c = (t1 < t2);
+ if (--n <= 0)
+ break;
+
+ t1 = a[3];
+ t2 = b[3];
+ r[3] = (t1 - t2 - c) & BN_MASK2;
+ if (t1 != t2)
+ c = (t1 < t2);
+ if (--n <= 0)
+ break;
+
+ a += 4;
+ b += 4;
+ r += 4;
+ }
+ return c;
+}
+# endif
+
+/* mul_add_c(a,b,c0,c1,c2) -- c+=a*b for three word number c=(c2,c1,c0) */
+/* mul_add_c2(a,b,c0,c1,c2) -- c+=2*a*b for three word number c=(c2,c1,c0) */
+/* sqr_add_c(a,i,c0,c1,c2) -- c+=a[i]^2 for three word number c=(c2,c1,c0) */
+/*
+ * sqr_add_c2(a,i,c0,c1,c2) -- c+=2*a[i]*a[j] for three word number
+ * c=(c2,c1,c0)
+ */
+
+/*
+ * Keep in mind that carrying into high part of multiplication result
+ * can not overflow, because it cannot be all-ones.
+ */
+# if 0
+/* original macros are kept for reference purposes */
+# define mul_add_c(a,b,c0,c1,c2) do { \
+ BN_ULONG ta = (a), tb = (b); \
+ BN_ULONG lo, hi; \
+ BN_UMULT_LOHI(lo,hi,ta,tb); \
+ c0 += lo; hi += (c0<lo)?1:0; \
+ c1 += hi; c2 += (c1<hi)?1:0; \
+ } while(0)
+
+# define mul_add_c2(a,b,c0,c1,c2) do { \
+ BN_ULONG ta = (a), tb = (b); \
+ BN_ULONG lo, hi, tt; \
+ BN_UMULT_LOHI(lo,hi,ta,tb); \
+ c0 += lo; tt = hi+((c0<lo)?1:0); \
+ c1 += tt; c2 += (c1<tt)?1:0; \
+ c0 += lo; hi += (c0<lo)?1:0; \
+ c1 += hi; c2 += (c1<hi)?1:0; \
+ } while(0)
+
+# define sqr_add_c(a,i,c0,c1,c2) do { \
+ BN_ULONG ta = (a)[i]; \
+ BN_ULONG lo, hi; \
+ BN_UMULT_LOHI(lo,hi,ta,ta); \
+ c0 += lo; hi += (c0<lo)?1:0; \
+ c1 += hi; c2 += (c1<hi)?1:0; \
+ } while(0)
+# else
+# define mul_add_c(a,b,c0,c1,c2) do { \
+ BN_ULONG t1,t2; \
+ asm ("mulq %3" \
+ : "=a"(t1),"=d"(t2) \
+ : "a"(a),"m"(b) \
+ : "cc"); \
+ asm ("addq %3,%0; adcq %4,%1; adcq %5,%2" \
+ : "+r"(c0),"+r"(c1),"+r"(c2) \
+ : "r"(t1),"r"(t2),"g"(0) \
+ : "cc"); \
+ } while (0)
+
+# define sqr_add_c(a,i,c0,c1,c2) do { \
+ BN_ULONG t1,t2; \
+ asm ("mulq %2" \
+ : "=a"(t1),"=d"(t2) \
+ : "a"(a[i]) \
+ : "cc"); \
+ asm ("addq %3,%0; adcq %4,%1; adcq %5,%2" \
+ : "+r"(c0),"+r"(c1),"+r"(c2) \
+ : "r"(t1),"r"(t2),"g"(0) \
+ : "cc"); \
+ } while (0)
+
+# define mul_add_c2(a,b,c0,c1,c2) do { \
+ BN_ULONG t1,t2; \
+ asm ("mulq %3" \
+ : "=a"(t1),"=d"(t2) \
+ : "a"(a),"m"(b) \
+ : "cc"); \
+ asm ("addq %3,%0; adcq %4,%1; adcq %5,%2" \
+ : "+r"(c0),"+r"(c1),"+r"(c2) \
+ : "r"(t1),"r"(t2),"g"(0) \
+ : "cc"); \
+ asm ("addq %3,%0; adcq %4,%1; adcq %5,%2" \
+ : "+r"(c0),"+r"(c1),"+r"(c2) \
+ : "r"(t1),"r"(t2),"g"(0) \
+ : "cc"); \
+ } while (0)
+# endif
+
+# define sqr_add_c2(a,i,j,c0,c1,c2) \
+ mul_add_c2((a)[i],(a)[j],c0,c1,c2)
+
+void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
+{
+ BN_ULONG c1, c2, c3;
+
+ c1 = 0;
+ c2 = 0;
+ c3 = 0;
+ mul_add_c(a[0], b[0], c1, c2, c3);
+ r[0] = c1;
+ c1 = 0;
+ mul_add_c(a[0], b[1], c2, c3, c1);
+ mul_add_c(a[1], b[0], c2, c3, c1);
+ r[1] = c2;
+ c2 = 0;
+ mul_add_c(a[2], b[0], c3, c1, c2);
+ mul_add_c(a[1], b[1], c3, c1, c2);
+ mul_add_c(a[0], b[2], c3, c1, c2);
+ r[2] = c3;
+ c3 = 0;
+ mul_add_c(a[0], b[3], c1, c2, c3);
+ mul_add_c(a[1], b[2], c1, c2, c3);
+ mul_add_c(a[2], b[1], c1, c2, c3);
+ mul_add_c(a[3], b[0], c1, c2, c3);
+ r[3] = c1;
+ c1 = 0;
+ mul_add_c(a[4], b[0], c2, c3, c1);
+ mul_add_c(a[3], b[1], c2, c3, c1);
+ mul_add_c(a[2], b[2], c2, c3, c1);
+ mul_add_c(a[1], b[3], c2, c3, c1);
+ mul_add_c(a[0], b[4], c2, c3, c1);
+ r[4] = c2;
+ c2 = 0;
+ mul_add_c(a[0], b[5], c3, c1, c2);
+ mul_add_c(a[1], b[4], c3, c1, c2);
+ mul_add_c(a[2], b[3], c3, c1, c2);
+ mul_add_c(a[3], b[2], c3, c1, c2);
+ mul_add_c(a[4], b[1], c3, c1, c2);
+ mul_add_c(a[5], b[0], c3, c1, c2);
+ r[5] = c3;
+ c3 = 0;
+ mul_add_c(a[6], b[0], c1, c2, c3);
+ mul_add_c(a[5], b[1], c1, c2, c3);
+ mul_add_c(a[4], b[2], c1, c2, c3);
+ mul_add_c(a[3], b[3], c1, c2, c3);
+ mul_add_c(a[2], b[4], c1, c2, c3);
+ mul_add_c(a[1], b[5], c1, c2, c3);
+ mul_add_c(a[0], b[6], c1, c2, c3);
+ r[6] = c1;
+ c1 = 0;
+ mul_add_c(a[0], b[7], c2, c3, c1);
+ mul_add_c(a[1], b[6], c2, c3, c1);
+ mul_add_c(a[2], b[5], c2, c3, c1);
+ mul_add_c(a[3], b[4], c2, c3, c1);
+ mul_add_c(a[4], b[3], c2, c3, c1);
+ mul_add_c(a[5], b[2], c2, c3, c1);
+ mul_add_c(a[6], b[1], c2, c3, c1);
+ mul_add_c(a[7], b[0], c2, c3, c1);
+ r[7] = c2;
+ c2 = 0;
+ mul_add_c(a[7], b[1], c3, c1, c2);
+ mul_add_c(a[6], b[2], c3, c1, c2);
+ mul_add_c(a[5], b[3], c3, c1, c2);
+ mul_add_c(a[4], b[4], c3, c1, c2);
+ mul_add_c(a[3], b[5], c3, c1, c2);
+ mul_add_c(a[2], b[6], c3, c1, c2);
+ mul_add_c(a[1], b[7], c3, c1, c2);
+ r[8] = c3;
+ c3 = 0;
+ mul_add_c(a[2], b[7], c1, c2, c3);
+ mul_add_c(a[3], b[6], c1, c2, c3);
+ mul_add_c(a[4], b[5], c1, c2, c3);
+ mul_add_c(a[5], b[4], c1, c2, c3);
+ mul_add_c(a[6], b[3], c1, c2, c3);
+ mul_add_c(a[7], b[2], c1, c2, c3);
+ r[9] = c1;
+ c1 = 0;
+ mul_add_c(a[7], b[3], c2, c3, c1);
+ mul_add_c(a[6], b[4], c2, c3, c1);
+ mul_add_c(a[5], b[5], c2, c3, c1);
+ mul_add_c(a[4], b[6], c2, c3, c1);
+ mul_add_c(a[3], b[7], c2, c3, c1);
+ r[10] = c2;
+ c2 = 0;
+ mul_add_c(a[4], b[7], c3, c1, c2);
+ mul_add_c(a[5], b[6], c3, c1, c2);
+ mul_add_c(a[6], b[5], c3, c1, c2);
+ mul_add_c(a[7], b[4], c3, c1, c2);
+ r[11] = c3;
+ c3 = 0;
+ mul_add_c(a[7], b[5], c1, c2, c3);
+ mul_add_c(a[6], b[6], c1, c2, c3);
+ mul_add_c(a[5], b[7], c1, c2, c3);
+ r[12] = c1;
+ c1 = 0;
+ mul_add_c(a[6], b[7], c2, c3, c1);
+ mul_add_c(a[7], b[6], c2, c3, c1);
+ r[13] = c2;
+ c2 = 0;
+ mul_add_c(a[7], b[7], c3, c1, c2);
+ r[14] = c3;
+ r[15] = c1;
+}
+
+void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
+{
+ BN_ULONG c1, c2, c3;
+
+ c1 = 0;
+ c2 = 0;
+ c3 = 0;
+ mul_add_c(a[0], b[0], c1, c2, c3);
+ r[0] = c1;
+ c1 = 0;
+ mul_add_c(a[0], b[1], c2, c3, c1);
+ mul_add_c(a[1], b[0], c2, c3, c1);
+ r[1] = c2;
+ c2 = 0;
+ mul_add_c(a[2], b[0], c3, c1, c2);
+ mul_add_c(a[1], b[1], c3, c1, c2);
+ mul_add_c(a[0], b[2], c3, c1, c2);
+ r[2] = c3;
+ c3 = 0;
+ mul_add_c(a[0], b[3], c1, c2, c3);
+ mul_add_c(a[1], b[2], c1, c2, c3);
+ mul_add_c(a[2], b[1], c1, c2, c3);
+ mul_add_c(a[3], b[0], c1, c2, c3);
+ r[3] = c1;
+ c1 = 0;
+ mul_add_c(a[3], b[1], c2, c3, c1);
+ mul_add_c(a[2], b[2], c2, c3, c1);
+ mul_add_c(a[1], b[3], c2, c3, c1);
+ r[4] = c2;
+ c2 = 0;
+ mul_add_c(a[2], b[3], c3, c1, c2);
+ mul_add_c(a[3], b[2], c3, c1, c2);
+ r[5] = c3;
+ c3 = 0;
+ mul_add_c(a[3], b[3], c1, c2, c3);
+ r[6] = c1;
+ r[7] = c2;
+}
+
+void bn_sqr_comba8(BN_ULONG *r, const BN_ULONG *a)
+{
+ BN_ULONG c1, c2, c3;
+
+ c1 = 0;
+ c2 = 0;
+ c3 = 0;
+ sqr_add_c(a, 0, c1, c2, c3);
+ r[0] = c1;
+ c1 = 0;
+ sqr_add_c2(a, 1, 0, c2, c3, c1);
+ r[1] = c2;
+ c2 = 0;
+ sqr_add_c(a, 1, c3, c1, c2);
+ sqr_add_c2(a, 2, 0, c3, c1, c2);
+ r[2] = c3;
+ c3 = 0;
+ sqr_add_c2(a, 3, 0, c1, c2, c3);
+ sqr_add_c2(a, 2, 1, c1, c2, c3);
+ r[3] = c1;
+ c1 = 0;
+ sqr_add_c(a, 2, c2, c3, c1);
+ sqr_add_c2(a, 3, 1, c2, c3, c1);
+ sqr_add_c2(a, 4, 0, c2, c3, c1);
+ r[4] = c2;
+ c2 = 0;
+ sqr_add_c2(a, 5, 0, c3, c1, c2);
+ sqr_add_c2(a, 4, 1, c3, c1, c2);
+ sqr_add_c2(a, 3, 2, c3, c1, c2);
+ r[5] = c3;
+ c3 = 0;
+ sqr_add_c(a, 3, c1, c2, c3);
+ sqr_add_c2(a, 4, 2, c1, c2, c3);
+ sqr_add_c2(a, 5, 1, c1, c2, c3);
+ sqr_add_c2(a, 6, 0, c1, c2, c3);
+ r[6] = c1;
+ c1 = 0;
+ sqr_add_c2(a, 7, 0, c2, c3, c1);
+ sqr_add_c2(a, 6, 1, c2, c3, c1);
+ sqr_add_c2(a, 5, 2, c2, c3, c1);
+ sqr_add_c2(a, 4, 3, c2, c3, c1);
+ r[7] = c2;
+ c2 = 0;
+ sqr_add_c(a, 4, c3, c1, c2);
+ sqr_add_c2(a, 5, 3, c3, c1, c2);
+ sqr_add_c2(a, 6, 2, c3, c1, c2);
+ sqr_add_c2(a, 7, 1, c3, c1, c2);
+ r[8] = c3;
+ c3 = 0;
+ sqr_add_c2(a, 7, 2, c1, c2, c3);
+ sqr_add_c2(a, 6, 3, c1, c2, c3);
+ sqr_add_c2(a, 5, 4, c1, c2, c3);
+ r[9] = c1;
+ c1 = 0;
+ sqr_add_c(a, 5, c2, c3, c1);
+ sqr_add_c2(a, 6, 4, c2, c3, c1);
+ sqr_add_c2(a, 7, 3, c2, c3, c1);
+ r[10] = c2;
+ c2 = 0;
+ sqr_add_c2(a, 7, 4, c3, c1, c2);
+ sqr_add_c2(a, 6, 5, c3, c1, c2);
+ r[11] = c3;
+ c3 = 0;
+ sqr_add_c(a, 6, c1, c2, c3);
+ sqr_add_c2(a, 7, 5, c1, c2, c3);
+ r[12] = c1;
+ c1 = 0;
+ sqr_add_c2(a, 7, 6, c2, c3, c1);
+ r[13] = c2;
+ c2 = 0;
+ sqr_add_c(a, 7, c3, c1, c2);
+ r[14] = c3;
+ r[15] = c1;
+}
+
+void bn_sqr_comba4(BN_ULONG *r, const BN_ULONG *a)
+{
+ BN_ULONG c1, c2, c3;
+
+ c1 = 0;
+ c2 = 0;
+ c3 = 0;
+ sqr_add_c(a, 0, c1, c2, c3);
+ r[0] = c1;
+ c1 = 0;
+ sqr_add_c2(a, 1, 0, c2, c3, c1);
+ r[1] = c2;
+ c2 = 0;
+ sqr_add_c(a, 1, c3, c1, c2);
+ sqr_add_c2(a, 2, 0, c3, c1, c2);
+ r[2] = c3;
+ c3 = 0;
+ sqr_add_c2(a, 3, 0, c1, c2, c3);
+ sqr_add_c2(a, 2, 1, c1, c2, c3);
+ r[3] = c1;
+ c1 = 0;
+ sqr_add_c(a, 2, c2, c3, c1);
+ sqr_add_c2(a, 3, 1, c2, c3, c1);
+ r[4] = c2;
+ c2 = 0;
+ sqr_add_c2(a, 3, 2, c3, c1, c2);
+ r[5] = c3;
+ c3 = 0;
+ sqr_add_c(a, 3, c1, c2, c3);
+ r[6] = c1;
+ r[7] = c2;
+}
+#endif
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/x86_64-gf2m.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/x86_64-gf2m.pl
new file mode 100644
index 0000000..655f13c
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/x86_64-gf2m.pl
@@ -0,0 +1,424 @@
+#! /usr/bin/env perl
+# Copyright 2011-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+#
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+#
+# May 2011
+#
+# The module implements bn_GF2m_mul_2x2 polynomial multiplication used
+# in bn_gf2m.c. It's kind of low-hanging mechanical port from C for
+# the time being... Except that it has two code paths: code suitable
+# for any x86_64 CPU and PCLMULQDQ one suitable for Westmere and
+# later. Improvement varies from one benchmark and µ-arch to another.
+# Vanilla code path is at most 20% faster than compiler-generated code
+# [not very impressive], while PCLMULQDQ - whole 85%-160% better on
+# 163- and 571-bit ECDH benchmarks on Intel CPUs. Keep in mind that
+# these coefficients are not ones for bn_GF2m_mul_2x2 itself, as not
+# all CPU time is burnt in it...
+
+$flavour = shift;
+$output = shift;
+if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
+
+$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or
+( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or
+die "can't locate x86_64-xlate.pl";
+
+open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\"";
+*STDOUT=*OUT;
+
+($lo,$hi)=("%rax","%rdx"); $a=$lo;
+($i0,$i1)=("%rsi","%rdi");
+($t0,$t1)=("%rbx","%rcx");
+($b,$mask)=("%rbp","%r8");
+($a1,$a2,$a4,$a8,$a12,$a48)=map("%r$_",(9..15));
+($R,$Tx)=("%xmm0","%xmm1");
+
+$code.=<<___;
+.text
+
+.type _mul_1x1,\@abi-omnipotent
+.align 16
+_mul_1x1:
+.cfi_startproc
+ sub \$128+8,%rsp
+.cfi_adjust_cfa_offset 128+8
+ mov \$-1,$a1
+ lea ($a,$a),$i0
+ shr \$3,$a1
+ lea (,$a,4),$i1
+ and $a,$a1 # a1=a&0x1fffffffffffffff
+ lea (,$a,8),$a8
+ sar \$63,$a # broadcast 63rd bit
+ lea ($a1,$a1),$a2
+ sar \$63,$i0 # broadcast 62nd bit
+ lea (,$a1,4),$a4
+ and $b,$a
+ sar \$63,$i1 # broadcast 61st bit
+ mov $a,$hi # $a is $lo
+ shl \$63,$lo
+ and $b,$i0
+ shr \$1,$hi
+ mov $i0,$t1
+ shl \$62,$i0
+ and $b,$i1
+ shr \$2,$t1
+ xor $i0,$lo
+ mov $i1,$t0
+ shl \$61,$i1
+ xor $t1,$hi
+ shr \$3,$t0
+ xor $i1,$lo
+ xor $t0,$hi
+
+ mov $a1,$a12
+ movq \$0,0(%rsp) # tab[0]=0
+ xor $a2,$a12 # a1^a2
+ mov $a1,8(%rsp) # tab[1]=a1
+ mov $a4,$a48
+ mov $a2,16(%rsp) # tab[2]=a2
+ xor $a8,$a48 # a4^a8
+ mov $a12,24(%rsp) # tab[3]=a1^a2
+
+ xor $a4,$a1
+ mov $a4,32(%rsp) # tab[4]=a4
+ xor $a4,$a2
+ mov $a1,40(%rsp) # tab[5]=a1^a4
+ xor $a4,$a12
+ mov $a2,48(%rsp) # tab[6]=a2^a4
+ xor $a48,$a1 # a1^a4^a4^a8=a1^a8
+ mov $a12,56(%rsp) # tab[7]=a1^a2^a4
+ xor $a48,$a2 # a2^a4^a4^a8=a1^a8
+
+ mov $a8,64(%rsp) # tab[8]=a8
+ xor $a48,$a12 # a1^a2^a4^a4^a8=a1^a2^a8
+ mov $a1,72(%rsp) # tab[9]=a1^a8
+ xor $a4,$a1 # a1^a8^a4
+ mov $a2,80(%rsp) # tab[10]=a2^a8
+ xor $a4,$a2 # a2^a8^a4
+ mov $a12,88(%rsp) # tab[11]=a1^a2^a8
+
+ xor $a4,$a12 # a1^a2^a8^a4
+ mov $a48,96(%rsp) # tab[12]=a4^a8
+ mov $mask,$i0
+ mov $a1,104(%rsp) # tab[13]=a1^a4^a8
+ and $b,$i0
+ mov $a2,112(%rsp) # tab[14]=a2^a4^a8
+ shr \$4,$b
+ mov $a12,120(%rsp) # tab[15]=a1^a2^a4^a8
+ mov $mask,$i1
+ and $b,$i1
+ shr \$4,$b
+
+ movq (%rsp,$i0,8),$R # half of calculations is done in SSE2
+ mov $mask,$i0
+ and $b,$i0
+ shr \$4,$b
+___
+ for ($n=1;$n<8;$n++) {
+ $code.=<<___;
+ mov (%rsp,$i1,8),$t1
+ mov $mask,$i1
+ mov $t1,$t0
+ shl \$`8*$n-4`,$t1
+ and $b,$i1
+ movq (%rsp,$i0,8),$Tx
+ shr \$`64-(8*$n-4)`,$t0
+ xor $t1,$lo
+ pslldq \$$n,$Tx
+ mov $mask,$i0
+ shr \$4,$b
+ xor $t0,$hi
+ and $b,$i0
+ shr \$4,$b
+ pxor $Tx,$R
+___
+ }
+$code.=<<___;
+ mov (%rsp,$i1,8),$t1
+ mov $t1,$t0
+ shl \$`8*$n-4`,$t1
+ movq $R,$i0
+ shr \$`64-(8*$n-4)`,$t0
+ xor $t1,$lo
+ psrldq \$8,$R
+ xor $t0,$hi
+ movq $R,$i1
+ xor $i0,$lo
+ xor $i1,$hi
+
+ add \$128+8,%rsp
+.cfi_adjust_cfa_offset -128-8
+ ret
+.Lend_mul_1x1:
+.cfi_endproc
+.size _mul_1x1,.-_mul_1x1
+___
+
+($rp,$a1,$a0,$b1,$b0) = $win64? ("%rcx","%rdx","%r8", "%r9","%r10") : # Win64 order
+ ("%rdi","%rsi","%rdx","%rcx","%r8"); # Unix order
+
+$code.=<<___;
+.extern OPENSSL_ia32cap_P
+.globl bn_GF2m_mul_2x2
+.type bn_GF2m_mul_2x2,\@abi-omnipotent
+.align 16
+bn_GF2m_mul_2x2:
+.cfi_startproc
+ mov %rsp,%rax
+ mov OPENSSL_ia32cap_P(%rip),%r10
+ bt \$33,%r10
+ jnc .Lvanilla_mul_2x2
+
+ movq $a1,%xmm0
+ movq $b1,%xmm1
+ movq $a0,%xmm2
+___
+$code.=<<___ if ($win64);
+ movq 40(%rsp),%xmm3
+___
+$code.=<<___ if (!$win64);
+ movq $b0,%xmm3
+___
+$code.=<<___;
+ movdqa %xmm0,%xmm4
+ movdqa %xmm1,%xmm5
+ pclmulqdq \$0,%xmm1,%xmm0 # a1·b1
+ pxor %xmm2,%xmm4
+ pxor %xmm3,%xmm5
+ pclmulqdq \$0,%xmm3,%xmm2 # a0·b0
+ pclmulqdq \$0,%xmm5,%xmm4 # (a0+a1)·(b0+b1)
+ xorps %xmm0,%xmm4
+ xorps %xmm2,%xmm4 # (a0+a1)·(b0+b1)-a0·b0-a1·b1
+ movdqa %xmm4,%xmm5
+ pslldq \$8,%xmm4
+ psrldq \$8,%xmm5
+ pxor %xmm4,%xmm2
+ pxor %xmm5,%xmm0
+ movdqu %xmm2,0($rp)
+ movdqu %xmm0,16($rp)
+ ret
+
+.align 16
+.Lvanilla_mul_2x2:
+ lea -8*17(%rsp),%rsp
+.cfi_adjust_cfa_offset 8*17
+___
+$code.=<<___ if ($win64);
+ mov `8*17+40`(%rsp),$b0
+ mov %rdi,8*15(%rsp)
+ mov %rsi,8*16(%rsp)
+___
+$code.=<<___;
+ mov %r14,8*10(%rsp)
+.cfi_rel_offset %r14,8*10
+ mov %r13,8*11(%rsp)
+.cfi_rel_offset %r13,8*11
+ mov %r12,8*12(%rsp)
+.cfi_rel_offset %r12,8*12
+ mov %rbp,8*13(%rsp)
+.cfi_rel_offset %rbp,8*13
+ mov %rbx,8*14(%rsp)
+.cfi_rel_offset %rbx,8*14
+.Lbody_mul_2x2:
+ mov $rp,32(%rsp) # save the arguments
+ mov $a1,40(%rsp)
+ mov $a0,48(%rsp)
+ mov $b1,56(%rsp)
+ mov $b0,64(%rsp)
+
+ mov \$0xf,$mask
+ mov $a1,$a
+ mov $b1,$b
+ call _mul_1x1 # a1·b1
+ mov $lo,16(%rsp)
+ mov $hi,24(%rsp)
+
+ mov 48(%rsp),$a
+ mov 64(%rsp),$b
+ call _mul_1x1 # a0·b0
+ mov $lo,0(%rsp)
+ mov $hi,8(%rsp)
+
+ mov 40(%rsp),$a
+ mov 56(%rsp),$b
+ xor 48(%rsp),$a
+ xor 64(%rsp),$b
+ call _mul_1x1 # (a0+a1)·(b0+b1)
+___
+ @r=("%rbx","%rcx","%rdi","%rsi");
+$code.=<<___;
+ mov 0(%rsp),@r[0]
+ mov 8(%rsp),@r[1]
+ mov 16(%rsp),@r[2]
+ mov 24(%rsp),@r[3]
+ mov 32(%rsp),%rbp
+
+ xor $hi,$lo
+ xor @r[1],$hi
+ xor @r[0],$lo
+ mov @r[0],0(%rbp)
+ xor @r[2],$hi
+ mov @r[3],24(%rbp)
+ xor @r[3],$lo
+ xor @r[3],$hi
+ xor $hi,$lo
+ mov $hi,16(%rbp)
+ mov $lo,8(%rbp)
+
+ mov 8*10(%rsp),%r14
+.cfi_restore %r14
+ mov 8*11(%rsp),%r13
+.cfi_restore %r13
+ mov 8*12(%rsp),%r12
+.cfi_restore %r12
+ mov 8*13(%rsp),%rbp
+.cfi_restore %rbp
+ mov 8*14(%rsp),%rbx
+.cfi_restore %rbx
+___
+$code.=<<___ if ($win64);
+ mov 8*15(%rsp),%rdi
+ mov 8*16(%rsp),%rsi
+___
+$code.=<<___;
+ lea 8*17(%rsp),%rsp
+.cfi_adjust_cfa_offset -8*17
+.Lepilogue_mul_2x2:
+ ret
+.Lend_mul_2x2:
+.cfi_endproc
+.size bn_GF2m_mul_2x2,.-bn_GF2m_mul_2x2
+.asciz "GF(2^m) Multiplication for x86_64, CRYPTOGAMS by <appro\@openssl.org>"
+.align 16
+___
+
+# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame,
+# CONTEXT *context,DISPATCHER_CONTEXT *disp)
+if ($win64) {
+$rec="%rcx";
+$frame="%rdx";
+$context="%r8";
+$disp="%r9";
+
+$code.=<<___;
+.extern __imp_RtlVirtualUnwind
+
+.type se_handler,\@abi-omnipotent
+.align 16
+se_handler:
+ push %rsi
+ push %rdi
+ push %rbx
+ push %rbp
+ push %r12
+ push %r13
+ push %r14
+ push %r15
+ pushfq
+ sub \$64,%rsp
+
+ mov 120($context),%rax # pull context->Rax
+ mov 248($context),%rbx # pull context->Rip
+
+ lea .Lbody_mul_2x2(%rip),%r10
+ cmp %r10,%rbx # context->Rip<"prologue" label
+ jb .Lin_prologue
+
+ mov 152($context),%rax # pull context->Rsp
+
+ lea .Lepilogue_mul_2x2(%rip),%r10
+ cmp %r10,%rbx # context->Rip>="epilogue" label
+ jae .Lin_prologue
+
+ mov 8*10(%rax),%r14 # mimic epilogue
+ mov 8*11(%rax),%r13
+ mov 8*12(%rax),%r12
+ mov 8*13(%rax),%rbp
+ mov 8*14(%rax),%rbx
+ mov 8*15(%rax),%rdi
+ mov 8*16(%rax),%rsi
+
+ mov %rbx,144($context) # restore context->Rbx
+ mov %rbp,160($context) # restore context->Rbp
+ mov %rsi,168($context) # restore context->Rsi
+ mov %rdi,176($context) # restore context->Rdi
+ mov %r12,216($context) # restore context->R12
+ mov %r13,224($context) # restore context->R13
+ mov %r14,232($context) # restore context->R14
+
+ lea 8*17(%rax),%rax
+
+.Lin_prologue:
+ mov %rax,152($context) # restore context->Rsp
+
+ mov 40($disp),%rdi # disp->ContextRecord
+ mov $context,%rsi # context
+ mov \$154,%ecx # sizeof(CONTEXT)
+ .long 0xa548f3fc # cld; rep movsq
+
+ mov $disp,%rsi
+ xor %rcx,%rcx # arg1, UNW_FLAG_NHANDLER
+ mov 8(%rsi),%rdx # arg2, disp->ImageBase
+ mov 0(%rsi),%r8 # arg3, disp->ControlPc
+ mov 16(%rsi),%r9 # arg4, disp->FunctionEntry
+ mov 40(%rsi),%r10 # disp->ContextRecord
+ lea 56(%rsi),%r11 # &disp->HandlerData
+ lea 24(%rsi),%r12 # &disp->EstablisherFrame
+ mov %r10,32(%rsp) # arg5
+ mov %r11,40(%rsp) # arg6
+ mov %r12,48(%rsp) # arg7
+ mov %rcx,56(%rsp) # arg8, (NULL)
+ call *__imp_RtlVirtualUnwind(%rip)
+
+ mov \$1,%eax # ExceptionContinueSearch
+ add \$64,%rsp
+ popfq
+ pop %r15
+ pop %r14
+ pop %r13
+ pop %r12
+ pop %rbp
+ pop %rbx
+ pop %rdi
+ pop %rsi
+ ret
+.size se_handler,.-se_handler
+
+.section .pdata
+.align 4
+ .rva _mul_1x1
+ .rva .Lend_mul_1x1
+ .rva .LSEH_info_1x1
+
+ .rva .Lvanilla_mul_2x2
+ .rva .Lend_mul_2x2
+ .rva .LSEH_info_2x2
+.section .xdata
+.align 8
+.LSEH_info_1x1:
+ .byte 0x01,0x07,0x02,0x00
+ .byte 0x07,0x01,0x11,0x00 # sub rsp,128+8
+.LSEH_info_2x2:
+ .byte 9,0,0,0
+ .rva se_handler
+___
+}
+
+$code =~ s/\`([^\`]*)\`/eval($1)/gem;
+print $code;
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/x86_64-mont.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/x86_64-mont.pl
new file mode 100755
index 0000000..f14d4e6
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/x86_64-mont.pl
@@ -0,0 +1,1592 @@
+#! /usr/bin/env perl
+# Copyright 2005-2020 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+
+# October 2005.
+#
+# Montgomery multiplication routine for x86_64. While it gives modest
+# 9% improvement of rsa4096 sign on Opteron, rsa512 sign runs more
+# than twice, >2x, as fast. Most common rsa1024 sign is improved by
+# respectful 50%. It remains to be seen if loop unrolling and
+# dedicated squaring routine can provide further improvement...
+
+# July 2011.
+#
+# Add dedicated squaring procedure. Performance improvement varies
+# from platform to platform, but in average it's ~5%/15%/25%/33%
+# for 512-/1024-/2048-/4096-bit RSA *sign* benchmarks respectively.
+
+# August 2011.
+#
+# Unroll and modulo-schedule inner loops in such manner that they
+# are "fallen through" for input lengths of 8, which is critical for
+# 1024-bit RSA *sign*. Average performance improvement in comparison
+# to *initial* version of this module from 2005 is ~0%/30%/40%/45%
+# for 512-/1024-/2048-/4096-bit RSA *sign* benchmarks respectively.
+
+# June 2013.
+#
+# Optimize reduction in squaring procedure and improve 1024+-bit RSA
+# sign performance by 10-16% on Intel Sandy Bridge and later
+# (virtually same on non-Intel processors).
+
+# August 2013.
+#
+# Add MULX/ADOX/ADCX code path.
+
+$flavour = shift;
+$output = shift;
+if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
+
+$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or
+( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or
+die "can't locate x86_64-xlate.pl";
+
+open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\"";
+*STDOUT=*OUT;
+
+if (`$ENV{CC} -Wa,-v -c -o /dev/null -x assembler /dev/null 2>&1`
+ =~ /GNU assembler version ([2-9]\.[0-9]+)/) {
+ $addx = ($1>=2.23);
+}
+
+if (!$addx && $win64 && ($flavour =~ /nasm/ || $ENV{ASM} =~ /nasm/) &&
+ `nasm -v 2>&1` =~ /NASM version ([2-9]\.[0-9]+)/) {
+ $addx = ($1>=2.10);
+}
+
+if (!$addx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
+ `ml64 2>&1` =~ /Version ([0-9]+)\./) {
+ $addx = ($1>=12);
+}
+
+if (!$addx && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|.*based on LLVM) ([0-9]+)\.([0-9]+)/) {
+ my $ver = $2 + $3/100.0; # 3.1->3.01, 3.10->3.10
+ $addx = ($ver>=3.03);
+}
+
+# int bn_mul_mont(
+$rp="%rdi"; # BN_ULONG *rp,
+$ap="%rsi"; # const BN_ULONG *ap,
+$bp="%rdx"; # const BN_ULONG *bp,
+$np="%rcx"; # const BN_ULONG *np,
+$n0="%r8"; # const BN_ULONG *n0,
+$num="%r9"; # int num);
+$lo0="%r10";
+$hi0="%r11";
+$hi1="%r13";
+$i="%r14";
+$j="%r15";
+$m0="%rbx";
+$m1="%rbp";
+
+$code=<<___;
+.text
+
+.extern OPENSSL_ia32cap_P
+
+.globl bn_mul_mont
+.type bn_mul_mont,\@function,6
+.align 16
+bn_mul_mont:
+.cfi_startproc
+ mov ${num}d,${num}d
+ mov %rsp,%rax
+.cfi_def_cfa_register %rax
+ test \$3,${num}d
+ jnz .Lmul_enter
+ cmp \$8,${num}d
+ jb .Lmul_enter
+___
+$code.=<<___ if ($addx);
+ mov OPENSSL_ia32cap_P+8(%rip),%r11d
+___
+$code.=<<___;
+ cmp $ap,$bp
+ jne .Lmul4x_enter
+ test \$7,${num}d
+ jz .Lsqr8x_enter
+ jmp .Lmul4x_enter
+
+.align 16
+.Lmul_enter:
+ push %rbx
+.cfi_push %rbx
+ push %rbp
+.cfi_push %rbp
+ push %r12
+.cfi_push %r12
+ push %r13
+.cfi_push %r13
+ push %r14
+.cfi_push %r14
+ push %r15
+.cfi_push %r15
+
+ neg $num
+ mov %rsp,%r11
+ lea -16(%rsp,$num,8),%r10 # future alloca(8*(num+2))
+ neg $num # restore $num
+ and \$-1024,%r10 # minimize TLB usage
+
+ # An OS-agnostic version of __chkstk.
+ #
+ # Some OSes (Windows) insist on stack being "wired" to
+ # physical memory in strictly sequential manner, i.e. if stack
+ # allocation spans two pages, then reference to farmost one can
+ # be punishable by SEGV. But page walking can do good even on
+ # other OSes, because it guarantees that villain thread hits
+ # the guard page before it can make damage to innocent one...
+ sub %r10,%r11
+ and \$-4096,%r11
+ lea (%r10,%r11),%rsp
+ mov (%rsp),%r11
+ cmp %r10,%rsp
+ ja .Lmul_page_walk
+ jmp .Lmul_page_walk_done
+
+.align 16
+.Lmul_page_walk:
+ lea -4096(%rsp),%rsp
+ mov (%rsp),%r11
+ cmp %r10,%rsp
+ ja .Lmul_page_walk
+.Lmul_page_walk_done:
+
+ mov %rax,8(%rsp,$num,8) # tp[num+1]=%rsp
+.cfi_cfa_expression %rsp+8,$num,8,mul,plus,deref,+8
+.Lmul_body:
+ mov $bp,%r12 # reassign $bp
+___
+ $bp="%r12";
+$code.=<<___;
+ mov ($n0),$n0 # pull n0[0] value
+ mov ($bp),$m0 # m0=bp[0]
+ mov ($ap),%rax
+
+ xor $i,$i # i=0
+ xor $j,$j # j=0
+
+ mov $n0,$m1
+ mulq $m0 # ap[0]*bp[0]
+ mov %rax,$lo0
+ mov ($np),%rax
+
+ imulq $lo0,$m1 # "tp[0]"*n0
+ mov %rdx,$hi0
+
+ mulq $m1 # np[0]*m1
+ add %rax,$lo0 # discarded
+ mov 8($ap),%rax
+ adc \$0,%rdx
+ mov %rdx,$hi1
+
+ lea 1($j),$j # j++
+ jmp .L1st_enter
+
+.align 16
+.L1st:
+ add %rax,$hi1
+ mov ($ap,$j,8),%rax
+ adc \$0,%rdx
+ add $hi0,$hi1 # np[j]*m1+ap[j]*bp[0]
+ mov $lo0,$hi0
+ adc \$0,%rdx
+ mov $hi1,-16(%rsp,$j,8) # tp[j-1]
+ mov %rdx,$hi1
+
+.L1st_enter:
+ mulq $m0 # ap[j]*bp[0]
+ add %rax,$hi0
+ mov ($np,$j,8),%rax
+ adc \$0,%rdx
+ lea 1($j),$j # j++
+ mov %rdx,$lo0
+
+ mulq $m1 # np[j]*m1
+ cmp $num,$j
+ jne .L1st
+
+ add %rax,$hi1
+ mov ($ap),%rax # ap[0]
+ adc \$0,%rdx
+ add $hi0,$hi1 # np[j]*m1+ap[j]*bp[0]
+ adc \$0,%rdx
+ mov $hi1,-16(%rsp,$j,8) # tp[j-1]
+ mov %rdx,$hi1
+ mov $lo0,$hi0
+
+ xor %rdx,%rdx
+ add $hi0,$hi1
+ adc \$0,%rdx
+ mov $hi1,-8(%rsp,$num,8)
+ mov %rdx,(%rsp,$num,8) # store upmost overflow bit
+
+ lea 1($i),$i # i++
+ jmp .Louter
+.align 16
+.Louter:
+ mov ($bp,$i,8),$m0 # m0=bp[i]
+ xor $j,$j # j=0
+ mov $n0,$m1
+ mov (%rsp),$lo0
+ mulq $m0 # ap[0]*bp[i]
+ add %rax,$lo0 # ap[0]*bp[i]+tp[0]
+ mov ($np),%rax
+ adc \$0,%rdx
+
+ imulq $lo0,$m1 # tp[0]*n0
+ mov %rdx,$hi0
+
+ mulq $m1 # np[0]*m1
+ add %rax,$lo0 # discarded
+ mov 8($ap),%rax
+ adc \$0,%rdx
+ mov 8(%rsp),$lo0 # tp[1]
+ mov %rdx,$hi1
+
+ lea 1($j),$j # j++
+ jmp .Linner_enter
+
+.align 16
+.Linner:
+ add %rax,$hi1
+ mov ($ap,$j,8),%rax
+ adc \$0,%rdx
+ add $lo0,$hi1 # np[j]*m1+ap[j]*bp[i]+tp[j]
+ mov (%rsp,$j,8),$lo0
+ adc \$0,%rdx
+ mov $hi1,-16(%rsp,$j,8) # tp[j-1]
+ mov %rdx,$hi1
+
+.Linner_enter:
+ mulq $m0 # ap[j]*bp[i]
+ add %rax,$hi0
+ mov ($np,$j,8),%rax
+ adc \$0,%rdx
+ add $hi0,$lo0 # ap[j]*bp[i]+tp[j]
+ mov %rdx,$hi0
+ adc \$0,$hi0
+ lea 1($j),$j # j++
+
+ mulq $m1 # np[j]*m1
+ cmp $num,$j
+ jne .Linner
+
+ add %rax,$hi1
+ mov ($ap),%rax # ap[0]
+ adc \$0,%rdx
+ add $lo0,$hi1 # np[j]*m1+ap[j]*bp[i]+tp[j]
+ mov (%rsp,$j,8),$lo0
+ adc \$0,%rdx
+ mov $hi1,-16(%rsp,$j,8) # tp[j-1]
+ mov %rdx,$hi1
+
+ xor %rdx,%rdx
+ add $hi0,$hi1
+ adc \$0,%rdx
+ add $lo0,$hi1 # pull upmost overflow bit
+ adc \$0,%rdx
+ mov $hi1,-8(%rsp,$num,8)
+ mov %rdx,(%rsp,$num,8) # store upmost overflow bit
+
+ lea 1($i),$i # i++
+ cmp $num,$i
+ jb .Louter
+
+ xor $i,$i # i=0 and clear CF!
+ mov (%rsp),%rax # tp[0]
+ mov $num,$j # j=num
+
+.align 16
+.Lsub: sbb ($np,$i,8),%rax
+ mov %rax,($rp,$i,8) # rp[i]=tp[i]-np[i]
+ mov 8(%rsp,$i,8),%rax # tp[i+1]
+ lea 1($i),$i # i++
+ dec $j # doesn't affect CF!
+ jnz .Lsub
+
+ sbb \$0,%rax # handle upmost overflow bit
+ mov \$-1,%rbx
+ xor %rax,%rbx # not %rax
+ xor $i,$i
+ mov $num,$j # j=num
+
+.Lcopy: # conditional copy
+ mov ($rp,$i,8),%rcx
+ mov (%rsp,$i,8),%rdx
+ and %rbx,%rcx
+ and %rax,%rdx
+ mov $num,(%rsp,$i,8) # zap temporary vector
+ or %rcx,%rdx
+ mov %rdx,($rp,$i,8) # rp[i]=tp[i]
+ lea 1($i),$i
+ sub \$1,$j
+ jnz .Lcopy
+
+ mov 8(%rsp,$num,8),%rsi # restore %rsp
+.cfi_def_cfa %rsi,8
+ mov \$1,%rax
+ mov -48(%rsi),%r15
+.cfi_restore %r15
+ mov -40(%rsi),%r14
+.cfi_restore %r14
+ mov -32(%rsi),%r13
+.cfi_restore %r13
+ mov -24(%rsi),%r12
+.cfi_restore %r12
+ mov -16(%rsi),%rbp
+.cfi_restore %rbp
+ mov -8(%rsi),%rbx
+.cfi_restore %rbx
+ lea (%rsi),%rsp
+.cfi_def_cfa_register %rsp
+.Lmul_epilogue:
+ ret
+.cfi_endproc
+.size bn_mul_mont,.-bn_mul_mont
+___
+{{{
+my @A=("%r10","%r11");
+my @N=("%r13","%rdi");
+$code.=<<___;
+.type bn_mul4x_mont,\@function,6
+.align 16
+bn_mul4x_mont:
+.cfi_startproc
+ mov ${num}d,${num}d
+ mov %rsp,%rax
+.cfi_def_cfa_register %rax
+.Lmul4x_enter:
+___
+$code.=<<___ if ($addx);
+ and \$0x80100,%r11d
+ cmp \$0x80100,%r11d
+ je .Lmulx4x_enter
+___
+$code.=<<___;
+ push %rbx
+.cfi_push %rbx
+ push %rbp
+.cfi_push %rbp
+ push %r12
+.cfi_push %r12
+ push %r13
+.cfi_push %r13
+ push %r14
+.cfi_push %r14
+ push %r15
+.cfi_push %r15
+
+ neg $num
+ mov %rsp,%r11
+ lea -32(%rsp,$num,8),%r10 # future alloca(8*(num+4))
+ neg $num # restore
+ and \$-1024,%r10 # minimize TLB usage
+
+ sub %r10,%r11
+ and \$-4096,%r11
+ lea (%r10,%r11),%rsp
+ mov (%rsp),%r11
+ cmp %r10,%rsp
+ ja .Lmul4x_page_walk
+ jmp .Lmul4x_page_walk_done
+
+.Lmul4x_page_walk:
+ lea -4096(%rsp),%rsp
+ mov (%rsp),%r11
+ cmp %r10,%rsp
+ ja .Lmul4x_page_walk
+.Lmul4x_page_walk_done:
+
+ mov %rax,8(%rsp,$num,8) # tp[num+1]=%rsp
+.cfi_cfa_expression %rsp+8,$num,8,mul,plus,deref,+8
+.Lmul4x_body:
+ mov $rp,16(%rsp,$num,8) # tp[num+2]=$rp
+ mov %rdx,%r12 # reassign $bp
+___
+ $bp="%r12";
+$code.=<<___;
+ mov ($n0),$n0 # pull n0[0] value
+ mov ($bp),$m0 # m0=bp[0]
+ mov ($ap),%rax
+
+ xor $i,$i # i=0
+ xor $j,$j # j=0
+
+ mov $n0,$m1
+ mulq $m0 # ap[0]*bp[0]
+ mov %rax,$A[0]
+ mov ($np),%rax
+
+ imulq $A[0],$m1 # "tp[0]"*n0
+ mov %rdx,$A[1]
+
+ mulq $m1 # np[0]*m1
+ add %rax,$A[0] # discarded
+ mov 8($ap),%rax
+ adc \$0,%rdx
+ mov %rdx,$N[1]
+
+ mulq $m0
+ add %rax,$A[1]
+ mov 8($np),%rax
+ adc \$0,%rdx
+ mov %rdx,$A[0]
+
+ mulq $m1
+ add %rax,$N[1]
+ mov 16($ap),%rax
+ adc \$0,%rdx
+ add $A[1],$N[1]
+ lea 4($j),$j # j++
+ adc \$0,%rdx
+ mov $N[1],(%rsp)
+ mov %rdx,$N[0]
+ jmp .L1st4x
+.align 16
+.L1st4x:
+ mulq $m0 # ap[j]*bp[0]
+ add %rax,$A[0]
+ mov -16($np,$j,8),%rax
+ adc \$0,%rdx
+ mov %rdx,$A[1]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[0]
+ mov -8($ap,$j,8),%rax
+ adc \$0,%rdx
+ add $A[0],$N[0] # np[j]*m1+ap[j]*bp[0]
+ adc \$0,%rdx
+ mov $N[0],-24(%rsp,$j,8) # tp[j-1]
+ mov %rdx,$N[1]
+
+ mulq $m0 # ap[j]*bp[0]
+ add %rax,$A[1]
+ mov -8($np,$j,8),%rax
+ adc \$0,%rdx
+ mov %rdx,$A[0]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[1]
+ mov ($ap,$j,8),%rax
+ adc \$0,%rdx
+ add $A[1],$N[1] # np[j]*m1+ap[j]*bp[0]
+ adc \$0,%rdx
+ mov $N[1],-16(%rsp,$j,8) # tp[j-1]
+ mov %rdx,$N[0]
+
+ mulq $m0 # ap[j]*bp[0]
+ add %rax,$A[0]
+ mov ($np,$j,8),%rax
+ adc \$0,%rdx
+ mov %rdx,$A[1]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[0]
+ mov 8($ap,$j,8),%rax
+ adc \$0,%rdx
+ add $A[0],$N[0] # np[j]*m1+ap[j]*bp[0]
+ adc \$0,%rdx
+ mov $N[0],-8(%rsp,$j,8) # tp[j-1]
+ mov %rdx,$N[1]
+
+ mulq $m0 # ap[j]*bp[0]
+ add %rax,$A[1]
+ mov 8($np,$j,8),%rax
+ adc \$0,%rdx
+ lea 4($j),$j # j++
+ mov %rdx,$A[0]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[1]
+ mov -16($ap,$j,8),%rax
+ adc \$0,%rdx
+ add $A[1],$N[1] # np[j]*m1+ap[j]*bp[0]
+ adc \$0,%rdx
+ mov $N[1],-32(%rsp,$j,8) # tp[j-1]
+ mov %rdx,$N[0]
+ cmp $num,$j
+ jb .L1st4x
+
+ mulq $m0 # ap[j]*bp[0]
+ add %rax,$A[0]
+ mov -16($np,$j,8),%rax
+ adc \$0,%rdx
+ mov %rdx,$A[1]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[0]
+ mov -8($ap,$j,8),%rax
+ adc \$0,%rdx
+ add $A[0],$N[0] # np[j]*m1+ap[j]*bp[0]
+ adc \$0,%rdx
+ mov $N[0],-24(%rsp,$j,8) # tp[j-1]
+ mov %rdx,$N[1]
+
+ mulq $m0 # ap[j]*bp[0]
+ add %rax,$A[1]
+ mov -8($np,$j,8),%rax
+ adc \$0,%rdx
+ mov %rdx,$A[0]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[1]
+ mov ($ap),%rax # ap[0]
+ adc \$0,%rdx
+ add $A[1],$N[1] # np[j]*m1+ap[j]*bp[0]
+ adc \$0,%rdx
+ mov $N[1],-16(%rsp,$j,8) # tp[j-1]
+ mov %rdx,$N[0]
+
+ xor $N[1],$N[1]
+ add $A[0],$N[0]
+ adc \$0,$N[1]
+ mov $N[0],-8(%rsp,$j,8)
+ mov $N[1],(%rsp,$j,8) # store upmost overflow bit
+
+ lea 1($i),$i # i++
+.align 4
+.Louter4x:
+ mov ($bp,$i,8),$m0 # m0=bp[i]
+ xor $j,$j # j=0
+ mov (%rsp),$A[0]
+ mov $n0,$m1
+ mulq $m0 # ap[0]*bp[i]
+ add %rax,$A[0] # ap[0]*bp[i]+tp[0]
+ mov ($np),%rax
+ adc \$0,%rdx
+
+ imulq $A[0],$m1 # tp[0]*n0
+ mov %rdx,$A[1]
+
+ mulq $m1 # np[0]*m1
+ add %rax,$A[0] # "$N[0]", discarded
+ mov 8($ap),%rax
+ adc \$0,%rdx
+ mov %rdx,$N[1]
+
+ mulq $m0 # ap[j]*bp[i]
+ add %rax,$A[1]
+ mov 8($np),%rax
+ adc \$0,%rdx
+ add 8(%rsp),$A[1] # +tp[1]
+ adc \$0,%rdx
+ mov %rdx,$A[0]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[1]
+ mov 16($ap),%rax
+ adc \$0,%rdx
+ add $A[1],$N[1] # np[j]*m1+ap[j]*bp[i]+tp[j]
+ lea 4($j),$j # j+=2
+ adc \$0,%rdx
+ mov $N[1],(%rsp) # tp[j-1]
+ mov %rdx,$N[0]
+ jmp .Linner4x
+.align 16
+.Linner4x:
+ mulq $m0 # ap[j]*bp[i]
+ add %rax,$A[0]
+ mov -16($np,$j,8),%rax
+ adc \$0,%rdx
+ add -16(%rsp,$j,8),$A[0] # ap[j]*bp[i]+tp[j]
+ adc \$0,%rdx
+ mov %rdx,$A[1]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[0]
+ mov -8($ap,$j,8),%rax
+ adc \$0,%rdx
+ add $A[0],$N[0]
+ adc \$0,%rdx
+ mov $N[0],-24(%rsp,$j,8) # tp[j-1]
+ mov %rdx,$N[1]
+
+ mulq $m0 # ap[j]*bp[i]
+ add %rax,$A[1]
+ mov -8($np,$j,8),%rax
+ adc \$0,%rdx
+ add -8(%rsp,$j,8),$A[1]
+ adc \$0,%rdx
+ mov %rdx,$A[0]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[1]
+ mov ($ap,$j,8),%rax
+ adc \$0,%rdx
+ add $A[1],$N[1]
+ adc \$0,%rdx
+ mov $N[1],-16(%rsp,$j,8) # tp[j-1]
+ mov %rdx,$N[0]
+
+ mulq $m0 # ap[j]*bp[i]
+ add %rax,$A[0]
+ mov ($np,$j,8),%rax
+ adc \$0,%rdx
+ add (%rsp,$j,8),$A[0] # ap[j]*bp[i]+tp[j]
+ adc \$0,%rdx
+ mov %rdx,$A[1]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[0]
+ mov 8($ap,$j,8),%rax
+ adc \$0,%rdx
+ add $A[0],$N[0]
+ adc \$0,%rdx
+ mov $N[0],-8(%rsp,$j,8) # tp[j-1]
+ mov %rdx,$N[1]
+
+ mulq $m0 # ap[j]*bp[i]
+ add %rax,$A[1]
+ mov 8($np,$j,8),%rax
+ adc \$0,%rdx
+ add 8(%rsp,$j,8),$A[1]
+ adc \$0,%rdx
+ lea 4($j),$j # j++
+ mov %rdx,$A[0]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[1]
+ mov -16($ap,$j,8),%rax
+ adc \$0,%rdx
+ add $A[1],$N[1]
+ adc \$0,%rdx
+ mov $N[1],-32(%rsp,$j,8) # tp[j-1]
+ mov %rdx,$N[0]
+ cmp $num,$j
+ jb .Linner4x
+
+ mulq $m0 # ap[j]*bp[i]
+ add %rax,$A[0]
+ mov -16($np,$j,8),%rax
+ adc \$0,%rdx
+ add -16(%rsp,$j,8),$A[0] # ap[j]*bp[i]+tp[j]
+ adc \$0,%rdx
+ mov %rdx,$A[1]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[0]
+ mov -8($ap,$j,8),%rax
+ adc \$0,%rdx
+ add $A[0],$N[0]
+ adc \$0,%rdx
+ mov $N[0],-24(%rsp,$j,8) # tp[j-1]
+ mov %rdx,$N[1]
+
+ mulq $m0 # ap[j]*bp[i]
+ add %rax,$A[1]
+ mov -8($np,$j,8),%rax
+ adc \$0,%rdx
+ add -8(%rsp,$j,8),$A[1]
+ adc \$0,%rdx
+ lea 1($i),$i # i++
+ mov %rdx,$A[0]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[1]
+ mov ($ap),%rax # ap[0]
+ adc \$0,%rdx
+ add $A[1],$N[1]
+ adc \$0,%rdx
+ mov $N[1],-16(%rsp,$j,8) # tp[j-1]
+ mov %rdx,$N[0]
+
+ xor $N[1],$N[1]
+ add $A[0],$N[0]
+ adc \$0,$N[1]
+ add (%rsp,$num,8),$N[0] # pull upmost overflow bit
+ adc \$0,$N[1]
+ mov $N[0],-8(%rsp,$j,8)
+ mov $N[1],(%rsp,$j,8) # store upmost overflow bit
+
+ cmp $num,$i
+ jb .Louter4x
+___
+{
+my @ri=("%rax","%rdx",$m0,$m1);
+$code.=<<___;
+ mov 16(%rsp,$num,8),$rp # restore $rp
+ lea -4($num),$j
+ mov 0(%rsp),@ri[0] # tp[0]
+ mov 8(%rsp),@ri[1] # tp[1]
+ shr \$2,$j # j=num/4-1
+ lea (%rsp),$ap # borrow ap for tp
+ xor $i,$i # i=0 and clear CF!
+
+ sub 0($np),@ri[0]
+ mov 16($ap),@ri[2] # tp[2]
+ mov 24($ap),@ri[3] # tp[3]
+ sbb 8($np),@ri[1]
+
+.Lsub4x:
+ mov @ri[0],0($rp,$i,8) # rp[i]=tp[i]-np[i]
+ mov @ri[1],8($rp,$i,8) # rp[i]=tp[i]-np[i]
+ sbb 16($np,$i,8),@ri[2]
+ mov 32($ap,$i,8),@ri[0] # tp[i+1]
+ mov 40($ap,$i,8),@ri[1]
+ sbb 24($np,$i,8),@ri[3]
+ mov @ri[2],16($rp,$i,8) # rp[i]=tp[i]-np[i]
+ mov @ri[3],24($rp,$i,8) # rp[i]=tp[i]-np[i]
+ sbb 32($np,$i,8),@ri[0]
+ mov 48($ap,$i,8),@ri[2]
+ mov 56($ap,$i,8),@ri[3]
+ sbb 40($np,$i,8),@ri[1]
+ lea 4($i),$i # i++
+ dec $j # doesn't affect CF!
+ jnz .Lsub4x
+
+ mov @ri[0],0($rp,$i,8) # rp[i]=tp[i]-np[i]
+ mov 32($ap,$i,8),@ri[0] # load overflow bit
+ sbb 16($np,$i,8),@ri[2]
+ mov @ri[1],8($rp,$i,8) # rp[i]=tp[i]-np[i]
+ sbb 24($np,$i,8),@ri[3]
+ mov @ri[2],16($rp,$i,8) # rp[i]=tp[i]-np[i]
+
+ sbb \$0,@ri[0] # handle upmost overflow bit
+ mov @ri[3],24($rp,$i,8) # rp[i]=tp[i]-np[i]
+ pxor %xmm0,%xmm0
+ movq @ri[0],%xmm4
+ pcmpeqd %xmm5,%xmm5
+ pshufd \$0,%xmm4,%xmm4
+ mov $num,$j
+ pxor %xmm4,%xmm5
+ shr \$2,$j # j=num/4
+ xor %eax,%eax # i=0
+
+ jmp .Lcopy4x
+.align 16
+.Lcopy4x: # conditional copy
+ movdqa (%rsp,%rax),%xmm1
+ movdqu ($rp,%rax),%xmm2
+ pand %xmm4,%xmm1
+ pand %xmm5,%xmm2
+ movdqa 16(%rsp,%rax),%xmm3
+ movdqa %xmm0,(%rsp,%rax)
+ por %xmm2,%xmm1
+ movdqu 16($rp,%rax),%xmm2
+ movdqu %xmm1,($rp,%rax)
+ pand %xmm4,%xmm3
+ pand %xmm5,%xmm2
+ movdqa %xmm0,16(%rsp,%rax)
+ por %xmm2,%xmm3
+ movdqu %xmm3,16($rp,%rax)
+ lea 32(%rax),%rax
+ dec $j
+ jnz .Lcopy4x
+___
+}
+$code.=<<___;
+ mov 8(%rsp,$num,8),%rsi # restore %rsp
+.cfi_def_cfa %rsi, 8
+ mov \$1,%rax
+ mov -48(%rsi),%r15
+.cfi_restore %r15
+ mov -40(%rsi),%r14
+.cfi_restore %r14
+ mov -32(%rsi),%r13
+.cfi_restore %r13
+ mov -24(%rsi),%r12
+.cfi_restore %r12
+ mov -16(%rsi),%rbp
+.cfi_restore %rbp
+ mov -8(%rsi),%rbx
+.cfi_restore %rbx
+ lea (%rsi),%rsp
+.cfi_def_cfa_register %rsp
+.Lmul4x_epilogue:
+ ret
+.cfi_endproc
+.size bn_mul4x_mont,.-bn_mul4x_mont
+___
+}}}
+ {{{
+######################################################################
+# void bn_sqr8x_mont(
+my $rptr="%rdi"; # const BN_ULONG *rptr,
+my $aptr="%rsi"; # const BN_ULONG *aptr,
+my $bptr="%rdx"; # not used
+my $nptr="%rcx"; # const BN_ULONG *nptr,
+my $n0 ="%r8"; # const BN_ULONG *n0);
+my $num ="%r9"; # int num, has to be divisible by 8
+
+my ($i,$j,$tptr)=("%rbp","%rcx",$rptr);
+my @A0=("%r10","%r11");
+my @A1=("%r12","%r13");
+my ($a0,$a1,$ai)=("%r14","%r15","%rbx");
+
+$code.=<<___ if ($addx);
+.extern bn_sqrx8x_internal # see x86_64-mont5 module
+___
+$code.=<<___;
+.extern bn_sqr8x_internal # see x86_64-mont5 module
+
+.type bn_sqr8x_mont,\@function,6
+.align 32
+bn_sqr8x_mont:
+.cfi_startproc
+ mov %rsp,%rax
+.cfi_def_cfa_register %rax
+.Lsqr8x_enter:
+ push %rbx
+.cfi_push %rbx
+ push %rbp
+.cfi_push %rbp
+ push %r12
+.cfi_push %r12
+ push %r13
+.cfi_push %r13
+ push %r14
+.cfi_push %r14
+ push %r15
+.cfi_push %r15
+.Lsqr8x_prologue:
+
+ mov ${num}d,%r10d
+ shl \$3,${num}d # convert $num to bytes
+ shl \$3+2,%r10 # 4*$num
+ neg $num
+
+ ##############################################################
+ # ensure that stack frame doesn't alias with $aptr modulo
+ # 4096. this is done to allow memory disambiguation logic
+ # do its job.
+ #
+ lea -64(%rsp,$num,2),%r11
+ mov %rsp,%rbp
+ mov ($n0),$n0 # *n0
+ sub $aptr,%r11
+ and \$4095,%r11
+ cmp %r11,%r10
+ jb .Lsqr8x_sp_alt
+ sub %r11,%rbp # align with $aptr
+ lea -64(%rbp,$num,2),%rbp # future alloca(frame+2*$num)
+ jmp .Lsqr8x_sp_done
+
+.align 32
+.Lsqr8x_sp_alt:
+ lea 4096-64(,$num,2),%r10 # 4096-frame-2*$num
+ lea -64(%rbp,$num,2),%rbp # future alloca(frame+2*$num)
+ sub %r10,%r11
+ mov \$0,%r10
+ cmovc %r10,%r11
+ sub %r11,%rbp
+.Lsqr8x_sp_done:
+ and \$-64,%rbp
+ mov %rsp,%r11
+ sub %rbp,%r11
+ and \$-4096,%r11
+ lea (%rbp,%r11),%rsp
+ mov (%rsp),%r10
+ cmp %rbp,%rsp
+ ja .Lsqr8x_page_walk
+ jmp .Lsqr8x_page_walk_done
+
+.align 16
+.Lsqr8x_page_walk:
+ lea -4096(%rsp),%rsp
+ mov (%rsp),%r10
+ cmp %rbp,%rsp
+ ja .Lsqr8x_page_walk
+.Lsqr8x_page_walk_done:
+
+ mov $num,%r10
+ neg $num
+
+ mov $n0, 32(%rsp)
+ mov %rax, 40(%rsp) # save original %rsp
+.cfi_cfa_expression %rsp+40,deref,+8
+.Lsqr8x_body:
+
+ movq $nptr, %xmm2 # save pointer to modulus
+ pxor %xmm0,%xmm0
+ movq $rptr,%xmm1 # save $rptr
+ movq %r10, %xmm3 # -$num
+___
+$code.=<<___ if ($addx);
+ mov OPENSSL_ia32cap_P+8(%rip),%eax
+ and \$0x80100,%eax
+ cmp \$0x80100,%eax
+ jne .Lsqr8x_nox
+
+ call bn_sqrx8x_internal # see x86_64-mont5 module
+ # %rax top-most carry
+ # %rbp nptr
+ # %rcx -8*num
+ # %r8 end of tp[2*num]
+ lea (%r8,%rcx),%rbx
+ mov %rcx,$num
+ mov %rcx,%rdx
+ movq %xmm1,$rptr
+ sar \$3+2,%rcx # %cf=0
+ jmp .Lsqr8x_sub
+
+.align 32
+.Lsqr8x_nox:
+___
+$code.=<<___;
+ call bn_sqr8x_internal # see x86_64-mont5 module
+ # %rax top-most carry
+ # %rbp nptr
+ # %r8 -8*num
+ # %rdi end of tp[2*num]
+ lea (%rdi,$num),%rbx
+ mov $num,%rcx
+ mov $num,%rdx
+ movq %xmm1,$rptr
+ sar \$3+2,%rcx # %cf=0
+ jmp .Lsqr8x_sub
+
+.align 32
+.Lsqr8x_sub:
+ mov 8*0(%rbx),%r12
+ mov 8*1(%rbx),%r13
+ mov 8*2(%rbx),%r14
+ mov 8*3(%rbx),%r15
+ lea 8*4(%rbx),%rbx
+ sbb 8*0(%rbp),%r12
+ sbb 8*1(%rbp),%r13
+ sbb 8*2(%rbp),%r14
+ sbb 8*3(%rbp),%r15
+ lea 8*4(%rbp),%rbp
+ mov %r12,8*0($rptr)
+ mov %r13,8*1($rptr)
+ mov %r14,8*2($rptr)
+ mov %r15,8*3($rptr)
+ lea 8*4($rptr),$rptr
+ inc %rcx # preserves %cf
+ jnz .Lsqr8x_sub
+
+ sbb \$0,%rax # top-most carry
+ lea (%rbx,$num),%rbx # rewind
+ lea ($rptr,$num),$rptr # rewind
+
+ movq %rax,%xmm1
+ pxor %xmm0,%xmm0
+ pshufd \$0,%xmm1,%xmm1
+ mov 40(%rsp),%rsi # restore %rsp
+.cfi_def_cfa %rsi,8
+ jmp .Lsqr8x_cond_copy
+
+.align 32
+.Lsqr8x_cond_copy:
+ movdqa 16*0(%rbx),%xmm2
+ movdqa 16*1(%rbx),%xmm3
+ lea 16*2(%rbx),%rbx
+ movdqu 16*0($rptr),%xmm4
+ movdqu 16*1($rptr),%xmm5
+ lea 16*2($rptr),$rptr
+ movdqa %xmm0,-16*2(%rbx) # zero tp
+ movdqa %xmm0,-16*1(%rbx)
+ movdqa %xmm0,-16*2(%rbx,%rdx)
+ movdqa %xmm0,-16*1(%rbx,%rdx)
+ pcmpeqd %xmm1,%xmm0
+ pand %xmm1,%xmm2
+ pand %xmm1,%xmm3
+ pand %xmm0,%xmm4
+ pand %xmm0,%xmm5
+ pxor %xmm0,%xmm0
+ por %xmm2,%xmm4
+ por %xmm3,%xmm5
+ movdqu %xmm4,-16*2($rptr)
+ movdqu %xmm5,-16*1($rptr)
+ add \$32,$num
+ jnz .Lsqr8x_cond_copy
+
+ mov \$1,%rax
+ mov -48(%rsi),%r15
+.cfi_restore %r15
+ mov -40(%rsi),%r14
+.cfi_restore %r14
+ mov -32(%rsi),%r13
+.cfi_restore %r13
+ mov -24(%rsi),%r12
+.cfi_restore %r12
+ mov -16(%rsi),%rbp
+.cfi_restore %rbp
+ mov -8(%rsi),%rbx
+.cfi_restore %rbx
+ lea (%rsi),%rsp
+.cfi_def_cfa_register %rsp
+.Lsqr8x_epilogue:
+ ret
+.cfi_endproc
+.size bn_sqr8x_mont,.-bn_sqr8x_mont
+___
+}}}
+
+if ($addx) {{{
+my $bp="%rdx"; # original value
+
+$code.=<<___;
+.type bn_mulx4x_mont,\@function,6
+.align 32
+bn_mulx4x_mont:
+.cfi_startproc
+ mov %rsp,%rax
+.cfi_def_cfa_register %rax
+.Lmulx4x_enter:
+ push %rbx
+.cfi_push %rbx
+ push %rbp
+.cfi_push %rbp
+ push %r12
+.cfi_push %r12
+ push %r13
+.cfi_push %r13
+ push %r14
+.cfi_push %r14
+ push %r15
+.cfi_push %r15
+.Lmulx4x_prologue:
+
+ shl \$3,${num}d # convert $num to bytes
+ xor %r10,%r10
+ sub $num,%r10 # -$num
+ mov ($n0),$n0 # *n0
+ lea -72(%rsp,%r10),%rbp # future alloca(frame+$num+8)
+ and \$-128,%rbp
+ mov %rsp,%r11
+ sub %rbp,%r11
+ and \$-4096,%r11
+ lea (%rbp,%r11),%rsp
+ mov (%rsp),%r10
+ cmp %rbp,%rsp
+ ja .Lmulx4x_page_walk
+ jmp .Lmulx4x_page_walk_done
+
+.align 16
+.Lmulx4x_page_walk:
+ lea -4096(%rsp),%rsp
+ mov (%rsp),%r10
+ cmp %rbp,%rsp
+ ja .Lmulx4x_page_walk
+.Lmulx4x_page_walk_done:
+
+ lea ($bp,$num),%r10
+ ##############################################################
+ # Stack layout
+ # +0 num
+ # +8 off-loaded &b[i]
+ # +16 end of b[num]
+ # +24 saved n0
+ # +32 saved rp
+ # +40 saved %rsp
+ # +48 inner counter
+ # +56
+ # +64 tmp[num+1]
+ #
+ mov $num,0(%rsp) # save $num
+ shr \$5,$num
+ mov %r10,16(%rsp) # end of b[num]
+ sub \$1,$num
+ mov $n0, 24(%rsp) # save *n0
+ mov $rp, 32(%rsp) # save $rp
+ mov %rax,40(%rsp) # save original %rsp
+.cfi_cfa_expression %rsp+40,deref,+8
+ mov $num,48(%rsp) # inner counter
+ jmp .Lmulx4x_body
+
+.align 32
+.Lmulx4x_body:
+___
+my ($aptr, $bptr, $nptr, $tptr, $mi, $bi, $zero, $num)=
+ ("%rsi","%rdi","%rcx","%rbx","%r8","%r9","%rbp","%rax");
+my $rptr=$bptr;
+$code.=<<___;
+ lea 8($bp),$bptr
+ mov ($bp),%rdx # b[0], $bp==%rdx actually
+ lea 64+32(%rsp),$tptr
+ mov %rdx,$bi
+
+ mulx 0*8($aptr),$mi,%rax # a[0]*b[0]
+ mulx 1*8($aptr),%r11,%r14 # a[1]*b[0]
+ add %rax,%r11
+ mov $bptr,8(%rsp) # off-load &b[i]
+ mulx 2*8($aptr),%r12,%r13 # ...
+ adc %r14,%r12
+ adc \$0,%r13
+
+ mov $mi,$bptr # borrow $bptr
+ imulq 24(%rsp),$mi # "t[0]"*n0
+ xor $zero,$zero # cf=0, of=0
+
+ mulx 3*8($aptr),%rax,%r14
+ mov $mi,%rdx
+ lea 4*8($aptr),$aptr
+ adcx %rax,%r13
+ adcx $zero,%r14 # cf=0
+
+ mulx 0*8($nptr),%rax,%r10
+ adcx %rax,$bptr # discarded
+ adox %r11,%r10
+ mulx 1*8($nptr),%rax,%r11
+ adcx %rax,%r10
+ adox %r12,%r11
+ .byte 0xc4,0x62,0xfb,0xf6,0xa1,0x10,0x00,0x00,0x00 # mulx 2*8($nptr),%rax,%r12
+ mov 48(%rsp),$bptr # counter value
+ mov %r10,-4*8($tptr)
+ adcx %rax,%r11
+ adox %r13,%r12
+ mulx 3*8($nptr),%rax,%r15
+ mov $bi,%rdx
+ mov %r11,-3*8($tptr)
+ adcx %rax,%r12
+ adox $zero,%r15 # of=0
+ lea 4*8($nptr),$nptr
+ mov %r12,-2*8($tptr)
+
+ jmp .Lmulx4x_1st
+
+.align 32
+.Lmulx4x_1st:
+ adcx $zero,%r15 # cf=0, modulo-scheduled
+ mulx 0*8($aptr),%r10,%rax # a[4]*b[0]
+ adcx %r14,%r10
+ mulx 1*8($aptr),%r11,%r14 # a[5]*b[0]
+ adcx %rax,%r11
+ mulx 2*8($aptr),%r12,%rax # ...
+ adcx %r14,%r12
+ mulx 3*8($aptr),%r13,%r14
+ .byte 0x67,0x67
+ mov $mi,%rdx
+ adcx %rax,%r13
+ adcx $zero,%r14 # cf=0
+ lea 4*8($aptr),$aptr
+ lea 4*8($tptr),$tptr
+
+ adox %r15,%r10
+ mulx 0*8($nptr),%rax,%r15
+ adcx %rax,%r10
+ adox %r15,%r11
+ mulx 1*8($nptr),%rax,%r15
+ adcx %rax,%r11
+ adox %r15,%r12
+ mulx 2*8($nptr),%rax,%r15
+ mov %r10,-5*8($tptr)
+ adcx %rax,%r12
+ mov %r11,-4*8($tptr)
+ adox %r15,%r13
+ mulx 3*8($nptr),%rax,%r15
+ mov $bi,%rdx
+ mov %r12,-3*8($tptr)
+ adcx %rax,%r13
+ adox $zero,%r15
+ lea 4*8($nptr),$nptr
+ mov %r13,-2*8($tptr)
+
+ dec $bptr # of=0, pass cf
+ jnz .Lmulx4x_1st
+
+ mov 0(%rsp),$num # load num
+ mov 8(%rsp),$bptr # re-load &b[i]
+ adc $zero,%r15 # modulo-scheduled
+ add %r15,%r14
+ sbb %r15,%r15 # top-most carry
+ mov %r14,-1*8($tptr)
+ jmp .Lmulx4x_outer
+
+.align 32
+.Lmulx4x_outer:
+ mov ($bptr),%rdx # b[i]
+ lea 8($bptr),$bptr # b++
+ sub $num,$aptr # rewind $aptr
+ mov %r15,($tptr) # save top-most carry
+ lea 64+4*8(%rsp),$tptr
+ sub $num,$nptr # rewind $nptr
+
+ mulx 0*8($aptr),$mi,%r11 # a[0]*b[i]
+ xor %ebp,%ebp # xor $zero,$zero # cf=0, of=0
+ mov %rdx,$bi
+ mulx 1*8($aptr),%r14,%r12 # a[1]*b[i]
+ adox -4*8($tptr),$mi
+ adcx %r14,%r11
+ mulx 2*8($aptr),%r15,%r13 # ...
+ adox -3*8($tptr),%r11
+ adcx %r15,%r12
+ adox -2*8($tptr),%r12
+ adcx $zero,%r13
+ adox $zero,%r13
+
+ mov $bptr,8(%rsp) # off-load &b[i]
+ mov $mi,%r15
+ imulq 24(%rsp),$mi # "t[0]"*n0
+ xor %ebp,%ebp # xor $zero,$zero # cf=0, of=0
+
+ mulx 3*8($aptr),%rax,%r14
+ mov $mi,%rdx
+ adcx %rax,%r13
+ adox -1*8($tptr),%r13
+ adcx $zero,%r14
+ lea 4*8($aptr),$aptr
+ adox $zero,%r14
+
+ mulx 0*8($nptr),%rax,%r10
+ adcx %rax,%r15 # discarded
+ adox %r11,%r10
+ mulx 1*8($nptr),%rax,%r11
+ adcx %rax,%r10
+ adox %r12,%r11
+ mulx 2*8($nptr),%rax,%r12
+ mov %r10,-4*8($tptr)
+ adcx %rax,%r11
+ adox %r13,%r12
+ mulx 3*8($nptr),%rax,%r15
+ mov $bi,%rdx
+ mov %r11,-3*8($tptr)
+ lea 4*8($nptr),$nptr
+ adcx %rax,%r12
+ adox $zero,%r15 # of=0
+ mov 48(%rsp),$bptr # counter value
+ mov %r12,-2*8($tptr)
+
+ jmp .Lmulx4x_inner
+
+.align 32
+.Lmulx4x_inner:
+ mulx 0*8($aptr),%r10,%rax # a[4]*b[i]
+ adcx $zero,%r15 # cf=0, modulo-scheduled
+ adox %r14,%r10
+ mulx 1*8($aptr),%r11,%r14 # a[5]*b[i]
+ adcx 0*8($tptr),%r10
+ adox %rax,%r11
+ mulx 2*8($aptr),%r12,%rax # ...
+ adcx 1*8($tptr),%r11
+ adox %r14,%r12
+ mulx 3*8($aptr),%r13,%r14
+ mov $mi,%rdx
+ adcx 2*8($tptr),%r12
+ adox %rax,%r13
+ adcx 3*8($tptr),%r13
+ adox $zero,%r14 # of=0
+ lea 4*8($aptr),$aptr
+ lea 4*8($tptr),$tptr
+ adcx $zero,%r14 # cf=0
+
+ adox %r15,%r10
+ mulx 0*8($nptr),%rax,%r15
+ adcx %rax,%r10
+ adox %r15,%r11
+ mulx 1*8($nptr),%rax,%r15
+ adcx %rax,%r11
+ adox %r15,%r12
+ mulx 2*8($nptr),%rax,%r15
+ mov %r10,-5*8($tptr)
+ adcx %rax,%r12
+ adox %r15,%r13
+ mulx 3*8($nptr),%rax,%r15
+ mov $bi,%rdx
+ mov %r11,-4*8($tptr)
+ mov %r12,-3*8($tptr)
+ adcx %rax,%r13
+ adox $zero,%r15
+ lea 4*8($nptr),$nptr
+ mov %r13,-2*8($tptr)
+
+ dec $bptr # of=0, pass cf
+ jnz .Lmulx4x_inner
+
+ mov 0(%rsp),$num # load num
+ mov 8(%rsp),$bptr # re-load &b[i]
+ adc $zero,%r15 # modulo-scheduled
+ sub 0*8($tptr),$zero # pull top-most carry
+ adc %r15,%r14
+ sbb %r15,%r15 # top-most carry
+ mov %r14,-1*8($tptr)
+
+ cmp 16(%rsp),$bptr
+ jne .Lmulx4x_outer
+
+ lea 64(%rsp),$tptr
+ sub $num,$nptr # rewind $nptr
+ neg %r15
+ mov $num,%rdx
+ shr \$3+2,$num # %cf=0
+ mov 32(%rsp),$rptr # restore rp
+ jmp .Lmulx4x_sub
+
+.align 32
+.Lmulx4x_sub:
+ mov 8*0($tptr),%r11
+ mov 8*1($tptr),%r12
+ mov 8*2($tptr),%r13
+ mov 8*3($tptr),%r14
+ lea 8*4($tptr),$tptr
+ sbb 8*0($nptr),%r11
+ sbb 8*1($nptr),%r12
+ sbb 8*2($nptr),%r13
+ sbb 8*3($nptr),%r14
+ lea 8*4($nptr),$nptr
+ mov %r11,8*0($rptr)
+ mov %r12,8*1($rptr)
+ mov %r13,8*2($rptr)
+ mov %r14,8*3($rptr)
+ lea 8*4($rptr),$rptr
+ dec $num # preserves %cf
+ jnz .Lmulx4x_sub
+
+ sbb \$0,%r15 # top-most carry
+ lea 64(%rsp),$tptr
+ sub %rdx,$rptr # rewind
+
+ movq %r15,%xmm1
+ pxor %xmm0,%xmm0
+ pshufd \$0,%xmm1,%xmm1
+ mov 40(%rsp),%rsi # restore %rsp
+.cfi_def_cfa %rsi,8
+ jmp .Lmulx4x_cond_copy
+
+.align 32
+.Lmulx4x_cond_copy:
+ movdqa 16*0($tptr),%xmm2
+ movdqa 16*1($tptr),%xmm3
+ lea 16*2($tptr),$tptr
+ movdqu 16*0($rptr),%xmm4
+ movdqu 16*1($rptr),%xmm5
+ lea 16*2($rptr),$rptr
+ movdqa %xmm0,-16*2($tptr) # zero tp
+ movdqa %xmm0,-16*1($tptr)
+ pcmpeqd %xmm1,%xmm0
+ pand %xmm1,%xmm2
+ pand %xmm1,%xmm3
+ pand %xmm0,%xmm4
+ pand %xmm0,%xmm5
+ pxor %xmm0,%xmm0
+ por %xmm2,%xmm4
+ por %xmm3,%xmm5
+ movdqu %xmm4,-16*2($rptr)
+ movdqu %xmm5,-16*1($rptr)
+ sub \$32,%rdx
+ jnz .Lmulx4x_cond_copy
+
+ mov %rdx,($tptr)
+
+ mov \$1,%rax
+ mov -48(%rsi),%r15
+.cfi_restore %r15
+ mov -40(%rsi),%r14
+.cfi_restore %r14
+ mov -32(%rsi),%r13
+.cfi_restore %r13
+ mov -24(%rsi),%r12
+.cfi_restore %r12
+ mov -16(%rsi),%rbp
+.cfi_restore %rbp
+ mov -8(%rsi),%rbx
+.cfi_restore %rbx
+ lea (%rsi),%rsp
+.cfi_def_cfa_register %rsp
+.Lmulx4x_epilogue:
+ ret
+.cfi_endproc
+.size bn_mulx4x_mont,.-bn_mulx4x_mont
+___
+}}}
+$code.=<<___;
+.asciz "Montgomery Multiplication for x86_64, CRYPTOGAMS by <appro\@openssl.org>"
+.align 16
+___
+
+# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame,
+# CONTEXT *context,DISPATCHER_CONTEXT *disp)
+if ($win64) {
+$rec="%rcx";
+$frame="%rdx";
+$context="%r8";
+$disp="%r9";
+
+$code.=<<___;
+.extern __imp_RtlVirtualUnwind
+.type mul_handler,\@abi-omnipotent
+.align 16
+mul_handler:
+ push %rsi
+ push %rdi
+ push %rbx
+ push %rbp
+ push %r12
+ push %r13
+ push %r14
+ push %r15
+ pushfq
+ sub \$64,%rsp
+
+ mov 120($context),%rax # pull context->Rax
+ mov 248($context),%rbx # pull context->Rip
+
+ mov 8($disp),%rsi # disp->ImageBase
+ mov 56($disp),%r11 # disp->HandlerData
+
+ mov 0(%r11),%r10d # HandlerData[0]
+ lea (%rsi,%r10),%r10 # end of prologue label
+ cmp %r10,%rbx # context->Rip<end of prologue label
+ jb .Lcommon_seh_tail
+
+ mov 152($context),%rax # pull context->Rsp
+
+ mov 4(%r11),%r10d # HandlerData[1]
+ lea (%rsi,%r10),%r10 # epilogue label
+ cmp %r10,%rbx # context->Rip>=epilogue label
+ jae .Lcommon_seh_tail
+
+ mov 192($context),%r10 # pull $num
+ mov 8(%rax,%r10,8),%rax # pull saved stack pointer
+
+ jmp .Lcommon_pop_regs
+.size mul_handler,.-mul_handler
+
+.type sqr_handler,\@abi-omnipotent
+.align 16
+sqr_handler:
+ push %rsi
+ push %rdi
+ push %rbx
+ push %rbp
+ push %r12
+ push %r13
+ push %r14
+ push %r15
+ pushfq
+ sub \$64,%rsp
+
+ mov 120($context),%rax # pull context->Rax
+ mov 248($context),%rbx # pull context->Rip
+
+ mov 8($disp),%rsi # disp->ImageBase
+ mov 56($disp),%r11 # disp->HandlerData
+
+ mov 0(%r11),%r10d # HandlerData[0]
+ lea (%rsi,%r10),%r10 # end of prologue label
+ cmp %r10,%rbx # context->Rip<.Lsqr_prologue
+ jb .Lcommon_seh_tail
+
+ mov 4(%r11),%r10d # HandlerData[1]
+ lea (%rsi,%r10),%r10 # body label
+ cmp %r10,%rbx # context->Rip<.Lsqr_body
+ jb .Lcommon_pop_regs
+
+ mov 152($context),%rax # pull context->Rsp
+
+ mov 8(%r11),%r10d # HandlerData[2]
+ lea (%rsi,%r10),%r10 # epilogue label
+ cmp %r10,%rbx # context->Rip>=.Lsqr_epilogue
+ jae .Lcommon_seh_tail
+
+ mov 40(%rax),%rax # pull saved stack pointer
+
+.Lcommon_pop_regs:
+ mov -8(%rax),%rbx
+ mov -16(%rax),%rbp
+ mov -24(%rax),%r12
+ mov -32(%rax),%r13
+ mov -40(%rax),%r14
+ mov -48(%rax),%r15
+ mov %rbx,144($context) # restore context->Rbx
+ mov %rbp,160($context) # restore context->Rbp
+ mov %r12,216($context) # restore context->R12
+ mov %r13,224($context) # restore context->R13
+ mov %r14,232($context) # restore context->R14
+ mov %r15,240($context) # restore context->R15
+
+.Lcommon_seh_tail:
+ mov 8(%rax),%rdi
+ mov 16(%rax),%rsi
+ mov %rax,152($context) # restore context->Rsp
+ mov %rsi,168($context) # restore context->Rsi
+ mov %rdi,176($context) # restore context->Rdi
+
+ mov 40($disp),%rdi # disp->ContextRecord
+ mov $context,%rsi # context
+ mov \$154,%ecx # sizeof(CONTEXT)
+ .long 0xa548f3fc # cld; rep movsq
+
+ mov $disp,%rsi
+ xor %rcx,%rcx # arg1, UNW_FLAG_NHANDLER
+ mov 8(%rsi),%rdx # arg2, disp->ImageBase
+ mov 0(%rsi),%r8 # arg3, disp->ControlPc
+ mov 16(%rsi),%r9 # arg4, disp->FunctionEntry
+ mov 40(%rsi),%r10 # disp->ContextRecord
+ lea 56(%rsi),%r11 # &disp->HandlerData
+ lea 24(%rsi),%r12 # &disp->EstablisherFrame
+ mov %r10,32(%rsp) # arg5
+ mov %r11,40(%rsp) # arg6
+ mov %r12,48(%rsp) # arg7
+ mov %rcx,56(%rsp) # arg8, (NULL)
+ call *__imp_RtlVirtualUnwind(%rip)
+
+ mov \$1,%eax # ExceptionContinueSearch
+ add \$64,%rsp
+ popfq
+ pop %r15
+ pop %r14
+ pop %r13
+ pop %r12
+ pop %rbp
+ pop %rbx
+ pop %rdi
+ pop %rsi
+ ret
+.size sqr_handler,.-sqr_handler
+
+.section .pdata
+.align 4
+ .rva .LSEH_begin_bn_mul_mont
+ .rva .LSEH_end_bn_mul_mont
+ .rva .LSEH_info_bn_mul_mont
+
+ .rva .LSEH_begin_bn_mul4x_mont
+ .rva .LSEH_end_bn_mul4x_mont
+ .rva .LSEH_info_bn_mul4x_mont
+
+ .rva .LSEH_begin_bn_sqr8x_mont
+ .rva .LSEH_end_bn_sqr8x_mont
+ .rva .LSEH_info_bn_sqr8x_mont
+___
+$code.=<<___ if ($addx);
+ .rva .LSEH_begin_bn_mulx4x_mont
+ .rva .LSEH_end_bn_mulx4x_mont
+ .rva .LSEH_info_bn_mulx4x_mont
+___
+$code.=<<___;
+.section .xdata
+.align 8
+.LSEH_info_bn_mul_mont:
+ .byte 9,0,0,0
+ .rva mul_handler
+ .rva .Lmul_body,.Lmul_epilogue # HandlerData[]
+.LSEH_info_bn_mul4x_mont:
+ .byte 9,0,0,0
+ .rva mul_handler
+ .rva .Lmul4x_body,.Lmul4x_epilogue # HandlerData[]
+.LSEH_info_bn_sqr8x_mont:
+ .byte 9,0,0,0
+ .rva sqr_handler
+ .rva .Lsqr8x_prologue,.Lsqr8x_body,.Lsqr8x_epilogue # HandlerData[]
+.align 8
+___
+$code.=<<___ if ($addx);
+.LSEH_info_bn_mulx4x_mont:
+ .byte 9,0,0,0
+ .rva sqr_handler
+ .rva .Lmulx4x_prologue,.Lmulx4x_body,.Lmulx4x_epilogue # HandlerData[]
+.align 8
+___
+}
+
+print $code;
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/x86_64-mont5.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/x86_64-mont5.pl
new file mode 100755
index 0000000..33cb769
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/asm/x86_64-mont5.pl
@@ -0,0 +1,3767 @@
+#! /usr/bin/env perl
+# Copyright 2011-2022 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+
+# August 2011.
+#
+# Companion to x86_64-mont.pl that optimizes cache-timing attack
+# countermeasures. The subroutines are produced by replacing bp[i]
+# references in their x86_64-mont.pl counterparts with cache-neutral
+# references to powers table computed in BN_mod_exp_mont_consttime.
+# In addition subroutine that scatters elements of the powers table
+# is implemented, so that scatter-/gathering can be tuned without
+# bn_exp.c modifications.
+
+# August 2013.
+#
+# Add MULX/AD*X code paths and additional interfaces to optimize for
+# branch prediction unit. For input lengths that are multiples of 8
+# the np argument is not just modulus value, but one interleaved
+# with 0. This is to optimize post-condition...
+
+$flavour = shift;
+$output = shift;
+if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
+
+$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or
+( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or
+die "can't locate x86_64-xlate.pl";
+
+open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\"";
+*STDOUT=*OUT;
+
+if (`$ENV{CC} -Wa,-v -c -o /dev/null -x assembler /dev/null 2>&1`
+ =~ /GNU assembler version ([2-9]\.[0-9]+)/) {
+ $addx = ($1>=2.23);
+}
+
+if (!$addx && $win64 && ($flavour =~ /nasm/ || $ENV{ASM} =~ /nasm/) &&
+ `nasm -v 2>&1` =~ /NASM version ([2-9]\.[0-9]+)/) {
+ $addx = ($1>=2.10);
+}
+
+if (!$addx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
+ `ml64 2>&1` =~ /Version ([0-9]+)\./) {
+ $addx = ($1>=12);
+}
+
+if (!$addx && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|.*based on LLVM) ([0-9]+)\.([0-9]+)/) {
+ my $ver = $2 + $3/100.0; # 3.1->3.01, 3.10->3.10
+ $addx = ($ver>=3.03);
+}
+
+# int bn_mul_mont_gather5(
+$rp="%rdi"; # BN_ULONG *rp,
+$ap="%rsi"; # const BN_ULONG *ap,
+$bp="%rdx"; # const BN_ULONG *bp,
+$np="%rcx"; # const BN_ULONG *np,
+$n0="%r8"; # const BN_ULONG *n0,
+$num="%r9"; # int num,
+ # int idx); # 0 to 2^5-1, "index" in $bp holding
+ # pre-computed powers of a', interlaced
+ # in such manner that b[0] is $bp[idx],
+ # b[1] is [2^5+idx], etc.
+$lo0="%r10";
+$hi0="%r11";
+$hi1="%r13";
+$i="%r14";
+$j="%r15";
+$m0="%rbx";
+$m1="%rbp";
+
+$code=<<___;
+.text
+
+.extern OPENSSL_ia32cap_P
+
+.globl bn_mul_mont_gather5
+.type bn_mul_mont_gather5,\@function,6
+.align 64
+bn_mul_mont_gather5:
+.cfi_startproc
+ mov ${num}d,${num}d
+ mov %rsp,%rax
+.cfi_def_cfa_register %rax
+ test \$7,${num}d
+ jnz .Lmul_enter
+___
+$code.=<<___ if ($addx);
+ mov OPENSSL_ia32cap_P+8(%rip),%r11d
+___
+$code.=<<___;
+ jmp .Lmul4x_enter
+
+.align 16
+.Lmul_enter:
+ movd `($win64?56:8)`(%rsp),%xmm5 # load 7th argument
+ push %rbx
+.cfi_push %rbx
+ push %rbp
+.cfi_push %rbp
+ push %r12
+.cfi_push %r12
+ push %r13
+.cfi_push %r13
+ push %r14
+.cfi_push %r14
+ push %r15
+.cfi_push %r15
+
+ neg $num
+ mov %rsp,%r11
+ lea -280(%rsp,$num,8),%r10 # future alloca(8*(num+2)+256+8)
+ neg $num # restore $num
+ and \$-1024,%r10 # minimize TLB usage
+
+ # An OS-agnostic version of __chkstk.
+ #
+ # Some OSes (Windows) insist on stack being "wired" to
+ # physical memory in strictly sequential manner, i.e. if stack
+ # allocation spans two pages, then reference to farmost one can
+ # be punishable by SEGV. But page walking can do good even on
+ # other OSes, because it guarantees that villain thread hits
+ # the guard page before it can make damage to innocent one...
+ sub %r10,%r11
+ and \$-4096,%r11
+ lea (%r10,%r11),%rsp
+ mov (%rsp),%r11
+ cmp %r10,%rsp
+ ja .Lmul_page_walk
+ jmp .Lmul_page_walk_done
+
+.Lmul_page_walk:
+ lea -4096(%rsp),%rsp
+ mov (%rsp),%r11
+ cmp %r10,%rsp
+ ja .Lmul_page_walk
+.Lmul_page_walk_done:
+
+ lea .Linc(%rip),%r10
+ mov %rax,8(%rsp,$num,8) # tp[num+1]=%rsp
+.cfi_cfa_expression %rsp+8,$num,8,mul,plus,deref,+8
+.Lmul_body:
+
+ lea 128($bp),%r12 # reassign $bp (+size optimization)
+___
+ $bp="%r12";
+ $STRIDE=2**5*8; # 5 is "window size"
+ $N=$STRIDE/4; # should match cache line size
+$code.=<<___;
+ movdqa 0(%r10),%xmm0 # 00000001000000010000000000000000
+ movdqa 16(%r10),%xmm1 # 00000002000000020000000200000002
+ lea 24-112(%rsp,$num,8),%r10# place the mask after tp[num+3] (+ICache optimization)
+ and \$-16,%r10
+
+ pshufd \$0,%xmm5,%xmm5 # broadcast index
+ movdqa %xmm1,%xmm4
+ movdqa %xmm1,%xmm2
+___
+########################################################################
+# calculate mask by comparing 0..31 to index and save result to stack
+#
+$code.=<<___;
+ paddd %xmm0,%xmm1
+ pcmpeqd %xmm5,%xmm0 # compare to 1,0
+ .byte 0x67
+ movdqa %xmm4,%xmm3
+___
+for($k=0;$k<$STRIDE/16-4;$k+=4) {
+$code.=<<___;
+ paddd %xmm1,%xmm2
+ pcmpeqd %xmm5,%xmm1 # compare to 3,2
+ movdqa %xmm0,`16*($k+0)+112`(%r10)
+ movdqa %xmm4,%xmm0
+
+ paddd %xmm2,%xmm3
+ pcmpeqd %xmm5,%xmm2 # compare to 5,4
+ movdqa %xmm1,`16*($k+1)+112`(%r10)
+ movdqa %xmm4,%xmm1
+
+ paddd %xmm3,%xmm0
+ pcmpeqd %xmm5,%xmm3 # compare to 7,6
+ movdqa %xmm2,`16*($k+2)+112`(%r10)
+ movdqa %xmm4,%xmm2
+
+ paddd %xmm0,%xmm1
+ pcmpeqd %xmm5,%xmm0
+ movdqa %xmm3,`16*($k+3)+112`(%r10)
+ movdqa %xmm4,%xmm3
+___
+}
+$code.=<<___; # last iteration can be optimized
+ paddd %xmm1,%xmm2
+ pcmpeqd %xmm5,%xmm1
+ movdqa %xmm0,`16*($k+0)+112`(%r10)
+
+ paddd %xmm2,%xmm3
+ .byte 0x67
+ pcmpeqd %xmm5,%xmm2
+ movdqa %xmm1,`16*($k+1)+112`(%r10)
+
+ pcmpeqd %xmm5,%xmm3
+ movdqa %xmm2,`16*($k+2)+112`(%r10)
+ pand `16*($k+0)-128`($bp),%xmm0 # while it's still in register
+
+ pand `16*($k+1)-128`($bp),%xmm1
+ pand `16*($k+2)-128`($bp),%xmm2
+ movdqa %xmm3,`16*($k+3)+112`(%r10)
+ pand `16*($k+3)-128`($bp),%xmm3
+ por %xmm2,%xmm0
+ por %xmm3,%xmm1
+___
+for($k=0;$k<$STRIDE/16-4;$k+=4) {
+$code.=<<___;
+ movdqa `16*($k+0)-128`($bp),%xmm4
+ movdqa `16*($k+1)-128`($bp),%xmm5
+ movdqa `16*($k+2)-128`($bp),%xmm2
+ pand `16*($k+0)+112`(%r10),%xmm4
+ movdqa `16*($k+3)-128`($bp),%xmm3
+ pand `16*($k+1)+112`(%r10),%xmm5
+ por %xmm4,%xmm0
+ pand `16*($k+2)+112`(%r10),%xmm2
+ por %xmm5,%xmm1
+ pand `16*($k+3)+112`(%r10),%xmm3
+ por %xmm2,%xmm0
+ por %xmm3,%xmm1
+___
+}
+$code.=<<___;
+ por %xmm1,%xmm0
+ pshufd \$0x4e,%xmm0,%xmm1
+ por %xmm1,%xmm0
+ lea $STRIDE($bp),$bp
+ movq %xmm0,$m0 # m0=bp[0]
+
+ mov ($n0),$n0 # pull n0[0] value
+ mov ($ap),%rax
+
+ xor $i,$i # i=0
+ xor $j,$j # j=0
+
+ mov $n0,$m1
+ mulq $m0 # ap[0]*bp[0]
+ mov %rax,$lo0
+ mov ($np),%rax
+
+ imulq $lo0,$m1 # "tp[0]"*n0
+ mov %rdx,$hi0
+
+ mulq $m1 # np[0]*m1
+ add %rax,$lo0 # discarded
+ mov 8($ap),%rax
+ adc \$0,%rdx
+ mov %rdx,$hi1
+
+ lea 1($j),$j # j++
+ jmp .L1st_enter
+
+.align 16
+.L1st:
+ add %rax,$hi1
+ mov ($ap,$j,8),%rax
+ adc \$0,%rdx
+ add $hi0,$hi1 # np[j]*m1+ap[j]*bp[0]
+ mov $lo0,$hi0
+ adc \$0,%rdx
+ mov $hi1,-16(%rsp,$j,8) # tp[j-1]
+ mov %rdx,$hi1
+
+.L1st_enter:
+ mulq $m0 # ap[j]*bp[0]
+ add %rax,$hi0
+ mov ($np,$j,8),%rax
+ adc \$0,%rdx
+ lea 1($j),$j # j++
+ mov %rdx,$lo0
+
+ mulq $m1 # np[j]*m1
+ cmp $num,$j
+ jne .L1st # note that upon exit $j==$num, so
+ # they can be used interchangeably
+
+ add %rax,$hi1
+ adc \$0,%rdx
+ add $hi0,$hi1 # np[j]*m1+ap[j]*bp[0]
+ adc \$0,%rdx
+ mov $hi1,-16(%rsp,$num,8) # tp[num-1]
+ mov %rdx,$hi1
+ mov $lo0,$hi0
+
+ xor %rdx,%rdx
+ add $hi0,$hi1
+ adc \$0,%rdx
+ mov $hi1,-8(%rsp,$num,8)
+ mov %rdx,(%rsp,$num,8) # store upmost overflow bit
+
+ lea 1($i),$i # i++
+ jmp .Louter
+.align 16
+.Louter:
+ lea 24+128(%rsp,$num,8),%rdx # where 256-byte mask is (+size optimization)
+ and \$-16,%rdx
+ pxor %xmm4,%xmm4
+ pxor %xmm5,%xmm5
+___
+for($k=0;$k<$STRIDE/16;$k+=4) {
+$code.=<<___;
+ movdqa `16*($k+0)-128`($bp),%xmm0
+ movdqa `16*($k+1)-128`($bp),%xmm1
+ movdqa `16*($k+2)-128`($bp),%xmm2
+ movdqa `16*($k+3)-128`($bp),%xmm3
+ pand `16*($k+0)-128`(%rdx),%xmm0
+ pand `16*($k+1)-128`(%rdx),%xmm1
+ por %xmm0,%xmm4
+ pand `16*($k+2)-128`(%rdx),%xmm2
+ por %xmm1,%xmm5
+ pand `16*($k+3)-128`(%rdx),%xmm3
+ por %xmm2,%xmm4
+ por %xmm3,%xmm5
+___
+}
+$code.=<<___;
+ por %xmm5,%xmm4
+ pshufd \$0x4e,%xmm4,%xmm0
+ por %xmm4,%xmm0
+ lea $STRIDE($bp),$bp
+
+ mov ($ap),%rax # ap[0]
+ movq %xmm0,$m0 # m0=bp[i]
+
+ xor $j,$j # j=0
+ mov $n0,$m1
+ mov (%rsp),$lo0
+
+ mulq $m0 # ap[0]*bp[i]
+ add %rax,$lo0 # ap[0]*bp[i]+tp[0]
+ mov ($np),%rax
+ adc \$0,%rdx
+
+ imulq $lo0,$m1 # tp[0]*n0
+ mov %rdx,$hi0
+
+ mulq $m1 # np[0]*m1
+ add %rax,$lo0 # discarded
+ mov 8($ap),%rax
+ adc \$0,%rdx
+ mov 8(%rsp),$lo0 # tp[1]
+ mov %rdx,$hi1
+
+ lea 1($j),$j # j++
+ jmp .Linner_enter
+
+.align 16
+.Linner:
+ add %rax,$hi1
+ mov ($ap,$j,8),%rax
+ adc \$0,%rdx
+ add $lo0,$hi1 # np[j]*m1+ap[j]*bp[i]+tp[j]
+ mov (%rsp,$j,8),$lo0
+ adc \$0,%rdx
+ mov $hi1,-16(%rsp,$j,8) # tp[j-1]
+ mov %rdx,$hi1
+
+.Linner_enter:
+ mulq $m0 # ap[j]*bp[i]
+ add %rax,$hi0
+ mov ($np,$j,8),%rax
+ adc \$0,%rdx
+ add $hi0,$lo0 # ap[j]*bp[i]+tp[j]
+ mov %rdx,$hi0
+ adc \$0,$hi0
+ lea 1($j),$j # j++
+
+ mulq $m1 # np[j]*m1
+ cmp $num,$j
+ jne .Linner # note that upon exit $j==$num, so
+ # they can be used interchangeably
+ add %rax,$hi1
+ adc \$0,%rdx
+ add $lo0,$hi1 # np[j]*m1+ap[j]*bp[i]+tp[j]
+ mov (%rsp,$num,8),$lo0
+ adc \$0,%rdx
+ mov $hi1,-16(%rsp,$num,8) # tp[num-1]
+ mov %rdx,$hi1
+
+ xor %rdx,%rdx
+ add $hi0,$hi1
+ adc \$0,%rdx
+ add $lo0,$hi1 # pull upmost overflow bit
+ adc \$0,%rdx
+ mov $hi1,-8(%rsp,$num,8)
+ mov %rdx,(%rsp,$num,8) # store upmost overflow bit
+
+ lea 1($i),$i # i++
+ cmp $num,$i
+ jb .Louter
+
+ xor $i,$i # i=0 and clear CF!
+ mov (%rsp),%rax # tp[0]
+ lea (%rsp),$ap # borrow ap for tp
+ mov $num,$j # j=num
+ jmp .Lsub
+.align 16
+.Lsub: sbb ($np,$i,8),%rax
+ mov %rax,($rp,$i,8) # rp[i]=tp[i]-np[i]
+ mov 8($ap,$i,8),%rax # tp[i+1]
+ lea 1($i),$i # i++
+ dec $j # doesn't affect CF!
+ jnz .Lsub
+
+ sbb \$0,%rax # handle upmost overflow bit
+ mov \$-1,%rbx
+ xor %rax,%rbx
+ xor $i,$i
+ mov $num,$j # j=num
+
+.Lcopy: # conditional copy
+ mov ($rp,$i,8),%rcx
+ mov (%rsp,$i,8),%rdx
+ and %rbx,%rcx
+ and %rax,%rdx
+ mov $i,(%rsp,$i,8) # zap temporary vector
+ or %rcx,%rdx
+ mov %rdx,($rp,$i,8) # rp[i]=tp[i]
+ lea 1($i),$i
+ sub \$1,$j
+ jnz .Lcopy
+
+ mov 8(%rsp,$num,8),%rsi # restore %rsp
+.cfi_def_cfa %rsi,8
+ mov \$1,%rax
+
+ mov -48(%rsi),%r15
+.cfi_restore %r15
+ mov -40(%rsi),%r14
+.cfi_restore %r14
+ mov -32(%rsi),%r13
+.cfi_restore %r13
+ mov -24(%rsi),%r12
+.cfi_restore %r12
+ mov -16(%rsi),%rbp
+.cfi_restore %rbp
+ mov -8(%rsi),%rbx
+.cfi_restore %rbx
+ lea (%rsi),%rsp
+.cfi_def_cfa_register %rsp
+.Lmul_epilogue:
+ ret
+.cfi_endproc
+.size bn_mul_mont_gather5,.-bn_mul_mont_gather5
+___
+{{{
+my @A=("%r10","%r11");
+my @N=("%r13","%rdi");
+$code.=<<___;
+.type bn_mul4x_mont_gather5,\@function,6
+.align 32
+bn_mul4x_mont_gather5:
+.cfi_startproc
+ .byte 0x67
+ mov %rsp,%rax
+.cfi_def_cfa_register %rax
+.Lmul4x_enter:
+___
+$code.=<<___ if ($addx);
+ and \$0x80108,%r11d
+ cmp \$0x80108,%r11d # check for AD*X+BMI2+BMI1
+ je .Lmulx4x_enter
+___
+$code.=<<___;
+ push %rbx
+.cfi_push %rbx
+ push %rbp
+.cfi_push %rbp
+ push %r12
+.cfi_push %r12
+ push %r13
+.cfi_push %r13
+ push %r14
+.cfi_push %r14
+ push %r15
+.cfi_push %r15
+.Lmul4x_prologue:
+
+ .byte 0x67
+ shl \$3,${num}d # convert $num to bytes
+ lea ($num,$num,2),%r10 # 3*$num in bytes
+ neg $num # -$num
+
+ ##############################################################
+ # Ensure that stack frame doesn't alias with $rptr+3*$num
+ # modulo 4096, which covers ret[num], am[num] and n[num]
+ # (see bn_exp.c). This is done to allow memory disambiguation
+ # logic do its magic. [Extra [num] is allocated in order
+ # to align with bn_power5's frame, which is cleansed after
+ # completing exponentiation. Extra 256 bytes is for power mask
+ # calculated from 7th argument, the index.]
+ #
+ lea -320(%rsp,$num,2),%r11
+ mov %rsp,%rbp
+ sub $rp,%r11
+ and \$4095,%r11
+ cmp %r11,%r10
+ jb .Lmul4xsp_alt
+ sub %r11,%rbp # align with $rp
+ lea -320(%rbp,$num,2),%rbp # future alloca(frame+2*num*8+256)
+ jmp .Lmul4xsp_done
+
+.align 32
+.Lmul4xsp_alt:
+ lea 4096-320(,$num,2),%r10
+ lea -320(%rbp,$num,2),%rbp # future alloca(frame+2*num*8+256)
+ sub %r10,%r11
+ mov \$0,%r10
+ cmovc %r10,%r11
+ sub %r11,%rbp
+.Lmul4xsp_done:
+ and \$-64,%rbp
+ mov %rsp,%r11
+ sub %rbp,%r11
+ and \$-4096,%r11
+ lea (%rbp,%r11),%rsp
+ mov (%rsp),%r10
+ cmp %rbp,%rsp
+ ja .Lmul4x_page_walk
+ jmp .Lmul4x_page_walk_done
+
+.Lmul4x_page_walk:
+ lea -4096(%rsp),%rsp
+ mov (%rsp),%r10
+ cmp %rbp,%rsp
+ ja .Lmul4x_page_walk
+.Lmul4x_page_walk_done:
+
+ neg $num
+
+ mov %rax,40(%rsp)
+.cfi_cfa_expression %rsp+40,deref,+8
+.Lmul4x_body:
+
+ call mul4x_internal
+
+ mov 40(%rsp),%rsi # restore %rsp
+.cfi_def_cfa %rsi,8
+ mov \$1,%rax
+
+ mov -48(%rsi),%r15
+.cfi_restore %r15
+ mov -40(%rsi),%r14
+.cfi_restore %r14
+ mov -32(%rsi),%r13
+.cfi_restore %r13
+ mov -24(%rsi),%r12
+.cfi_restore %r12
+ mov -16(%rsi),%rbp
+.cfi_restore %rbp
+ mov -8(%rsi),%rbx
+.cfi_restore %rbx
+ lea (%rsi),%rsp
+.cfi_def_cfa_register %rsp
+.Lmul4x_epilogue:
+ ret
+.cfi_endproc
+.size bn_mul4x_mont_gather5,.-bn_mul4x_mont_gather5
+
+.type mul4x_internal,\@abi-omnipotent
+.align 32
+mul4x_internal:
+.cfi_startproc
+ shl \$5,$num # $num was in bytes
+ movd `($win64?56:8)`(%rax),%xmm5 # load 7th argument, index
+ lea .Linc(%rip),%rax
+ lea 128(%rdx,$num),%r13 # end of powers table (+size optimization)
+ shr \$5,$num # restore $num
+___
+ $bp="%r12";
+ $STRIDE=2**5*8; # 5 is "window size"
+ $N=$STRIDE/4; # should match cache line size
+ $tp=$i;
+$code.=<<___;
+ movdqa 0(%rax),%xmm0 # 00000001000000010000000000000000
+ movdqa 16(%rax),%xmm1 # 00000002000000020000000200000002
+ lea 88-112(%rsp,$num),%r10 # place the mask after tp[num+1] (+ICache optimization)
+ lea 128(%rdx),$bp # size optimization
+
+ pshufd \$0,%xmm5,%xmm5 # broadcast index
+ movdqa %xmm1,%xmm4
+ .byte 0x67,0x67
+ movdqa %xmm1,%xmm2
+___
+########################################################################
+# calculate mask by comparing 0..31 to index and save result to stack
+#
+$code.=<<___;
+ paddd %xmm0,%xmm1
+ pcmpeqd %xmm5,%xmm0 # compare to 1,0
+ .byte 0x67
+ movdqa %xmm4,%xmm3
+___
+for($i=0;$i<$STRIDE/16-4;$i+=4) {
+$code.=<<___;
+ paddd %xmm1,%xmm2
+ pcmpeqd %xmm5,%xmm1 # compare to 3,2
+ movdqa %xmm0,`16*($i+0)+112`(%r10)
+ movdqa %xmm4,%xmm0
+
+ paddd %xmm2,%xmm3
+ pcmpeqd %xmm5,%xmm2 # compare to 5,4
+ movdqa %xmm1,`16*($i+1)+112`(%r10)
+ movdqa %xmm4,%xmm1
+
+ paddd %xmm3,%xmm0
+ pcmpeqd %xmm5,%xmm3 # compare to 7,6
+ movdqa %xmm2,`16*($i+2)+112`(%r10)
+ movdqa %xmm4,%xmm2
+
+ paddd %xmm0,%xmm1
+ pcmpeqd %xmm5,%xmm0
+ movdqa %xmm3,`16*($i+3)+112`(%r10)
+ movdqa %xmm4,%xmm3
+___
+}
+$code.=<<___; # last iteration can be optimized
+ paddd %xmm1,%xmm2
+ pcmpeqd %xmm5,%xmm1
+ movdqa %xmm0,`16*($i+0)+112`(%r10)
+
+ paddd %xmm2,%xmm3
+ .byte 0x67
+ pcmpeqd %xmm5,%xmm2
+ movdqa %xmm1,`16*($i+1)+112`(%r10)
+
+ pcmpeqd %xmm5,%xmm3
+ movdqa %xmm2,`16*($i+2)+112`(%r10)
+ pand `16*($i+0)-128`($bp),%xmm0 # while it's still in register
+
+ pand `16*($i+1)-128`($bp),%xmm1
+ pand `16*($i+2)-128`($bp),%xmm2
+ movdqa %xmm3,`16*($i+3)+112`(%r10)
+ pand `16*($i+3)-128`($bp),%xmm3
+ por %xmm2,%xmm0
+ por %xmm3,%xmm1
+___
+for($i=0;$i<$STRIDE/16-4;$i+=4) {
+$code.=<<___;
+ movdqa `16*($i+0)-128`($bp),%xmm4
+ movdqa `16*($i+1)-128`($bp),%xmm5
+ movdqa `16*($i+2)-128`($bp),%xmm2
+ pand `16*($i+0)+112`(%r10),%xmm4
+ movdqa `16*($i+3)-128`($bp),%xmm3
+ pand `16*($i+1)+112`(%r10),%xmm5
+ por %xmm4,%xmm0
+ pand `16*($i+2)+112`(%r10),%xmm2
+ por %xmm5,%xmm1
+ pand `16*($i+3)+112`(%r10),%xmm3
+ por %xmm2,%xmm0
+ por %xmm3,%xmm1
+___
+}
+$code.=<<___;
+ por %xmm1,%xmm0
+ pshufd \$0x4e,%xmm0,%xmm1
+ por %xmm1,%xmm0
+ lea $STRIDE($bp),$bp
+ movq %xmm0,$m0 # m0=bp[0]
+
+ mov %r13,16+8(%rsp) # save end of b[num]
+ mov $rp, 56+8(%rsp) # save $rp
+
+ mov ($n0),$n0 # pull n0[0] value
+ mov ($ap),%rax
+ lea ($ap,$num),$ap # end of a[num]
+ neg $num
+
+ mov $n0,$m1
+ mulq $m0 # ap[0]*bp[0]
+ mov %rax,$A[0]
+ mov ($np),%rax
+
+ imulq $A[0],$m1 # "tp[0]"*n0
+ lea 64+8(%rsp),$tp
+ mov %rdx,$A[1]
+
+ mulq $m1 # np[0]*m1
+ add %rax,$A[0] # discarded
+ mov 8($ap,$num),%rax
+ adc \$0,%rdx
+ mov %rdx,$N[1]
+
+ mulq $m0
+ add %rax,$A[1]
+ mov 8*1($np),%rax
+ adc \$0,%rdx
+ mov %rdx,$A[0]
+
+ mulq $m1
+ add %rax,$N[1]
+ mov 16($ap,$num),%rax
+ adc \$0,%rdx
+ add $A[1],$N[1]
+ lea 4*8($num),$j # j=4
+ lea 8*4($np),$np
+ adc \$0,%rdx
+ mov $N[1],($tp)
+ mov %rdx,$N[0]
+ jmp .L1st4x
+
+.align 32
+.L1st4x:
+ mulq $m0 # ap[j]*bp[0]
+ add %rax,$A[0]
+ mov -8*2($np),%rax
+ lea 32($tp),$tp
+ adc \$0,%rdx
+ mov %rdx,$A[1]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[0]
+ mov -8($ap,$j),%rax
+ adc \$0,%rdx
+ add $A[0],$N[0] # np[j]*m1+ap[j]*bp[0]
+ adc \$0,%rdx
+ mov $N[0],-24($tp) # tp[j-1]
+ mov %rdx,$N[1]
+
+ mulq $m0 # ap[j]*bp[0]
+ add %rax,$A[1]
+ mov -8*1($np),%rax
+ adc \$0,%rdx
+ mov %rdx,$A[0]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[1]
+ mov ($ap,$j),%rax
+ adc \$0,%rdx
+ add $A[1],$N[1] # np[j]*m1+ap[j]*bp[0]
+ adc \$0,%rdx
+ mov $N[1],-16($tp) # tp[j-1]
+ mov %rdx,$N[0]
+
+ mulq $m0 # ap[j]*bp[0]
+ add %rax,$A[0]
+ mov 8*0($np),%rax
+ adc \$0,%rdx
+ mov %rdx,$A[1]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[0]
+ mov 8($ap,$j),%rax
+ adc \$0,%rdx
+ add $A[0],$N[0] # np[j]*m1+ap[j]*bp[0]
+ adc \$0,%rdx
+ mov $N[0],-8($tp) # tp[j-1]
+ mov %rdx,$N[1]
+
+ mulq $m0 # ap[j]*bp[0]
+ add %rax,$A[1]
+ mov 8*1($np),%rax
+ adc \$0,%rdx
+ mov %rdx,$A[0]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[1]
+ mov 16($ap,$j),%rax
+ adc \$0,%rdx
+ add $A[1],$N[1] # np[j]*m1+ap[j]*bp[0]
+ lea 8*4($np),$np
+ adc \$0,%rdx
+ mov $N[1],($tp) # tp[j-1]
+ mov %rdx,$N[0]
+
+ add \$32,$j # j+=4
+ jnz .L1st4x
+
+ mulq $m0 # ap[j]*bp[0]
+ add %rax,$A[0]
+ mov -8*2($np),%rax
+ lea 32($tp),$tp
+ adc \$0,%rdx
+ mov %rdx,$A[1]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[0]
+ mov -8($ap),%rax
+ adc \$0,%rdx
+ add $A[0],$N[0] # np[j]*m1+ap[j]*bp[0]
+ adc \$0,%rdx
+ mov $N[0],-24($tp) # tp[j-1]
+ mov %rdx,$N[1]
+
+ mulq $m0 # ap[j]*bp[0]
+ add %rax,$A[1]
+ mov -8*1($np),%rax
+ adc \$0,%rdx
+ mov %rdx,$A[0]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[1]
+ mov ($ap,$num),%rax # ap[0]
+ adc \$0,%rdx
+ add $A[1],$N[1] # np[j]*m1+ap[j]*bp[0]
+ adc \$0,%rdx
+ mov $N[1],-16($tp) # tp[j-1]
+ mov %rdx,$N[0]
+
+ lea ($np,$num),$np # rewind $np
+
+ xor $N[1],$N[1]
+ add $A[0],$N[0]
+ adc \$0,$N[1]
+ mov $N[0],-8($tp)
+
+ jmp .Louter4x
+
+.align 32
+.Louter4x:
+ lea 16+128($tp),%rdx # where 256-byte mask is (+size optimization)
+ pxor %xmm4,%xmm4
+ pxor %xmm5,%xmm5
+___
+for($i=0;$i<$STRIDE/16;$i+=4) {
+$code.=<<___;
+ movdqa `16*($i+0)-128`($bp),%xmm0
+ movdqa `16*($i+1)-128`($bp),%xmm1
+ movdqa `16*($i+2)-128`($bp),%xmm2
+ movdqa `16*($i+3)-128`($bp),%xmm3
+ pand `16*($i+0)-128`(%rdx),%xmm0
+ pand `16*($i+1)-128`(%rdx),%xmm1
+ por %xmm0,%xmm4
+ pand `16*($i+2)-128`(%rdx),%xmm2
+ por %xmm1,%xmm5
+ pand `16*($i+3)-128`(%rdx),%xmm3
+ por %xmm2,%xmm4
+ por %xmm3,%xmm5
+___
+}
+$code.=<<___;
+ por %xmm5,%xmm4
+ pshufd \$0x4e,%xmm4,%xmm0
+ por %xmm4,%xmm0
+ lea $STRIDE($bp),$bp
+ movq %xmm0,$m0 # m0=bp[i]
+
+ mov ($tp,$num),$A[0]
+ mov $n0,$m1
+ mulq $m0 # ap[0]*bp[i]
+ add %rax,$A[0] # ap[0]*bp[i]+tp[0]
+ mov ($np),%rax
+ adc \$0,%rdx
+
+ imulq $A[0],$m1 # tp[0]*n0
+ mov %rdx,$A[1]
+ mov $N[1],($tp) # store upmost overflow bit
+
+ lea ($tp,$num),$tp # rewind $tp
+
+ mulq $m1 # np[0]*m1
+ add %rax,$A[0] # "$N[0]", discarded
+ mov 8($ap,$num),%rax
+ adc \$0,%rdx
+ mov %rdx,$N[1]
+
+ mulq $m0 # ap[j]*bp[i]
+ add %rax,$A[1]
+ mov 8*1($np),%rax
+ adc \$0,%rdx
+ add 8($tp),$A[1] # +tp[1]
+ adc \$0,%rdx
+ mov %rdx,$A[0]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[1]
+ mov 16($ap,$num),%rax
+ adc \$0,%rdx
+ add $A[1],$N[1] # np[j]*m1+ap[j]*bp[i]+tp[j]
+ lea 4*8($num),$j # j=4
+ lea 8*4($np),$np
+ adc \$0,%rdx
+ mov %rdx,$N[0]
+ jmp .Linner4x
+
+.align 32
+.Linner4x:
+ mulq $m0 # ap[j]*bp[i]
+ add %rax,$A[0]
+ mov -8*2($np),%rax
+ adc \$0,%rdx
+ add 16($tp),$A[0] # ap[j]*bp[i]+tp[j]
+ lea 32($tp),$tp
+ adc \$0,%rdx
+ mov %rdx,$A[1]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[0]
+ mov -8($ap,$j),%rax
+ adc \$0,%rdx
+ add $A[0],$N[0]
+ adc \$0,%rdx
+ mov $N[1],-32($tp) # tp[j-1]
+ mov %rdx,$N[1]
+
+ mulq $m0 # ap[j]*bp[i]
+ add %rax,$A[1]
+ mov -8*1($np),%rax
+ adc \$0,%rdx
+ add -8($tp),$A[1]
+ adc \$0,%rdx
+ mov %rdx,$A[0]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[1]
+ mov ($ap,$j),%rax
+ adc \$0,%rdx
+ add $A[1],$N[1]
+ adc \$0,%rdx
+ mov $N[0],-24($tp) # tp[j-1]
+ mov %rdx,$N[0]
+
+ mulq $m0 # ap[j]*bp[i]
+ add %rax,$A[0]
+ mov 8*0($np),%rax
+ adc \$0,%rdx
+ add ($tp),$A[0] # ap[j]*bp[i]+tp[j]
+ adc \$0,%rdx
+ mov %rdx,$A[1]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[0]
+ mov 8($ap,$j),%rax
+ adc \$0,%rdx
+ add $A[0],$N[0]
+ adc \$0,%rdx
+ mov $N[1],-16($tp) # tp[j-1]
+ mov %rdx,$N[1]
+
+ mulq $m0 # ap[j]*bp[i]
+ add %rax,$A[1]
+ mov 8*1($np),%rax
+ adc \$0,%rdx
+ add 8($tp),$A[1]
+ adc \$0,%rdx
+ mov %rdx,$A[0]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[1]
+ mov 16($ap,$j),%rax
+ adc \$0,%rdx
+ add $A[1],$N[1]
+ lea 8*4($np),$np
+ adc \$0,%rdx
+ mov $N[0],-8($tp) # tp[j-1]
+ mov %rdx,$N[0]
+
+ add \$32,$j # j+=4
+ jnz .Linner4x
+
+ mulq $m0 # ap[j]*bp[i]
+ add %rax,$A[0]
+ mov -8*2($np),%rax
+ adc \$0,%rdx
+ add 16($tp),$A[0] # ap[j]*bp[i]+tp[j]
+ lea 32($tp),$tp
+ adc \$0,%rdx
+ mov %rdx,$A[1]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[0]
+ mov -8($ap),%rax
+ adc \$0,%rdx
+ add $A[0],$N[0]
+ adc \$0,%rdx
+ mov $N[1],-32($tp) # tp[j-1]
+ mov %rdx,$N[1]
+
+ mulq $m0 # ap[j]*bp[i]
+ add %rax,$A[1]
+ mov $m1,%rax
+ mov -8*1($np),$m1
+ adc \$0,%rdx
+ add -8($tp),$A[1]
+ adc \$0,%rdx
+ mov %rdx,$A[0]
+
+ mulq $m1 # np[j]*m1
+ add %rax,$N[1]
+ mov ($ap,$num),%rax # ap[0]
+ adc \$0,%rdx
+ add $A[1],$N[1]
+ adc \$0,%rdx
+ mov $N[0],-24($tp) # tp[j-1]
+ mov %rdx,$N[0]
+
+ mov $N[1],-16($tp) # tp[j-1]
+ lea ($np,$num),$np # rewind $np
+
+ xor $N[1],$N[1]
+ add $A[0],$N[0]
+ adc \$0,$N[1]
+ add ($tp),$N[0] # pull upmost overflow bit
+ adc \$0,$N[1] # upmost overflow bit
+ mov $N[0],-8($tp)
+
+ cmp 16+8(%rsp),$bp
+ jb .Louter4x
+___
+if (1) {
+$code.=<<___;
+ xor %rax,%rax
+ sub $N[0],$m1 # compare top-most words
+ adc $j,$j # $j is zero
+ or $j,$N[1]
+ sub $N[1],%rax # %rax=-$N[1]
+ lea ($tp,$num),%rbx # tptr in .sqr4x_sub
+ mov ($np),%r12
+ lea ($np),%rbp # nptr in .sqr4x_sub
+ mov %r9,%rcx
+ sar \$3+2,%rcx
+ mov 56+8(%rsp),%rdi # rptr in .sqr4x_sub
+ dec %r12 # so that after 'not' we get -n[0]
+ xor %r10,%r10
+ mov 8*1(%rbp),%r13
+ mov 8*2(%rbp),%r14
+ mov 8*3(%rbp),%r15
+ jmp .Lsqr4x_sub_entry
+___
+} else {
+my @ri=("%rax",$bp,$m0,$m1);
+my $rp="%rdx";
+$code.=<<___
+ xor \$1,$N[1]
+ lea ($tp,$num),$tp # rewind $tp
+ sar \$5,$num # cf=0
+ lea ($np,$N[1],8),$np
+ mov 56+8(%rsp),$rp # restore $rp
+ jmp .Lsub4x
+
+.align 32
+.Lsub4x:
+ .byte 0x66
+ mov 8*0($tp),@ri[0]
+ mov 8*1($tp),@ri[1]
+ .byte 0x66
+ sbb 16*0($np),@ri[0]
+ mov 8*2($tp),@ri[2]
+ sbb 16*1($np),@ri[1]
+ mov 3*8($tp),@ri[3]
+ lea 4*8($tp),$tp
+ sbb 16*2($np),@ri[2]
+ mov @ri[0],8*0($rp)
+ sbb 16*3($np),@ri[3]
+ lea 16*4($np),$np
+ mov @ri[1],8*1($rp)
+ mov @ri[2],8*2($rp)
+ mov @ri[3],8*3($rp)
+ lea 8*4($rp),$rp
+
+ inc $num
+ jnz .Lsub4x
+
+ ret
+___
+}
+$code.=<<___;
+.cfi_endproc
+.size mul4x_internal,.-mul4x_internal
+___
+}}}
+ {{{
+######################################################################
+# void bn_power5(
+my $rptr="%rdi"; # BN_ULONG *rptr,
+my $aptr="%rsi"; # const BN_ULONG *aptr,
+my $bptr="%rdx"; # const void *table,
+my $nptr="%rcx"; # const BN_ULONG *nptr,
+my $n0 ="%r8"; # const BN_ULONG *n0);
+my $num ="%r9"; # int num, has to be divisible by 8
+ # int pwr
+
+my ($i,$j,$tptr)=("%rbp","%rcx",$rptr);
+my @A0=("%r10","%r11");
+my @A1=("%r12","%r13");
+my ($a0,$a1,$ai)=("%r14","%r15","%rbx");
+
+$code.=<<___;
+.globl bn_power5
+.type bn_power5,\@function,6
+.align 32
+bn_power5:
+.cfi_startproc
+ mov %rsp,%rax
+.cfi_def_cfa_register %rax
+___
+$code.=<<___ if ($addx);
+ mov OPENSSL_ia32cap_P+8(%rip),%r11d
+ and \$0x80108,%r11d
+ cmp \$0x80108,%r11d # check for AD*X+BMI2+BMI1
+ je .Lpowerx5_enter
+___
+$code.=<<___;
+ push %rbx
+.cfi_push %rbx
+ push %rbp
+.cfi_push %rbp
+ push %r12
+.cfi_push %r12
+ push %r13
+.cfi_push %r13
+ push %r14
+.cfi_push %r14
+ push %r15
+.cfi_push %r15
+.Lpower5_prologue:
+
+ shl \$3,${num}d # convert $num to bytes
+ lea ($num,$num,2),%r10d # 3*$num
+ neg $num
+ mov ($n0),$n0 # *n0
+
+ ##############################################################
+ # Ensure that stack frame doesn't alias with $rptr+3*$num
+ # modulo 4096, which covers ret[num], am[num] and n[num]
+ # (see bn_exp.c). This is done to allow memory disambiguation
+ # logic do its magic. [Extra 256 bytes is for power mask
+ # calculated from 7th argument, the index.]
+ #
+ lea -320(%rsp,$num,2),%r11
+ mov %rsp,%rbp
+ sub $rptr,%r11
+ and \$4095,%r11
+ cmp %r11,%r10
+ jb .Lpwr_sp_alt
+ sub %r11,%rbp # align with $aptr
+ lea -320(%rbp,$num,2),%rbp # future alloca(frame+2*num*8+256)
+ jmp .Lpwr_sp_done
+
+.align 32
+.Lpwr_sp_alt:
+ lea 4096-320(,$num,2),%r10
+ lea -320(%rbp,$num,2),%rbp # future alloca(frame+2*num*8+256)
+ sub %r10,%r11
+ mov \$0,%r10
+ cmovc %r10,%r11
+ sub %r11,%rbp
+.Lpwr_sp_done:
+ and \$-64,%rbp
+ mov %rsp,%r11
+ sub %rbp,%r11
+ and \$-4096,%r11
+ lea (%rbp,%r11),%rsp
+ mov (%rsp),%r10
+ cmp %rbp,%rsp
+ ja .Lpwr_page_walk
+ jmp .Lpwr_page_walk_done
+
+.Lpwr_page_walk:
+ lea -4096(%rsp),%rsp
+ mov (%rsp),%r10
+ cmp %rbp,%rsp
+ ja .Lpwr_page_walk
+.Lpwr_page_walk_done:
+
+ mov $num,%r10
+ neg $num
+
+ ##############################################################
+ # Stack layout
+ #
+ # +0 saved $num, used in reduction section
+ # +8 &t[2*$num], used in reduction section
+ # +32 saved *n0
+ # +40 saved %rsp
+ # +48 t[2*$num]
+ #
+ mov $n0, 32(%rsp)
+ mov %rax, 40(%rsp) # save original %rsp
+.cfi_cfa_expression %rsp+40,deref,+8
+.Lpower5_body:
+ movq $rptr,%xmm1 # save $rptr, used in sqr8x
+ movq $nptr,%xmm2 # save $nptr
+ movq %r10, %xmm3 # -$num, used in sqr8x
+ movq $bptr,%xmm4
+
+ call __bn_sqr8x_internal
+ call __bn_post4x_internal
+ call __bn_sqr8x_internal
+ call __bn_post4x_internal
+ call __bn_sqr8x_internal
+ call __bn_post4x_internal
+ call __bn_sqr8x_internal
+ call __bn_post4x_internal
+ call __bn_sqr8x_internal
+ call __bn_post4x_internal
+
+ movq %xmm2,$nptr
+ movq %xmm4,$bptr
+ mov $aptr,$rptr
+ mov 40(%rsp),%rax
+ lea 32(%rsp),$n0
+
+ call mul4x_internal
+
+ mov 40(%rsp),%rsi # restore %rsp
+.cfi_def_cfa %rsi,8
+ mov \$1,%rax
+ mov -48(%rsi),%r15
+.cfi_restore %r15
+ mov -40(%rsi),%r14
+.cfi_restore %r14
+ mov -32(%rsi),%r13
+.cfi_restore %r13
+ mov -24(%rsi),%r12
+.cfi_restore %r12
+ mov -16(%rsi),%rbp
+.cfi_restore %rbp
+ mov -8(%rsi),%rbx
+.cfi_restore %rbx
+ lea (%rsi),%rsp
+.cfi_def_cfa_register %rsp
+.Lpower5_epilogue:
+ ret
+.cfi_endproc
+.size bn_power5,.-bn_power5
+
+.globl bn_sqr8x_internal
+.hidden bn_sqr8x_internal
+.type bn_sqr8x_internal,\@abi-omnipotent
+.align 32
+bn_sqr8x_internal:
+__bn_sqr8x_internal:
+.cfi_startproc
+ ##############################################################
+ # Squaring part:
+ #
+ # a) multiply-n-add everything but a[i]*a[i];
+ # b) shift result of a) by 1 to the left and accumulate
+ # a[i]*a[i] products;
+ #
+ ##############################################################
+ # a[1]a[0]
+ # a[2]a[0]
+ # a[3]a[0]
+ # a[2]a[1]
+ # a[4]a[0]
+ # a[3]a[1]
+ # a[5]a[0]
+ # a[4]a[1]
+ # a[3]a[2]
+ # a[6]a[0]
+ # a[5]a[1]
+ # a[4]a[2]
+ # a[7]a[0]
+ # a[6]a[1]
+ # a[5]a[2]
+ # a[4]a[3]
+ # a[7]a[1]
+ # a[6]a[2]
+ # a[5]a[3]
+ # a[7]a[2]
+ # a[6]a[3]
+ # a[5]a[4]
+ # a[7]a[3]
+ # a[6]a[4]
+ # a[7]a[4]
+ # a[6]a[5]
+ # a[7]a[5]
+ # a[7]a[6]
+ # a[1]a[0]
+ # a[2]a[0]
+ # a[3]a[0]
+ # a[4]a[0]
+ # a[5]a[0]
+ # a[6]a[0]
+ # a[7]a[0]
+ # a[2]a[1]
+ # a[3]a[1]
+ # a[4]a[1]
+ # a[5]a[1]
+ # a[6]a[1]
+ # a[7]a[1]
+ # a[3]a[2]
+ # a[4]a[2]
+ # a[5]a[2]
+ # a[6]a[2]
+ # a[7]a[2]
+ # a[4]a[3]
+ # a[5]a[3]
+ # a[6]a[3]
+ # a[7]a[3]
+ # a[5]a[4]
+ # a[6]a[4]
+ # a[7]a[4]
+ # a[6]a[5]
+ # a[7]a[5]
+ # a[7]a[6]
+ # a[0]a[0]
+ # a[1]a[1]
+ # a[2]a[2]
+ # a[3]a[3]
+ # a[4]a[4]
+ # a[5]a[5]
+ # a[6]a[6]
+ # a[7]a[7]
+
+ lea 32(%r10),$i # $i=-($num-32)
+ lea ($aptr,$num),$aptr # end of a[] buffer, ($aptr,$i)=&ap[2]
+
+ mov $num,$j # $j=$num
+
+ # comments apply to $num==8 case
+ mov -32($aptr,$i),$a0 # a[0]
+ lea 48+8(%rsp,$num,2),$tptr # end of tp[] buffer, &tp[2*$num]
+ mov -24($aptr,$i),%rax # a[1]
+ lea -32($tptr,$i),$tptr # end of tp[] window, &tp[2*$num-"$i"]
+ mov -16($aptr,$i),$ai # a[2]
+ mov %rax,$a1
+
+ mul $a0 # a[1]*a[0]
+ mov %rax,$A0[0] # a[1]*a[0]
+ mov $ai,%rax # a[2]
+ mov %rdx,$A0[1]
+ mov $A0[0],-24($tptr,$i) # t[1]
+
+ mul $a0 # a[2]*a[0]
+ add %rax,$A0[1]
+ mov $ai,%rax
+ adc \$0,%rdx
+ mov $A0[1],-16($tptr,$i) # t[2]
+ mov %rdx,$A0[0]
+
+
+ mov -8($aptr,$i),$ai # a[3]
+ mul $a1 # a[2]*a[1]
+ mov %rax,$A1[0] # a[2]*a[1]+t[3]
+ mov $ai,%rax
+ mov %rdx,$A1[1]
+
+ lea ($i),$j
+ mul $a0 # a[3]*a[0]
+ add %rax,$A0[0] # a[3]*a[0]+a[2]*a[1]+t[3]
+ mov $ai,%rax
+ mov %rdx,$A0[1]
+ adc \$0,$A0[1]
+ add $A1[0],$A0[0]
+ adc \$0,$A0[1]
+ mov $A0[0],-8($tptr,$j) # t[3]
+ jmp .Lsqr4x_1st
+
+.align 32
+.Lsqr4x_1st:
+ mov ($aptr,$j),$ai # a[4]
+ mul $a1 # a[3]*a[1]
+ add %rax,$A1[1] # a[3]*a[1]+t[4]
+ mov $ai,%rax
+ mov %rdx,$A1[0]
+ adc \$0,$A1[0]
+
+ mul $a0 # a[4]*a[0]
+ add %rax,$A0[1] # a[4]*a[0]+a[3]*a[1]+t[4]
+ mov $ai,%rax # a[3]
+ mov 8($aptr,$j),$ai # a[5]
+ mov %rdx,$A0[0]
+ adc \$0,$A0[0]
+ add $A1[1],$A0[1]
+ adc \$0,$A0[0]
+
+
+ mul $a1 # a[4]*a[3]
+ add %rax,$A1[0] # a[4]*a[3]+t[5]
+ mov $ai,%rax
+ mov $A0[1],($tptr,$j) # t[4]
+ mov %rdx,$A1[1]
+ adc \$0,$A1[1]
+
+ mul $a0 # a[5]*a[2]
+ add %rax,$A0[0] # a[5]*a[2]+a[4]*a[3]+t[5]
+ mov $ai,%rax
+ mov 16($aptr,$j),$ai # a[6]
+ mov %rdx,$A0[1]
+ adc \$0,$A0[1]
+ add $A1[0],$A0[0]
+ adc \$0,$A0[1]
+
+ mul $a1 # a[5]*a[3]
+ add %rax,$A1[1] # a[5]*a[3]+t[6]
+ mov $ai,%rax
+ mov $A0[0],8($tptr,$j) # t[5]
+ mov %rdx,$A1[0]
+ adc \$0,$A1[0]
+
+ mul $a0 # a[6]*a[2]
+ add %rax,$A0[1] # a[6]*a[2]+a[5]*a[3]+t[6]
+ mov $ai,%rax # a[3]
+ mov 24($aptr,$j),$ai # a[7]
+ mov %rdx,$A0[0]
+ adc \$0,$A0[0]
+ add $A1[1],$A0[1]
+ adc \$0,$A0[0]
+
+
+ mul $a1 # a[6]*a[5]
+ add %rax,$A1[0] # a[6]*a[5]+t[7]
+ mov $ai,%rax
+ mov $A0[1],16($tptr,$j) # t[6]
+ mov %rdx,$A1[1]
+ adc \$0,$A1[1]
+ lea 32($j),$j
+
+ mul $a0 # a[7]*a[4]
+ add %rax,$A0[0] # a[7]*a[4]+a[6]*a[5]+t[6]
+ mov $ai,%rax
+ mov %rdx,$A0[1]
+ adc \$0,$A0[1]
+ add $A1[0],$A0[0]
+ adc \$0,$A0[1]
+ mov $A0[0],-8($tptr,$j) # t[7]
+
+ cmp \$0,$j
+ jne .Lsqr4x_1st
+
+ mul $a1 # a[7]*a[5]
+ add %rax,$A1[1]
+ lea 16($i),$i
+ adc \$0,%rdx
+ add $A0[1],$A1[1]
+ adc \$0,%rdx
+
+ mov $A1[1],($tptr) # t[8]
+ mov %rdx,$A1[0]
+ mov %rdx,8($tptr) # t[9]
+ jmp .Lsqr4x_outer
+
+.align 32
+.Lsqr4x_outer: # comments apply to $num==6 case
+ mov -32($aptr,$i),$a0 # a[0]
+ lea 48+8(%rsp,$num,2),$tptr # end of tp[] buffer, &tp[2*$num]
+ mov -24($aptr,$i),%rax # a[1]
+ lea -32($tptr,$i),$tptr # end of tp[] window, &tp[2*$num-"$i"]
+ mov -16($aptr,$i),$ai # a[2]
+ mov %rax,$a1
+
+ mul $a0 # a[1]*a[0]
+ mov -24($tptr,$i),$A0[0] # t[1]
+ add %rax,$A0[0] # a[1]*a[0]+t[1]
+ mov $ai,%rax # a[2]
+ adc \$0,%rdx
+ mov $A0[0],-24($tptr,$i) # t[1]
+ mov %rdx,$A0[1]
+
+ mul $a0 # a[2]*a[0]
+ add %rax,$A0[1]
+ mov $ai,%rax
+ adc \$0,%rdx
+ add -16($tptr,$i),$A0[1] # a[2]*a[0]+t[2]
+ mov %rdx,$A0[0]
+ adc \$0,$A0[0]
+ mov $A0[1],-16($tptr,$i) # t[2]
+
+ xor $A1[0],$A1[0]
+
+ mov -8($aptr,$i),$ai # a[3]
+ mul $a1 # a[2]*a[1]
+ add %rax,$A1[0] # a[2]*a[1]+t[3]
+ mov $ai,%rax
+ adc \$0,%rdx
+ add -8($tptr,$i),$A1[0]
+ mov %rdx,$A1[1]
+ adc \$0,$A1[1]
+
+ mul $a0 # a[3]*a[0]
+ add %rax,$A0[0] # a[3]*a[0]+a[2]*a[1]+t[3]
+ mov $ai,%rax
+ adc \$0,%rdx
+ add $A1[0],$A0[0]
+ mov %rdx,$A0[1]
+ adc \$0,$A0[1]
+ mov $A0[0],-8($tptr,$i) # t[3]
+
+ lea ($i),$j
+ jmp .Lsqr4x_inner
+
+.align 32
+.Lsqr4x_inner:
+ mov ($aptr,$j),$ai # a[4]
+ mul $a1 # a[3]*a[1]
+ add %rax,$A1[1] # a[3]*a[1]+t[4]
+ mov $ai,%rax
+ mov %rdx,$A1[0]
+ adc \$0,$A1[0]
+ add ($tptr,$j),$A1[1]
+ adc \$0,$A1[0]
+
+ .byte 0x67
+ mul $a0 # a[4]*a[0]
+ add %rax,$A0[1] # a[4]*a[0]+a[3]*a[1]+t[4]
+ mov $ai,%rax # a[3]
+ mov 8($aptr,$j),$ai # a[5]
+ mov %rdx,$A0[0]
+ adc \$0,$A0[0]
+ add $A1[1],$A0[1]
+ adc \$0,$A0[0]
+
+ mul $a1 # a[4]*a[3]
+ add %rax,$A1[0] # a[4]*a[3]+t[5]
+ mov $A0[1],($tptr,$j) # t[4]
+ mov $ai,%rax
+ mov %rdx,$A1[1]
+ adc \$0,$A1[1]
+ add 8($tptr,$j),$A1[0]
+ lea 16($j),$j # j++
+ adc \$0,$A1[1]
+
+ mul $a0 # a[5]*a[2]
+ add %rax,$A0[0] # a[5]*a[2]+a[4]*a[3]+t[5]
+ mov $ai,%rax
+ adc \$0,%rdx
+ add $A1[0],$A0[0]
+ mov %rdx,$A0[1]
+ adc \$0,$A0[1]
+ mov $A0[0],-8($tptr,$j) # t[5], "preloaded t[1]" below
+
+ cmp \$0,$j
+ jne .Lsqr4x_inner
+
+ .byte 0x67
+ mul $a1 # a[5]*a[3]
+ add %rax,$A1[1]
+ adc \$0,%rdx
+ add $A0[1],$A1[1]
+ adc \$0,%rdx
+
+ mov $A1[1],($tptr) # t[6], "preloaded t[2]" below
+ mov %rdx,$A1[0]
+ mov %rdx,8($tptr) # t[7], "preloaded t[3]" below
+
+ add \$16,$i
+ jnz .Lsqr4x_outer
+
+ # comments apply to $num==4 case
+ mov -32($aptr),$a0 # a[0]
+ lea 48+8(%rsp,$num,2),$tptr # end of tp[] buffer, &tp[2*$num]
+ mov -24($aptr),%rax # a[1]
+ lea -32($tptr,$i),$tptr # end of tp[] window, &tp[2*$num-"$i"]
+ mov -16($aptr),$ai # a[2]
+ mov %rax,$a1
+
+ mul $a0 # a[1]*a[0]
+ add %rax,$A0[0] # a[1]*a[0]+t[1], preloaded t[1]
+ mov $ai,%rax # a[2]
+ mov %rdx,$A0[1]
+ adc \$0,$A0[1]
+
+ mul $a0 # a[2]*a[0]
+ add %rax,$A0[1]
+ mov $ai,%rax
+ mov $A0[0],-24($tptr) # t[1]
+ mov %rdx,$A0[0]
+ adc \$0,$A0[0]
+ add $A1[1],$A0[1] # a[2]*a[0]+t[2], preloaded t[2]
+ mov -8($aptr),$ai # a[3]
+ adc \$0,$A0[0]
+
+ mul $a1 # a[2]*a[1]
+ add %rax,$A1[0] # a[2]*a[1]+t[3], preloaded t[3]
+ mov $ai,%rax
+ mov $A0[1],-16($tptr) # t[2]
+ mov %rdx,$A1[1]
+ adc \$0,$A1[1]
+
+ mul $a0 # a[3]*a[0]
+ add %rax,$A0[0] # a[3]*a[0]+a[2]*a[1]+t[3]
+ mov $ai,%rax
+ mov %rdx,$A0[1]
+ adc \$0,$A0[1]
+ add $A1[0],$A0[0]
+ adc \$0,$A0[1]
+ mov $A0[0],-8($tptr) # t[3]
+
+ mul $a1 # a[3]*a[1]
+ add %rax,$A1[1]
+ mov -16($aptr),%rax # a[2]
+ adc \$0,%rdx
+ add $A0[1],$A1[1]
+ adc \$0,%rdx
+
+ mov $A1[1],($tptr) # t[4]
+ mov %rdx,$A1[0]
+ mov %rdx,8($tptr) # t[5]
+
+ mul $ai # a[2]*a[3]
+___
+{
+my ($shift,$carry)=($a0,$a1);
+my @S=(@A1,$ai,$n0);
+$code.=<<___;
+ add \$16,$i
+ xor $shift,$shift
+ sub $num,$i # $i=16-$num
+ xor $carry,$carry
+
+ add $A1[0],%rax # t[5]
+ adc \$0,%rdx
+ mov %rax,8($tptr) # t[5]
+ mov %rdx,16($tptr) # t[6]
+ mov $carry,24($tptr) # t[7]
+
+ mov -16($aptr,$i),%rax # a[0]
+ lea 48+8(%rsp),$tptr
+ xor $A0[0],$A0[0] # t[0]
+ mov 8($tptr),$A0[1] # t[1]
+
+ lea ($shift,$A0[0],2),$S[0] # t[2*i]<<1 | shift
+ shr \$63,$A0[0]
+ lea ($j,$A0[1],2),$S[1] # t[2*i+1]<<1 |
+ shr \$63,$A0[1]
+ or $A0[0],$S[1] # | t[2*i]>>63
+ mov 16($tptr),$A0[0] # t[2*i+2] # prefetch
+ mov $A0[1],$shift # shift=t[2*i+1]>>63
+ mul %rax # a[i]*a[i]
+ neg $carry # mov $carry,cf
+ mov 24($tptr),$A0[1] # t[2*i+2+1] # prefetch
+ adc %rax,$S[0]
+ mov -8($aptr,$i),%rax # a[i+1] # prefetch
+ mov $S[0],($tptr)
+ adc %rdx,$S[1]
+
+ lea ($shift,$A0[0],2),$S[2] # t[2*i]<<1 | shift
+ mov $S[1],8($tptr)
+ sbb $carry,$carry # mov cf,$carry
+ shr \$63,$A0[0]
+ lea ($j,$A0[1],2),$S[3] # t[2*i+1]<<1 |
+ shr \$63,$A0[1]
+ or $A0[0],$S[3] # | t[2*i]>>63
+ mov 32($tptr),$A0[0] # t[2*i+2] # prefetch
+ mov $A0[1],$shift # shift=t[2*i+1]>>63
+ mul %rax # a[i]*a[i]
+ neg $carry # mov $carry,cf
+ mov 40($tptr),$A0[1] # t[2*i+2+1] # prefetch
+ adc %rax,$S[2]
+ mov 0($aptr,$i),%rax # a[i+1] # prefetch
+ mov $S[2],16($tptr)
+ adc %rdx,$S[3]
+ lea 16($i),$i
+ mov $S[3],24($tptr)
+ sbb $carry,$carry # mov cf,$carry
+ lea 64($tptr),$tptr
+ jmp .Lsqr4x_shift_n_add
+
+.align 32
+.Lsqr4x_shift_n_add:
+ lea ($shift,$A0[0],2),$S[0] # t[2*i]<<1 | shift
+ shr \$63,$A0[0]
+ lea ($j,$A0[1],2),$S[1] # t[2*i+1]<<1 |
+ shr \$63,$A0[1]
+ or $A0[0],$S[1] # | t[2*i]>>63
+ mov -16($tptr),$A0[0] # t[2*i+2] # prefetch
+ mov $A0[1],$shift # shift=t[2*i+1]>>63
+ mul %rax # a[i]*a[i]
+ neg $carry # mov $carry,cf
+ mov -8($tptr),$A0[1] # t[2*i+2+1] # prefetch
+ adc %rax,$S[0]
+ mov -8($aptr,$i),%rax # a[i+1] # prefetch
+ mov $S[0],-32($tptr)
+ adc %rdx,$S[1]
+
+ lea ($shift,$A0[0],2),$S[2] # t[2*i]<<1 | shift
+ mov $S[1],-24($tptr)
+ sbb $carry,$carry # mov cf,$carry
+ shr \$63,$A0[0]
+ lea ($j,$A0[1],2),$S[3] # t[2*i+1]<<1 |
+ shr \$63,$A0[1]
+ or $A0[0],$S[3] # | t[2*i]>>63
+ mov 0($tptr),$A0[0] # t[2*i+2] # prefetch
+ mov $A0[1],$shift # shift=t[2*i+1]>>63
+ mul %rax # a[i]*a[i]
+ neg $carry # mov $carry,cf
+ mov 8($tptr),$A0[1] # t[2*i+2+1] # prefetch
+ adc %rax,$S[2]
+ mov 0($aptr,$i),%rax # a[i+1] # prefetch
+ mov $S[2],-16($tptr)
+ adc %rdx,$S[3]
+
+ lea ($shift,$A0[0],2),$S[0] # t[2*i]<<1 | shift
+ mov $S[3],-8($tptr)
+ sbb $carry,$carry # mov cf,$carry
+ shr \$63,$A0[0]
+ lea ($j,$A0[1],2),$S[1] # t[2*i+1]<<1 |
+ shr \$63,$A0[1]
+ or $A0[0],$S[1] # | t[2*i]>>63
+ mov 16($tptr),$A0[0] # t[2*i+2] # prefetch
+ mov $A0[1],$shift # shift=t[2*i+1]>>63
+ mul %rax # a[i]*a[i]
+ neg $carry # mov $carry,cf
+ mov 24($tptr),$A0[1] # t[2*i+2+1] # prefetch
+ adc %rax,$S[0]
+ mov 8($aptr,$i),%rax # a[i+1] # prefetch
+ mov $S[0],0($tptr)
+ adc %rdx,$S[1]
+
+ lea ($shift,$A0[0],2),$S[2] # t[2*i]<<1 | shift
+ mov $S[1],8($tptr)
+ sbb $carry,$carry # mov cf,$carry
+ shr \$63,$A0[0]
+ lea ($j,$A0[1],2),$S[3] # t[2*i+1]<<1 |
+ shr \$63,$A0[1]
+ or $A0[0],$S[3] # | t[2*i]>>63
+ mov 32($tptr),$A0[0] # t[2*i+2] # prefetch
+ mov $A0[1],$shift # shift=t[2*i+1]>>63
+ mul %rax # a[i]*a[i]
+ neg $carry # mov $carry,cf
+ mov 40($tptr),$A0[1] # t[2*i+2+1] # prefetch
+ adc %rax,$S[2]
+ mov 16($aptr,$i),%rax # a[i+1] # prefetch
+ mov $S[2],16($tptr)
+ adc %rdx,$S[3]
+ mov $S[3],24($tptr)
+ sbb $carry,$carry # mov cf,$carry
+ lea 64($tptr),$tptr
+ add \$32,$i
+ jnz .Lsqr4x_shift_n_add
+
+ lea ($shift,$A0[0],2),$S[0] # t[2*i]<<1 | shift
+ .byte 0x67
+ shr \$63,$A0[0]
+ lea ($j,$A0[1],2),$S[1] # t[2*i+1]<<1 |
+ shr \$63,$A0[1]
+ or $A0[0],$S[1] # | t[2*i]>>63
+ mov -16($tptr),$A0[0] # t[2*i+2] # prefetch
+ mov $A0[1],$shift # shift=t[2*i+1]>>63
+ mul %rax # a[i]*a[i]
+ neg $carry # mov $carry,cf
+ mov -8($tptr),$A0[1] # t[2*i+2+1] # prefetch
+ adc %rax,$S[0]
+ mov -8($aptr),%rax # a[i+1] # prefetch
+ mov $S[0],-32($tptr)
+ adc %rdx,$S[1]
+
+ lea ($shift,$A0[0],2),$S[2] # t[2*i]<<1|shift
+ mov $S[1],-24($tptr)
+ sbb $carry,$carry # mov cf,$carry
+ shr \$63,$A0[0]
+ lea ($j,$A0[1],2),$S[3] # t[2*i+1]<<1 |
+ shr \$63,$A0[1]
+ or $A0[0],$S[3] # | t[2*i]>>63
+ mul %rax # a[i]*a[i]
+ neg $carry # mov $carry,cf
+ adc %rax,$S[2]
+ adc %rdx,$S[3]
+ mov $S[2],-16($tptr)
+ mov $S[3],-8($tptr)
+___
+}
+######################################################################
+# Montgomery reduction part, "word-by-word" algorithm.
+#
+# This new path is inspired by multiple submissions from Intel, by
+# Shay Gueron, Vlad Krasnov, Erdinc Ozturk, James Guilford,
+# Vinodh Gopal...
+{
+my ($nptr,$tptr,$carry,$m0)=("%rbp","%rdi","%rsi","%rbx");
+
+$code.=<<___;
+ movq %xmm2,$nptr
+__bn_sqr8x_reduction:
+ xor %rax,%rax
+ lea ($nptr,$num),%rcx # end of n[]
+ lea 48+8(%rsp,$num,2),%rdx # end of t[] buffer
+ mov %rcx,0+8(%rsp)
+ lea 48+8(%rsp,$num),$tptr # end of initial t[] window
+ mov %rdx,8+8(%rsp)
+ neg $num
+ jmp .L8x_reduction_loop
+
+.align 32
+.L8x_reduction_loop:
+ lea ($tptr,$num),$tptr # start of current t[] window
+ .byte 0x66
+ mov 8*0($tptr),$m0
+ mov 8*1($tptr),%r9
+ mov 8*2($tptr),%r10
+ mov 8*3($tptr),%r11
+ mov 8*4($tptr),%r12
+ mov 8*5($tptr),%r13
+ mov 8*6($tptr),%r14
+ mov 8*7($tptr),%r15
+ mov %rax,(%rdx) # store top-most carry bit
+ lea 8*8($tptr),$tptr
+
+ .byte 0x67
+ mov $m0,%r8
+ imulq 32+8(%rsp),$m0 # n0*a[0]
+ mov 8*0($nptr),%rax # n[0]
+ mov \$8,%ecx
+ jmp .L8x_reduce
+
+.align 32
+.L8x_reduce:
+ mulq $m0
+ mov 8*1($nptr),%rax # n[1]
+ neg %r8
+ mov %rdx,%r8
+ adc \$0,%r8
+
+ mulq $m0
+ add %rax,%r9
+ mov 8*2($nptr),%rax
+ adc \$0,%rdx
+ add %r9,%r8
+ mov $m0,48-8+8(%rsp,%rcx,8) # put aside n0*a[i]
+ mov %rdx,%r9
+ adc \$0,%r9
+
+ mulq $m0
+ add %rax,%r10
+ mov 8*3($nptr),%rax
+ adc \$0,%rdx
+ add %r10,%r9
+ mov 32+8(%rsp),$carry # pull n0, borrow $carry
+ mov %rdx,%r10
+ adc \$0,%r10
+
+ mulq $m0
+ add %rax,%r11
+ mov 8*4($nptr),%rax
+ adc \$0,%rdx
+ imulq %r8,$carry # modulo-scheduled
+ add %r11,%r10
+ mov %rdx,%r11
+ adc \$0,%r11
+
+ mulq $m0
+ add %rax,%r12
+ mov 8*5($nptr),%rax
+ adc \$0,%rdx
+ add %r12,%r11
+ mov %rdx,%r12
+ adc \$0,%r12
+
+ mulq $m0
+ add %rax,%r13
+ mov 8*6($nptr),%rax
+ adc \$0,%rdx
+ add %r13,%r12
+ mov %rdx,%r13
+ adc \$0,%r13
+
+ mulq $m0
+ add %rax,%r14
+ mov 8*7($nptr),%rax
+ adc \$0,%rdx
+ add %r14,%r13
+ mov %rdx,%r14
+ adc \$0,%r14
+
+ mulq $m0
+ mov $carry,$m0 # n0*a[i]
+ add %rax,%r15
+ mov 8*0($nptr),%rax # n[0]
+ adc \$0,%rdx
+ add %r15,%r14
+ mov %rdx,%r15
+ adc \$0,%r15
+
+ dec %ecx
+ jnz .L8x_reduce
+
+ lea 8*8($nptr),$nptr
+ xor %rax,%rax
+ mov 8+8(%rsp),%rdx # pull end of t[]
+ cmp 0+8(%rsp),$nptr # end of n[]?
+ jae .L8x_no_tail
+
+ .byte 0x66
+ add 8*0($tptr),%r8
+ adc 8*1($tptr),%r9
+ adc 8*2($tptr),%r10
+ adc 8*3($tptr),%r11
+ adc 8*4($tptr),%r12
+ adc 8*5($tptr),%r13
+ adc 8*6($tptr),%r14
+ adc 8*7($tptr),%r15
+ sbb $carry,$carry # top carry
+
+ mov 48+56+8(%rsp),$m0 # pull n0*a[0]
+ mov \$8,%ecx
+ mov 8*0($nptr),%rax
+ jmp .L8x_tail
+
+.align 32
+.L8x_tail:
+ mulq $m0
+ add %rax,%r8
+ mov 8*1($nptr),%rax
+ mov %r8,($tptr) # save result
+ mov %rdx,%r8
+ adc \$0,%r8
+
+ mulq $m0
+ add %rax,%r9
+ mov 8*2($nptr),%rax
+ adc \$0,%rdx
+ add %r9,%r8
+ lea 8($tptr),$tptr # $tptr++
+ mov %rdx,%r9
+ adc \$0,%r9
+
+ mulq $m0
+ add %rax,%r10
+ mov 8*3($nptr),%rax
+ adc \$0,%rdx
+ add %r10,%r9
+ mov %rdx,%r10
+ adc \$0,%r10
+
+ mulq $m0
+ add %rax,%r11
+ mov 8*4($nptr),%rax
+ adc \$0,%rdx
+ add %r11,%r10
+ mov %rdx,%r11
+ adc \$0,%r11
+
+ mulq $m0
+ add %rax,%r12
+ mov 8*5($nptr),%rax
+ adc \$0,%rdx
+ add %r12,%r11
+ mov %rdx,%r12
+ adc \$0,%r12
+
+ mulq $m0
+ add %rax,%r13
+ mov 8*6($nptr),%rax
+ adc \$0,%rdx
+ add %r13,%r12
+ mov %rdx,%r13
+ adc \$0,%r13
+
+ mulq $m0
+ add %rax,%r14
+ mov 8*7($nptr),%rax
+ adc \$0,%rdx
+ add %r14,%r13
+ mov %rdx,%r14
+ adc \$0,%r14
+
+ mulq $m0
+ mov 48-16+8(%rsp,%rcx,8),$m0# pull n0*a[i]
+ add %rax,%r15
+ adc \$0,%rdx
+ add %r15,%r14
+ mov 8*0($nptr),%rax # pull n[0]
+ mov %rdx,%r15
+ adc \$0,%r15
+
+ dec %ecx
+ jnz .L8x_tail
+
+ lea 8*8($nptr),$nptr
+ mov 8+8(%rsp),%rdx # pull end of t[]
+ cmp 0+8(%rsp),$nptr # end of n[]?
+ jae .L8x_tail_done # break out of loop
+
+ mov 48+56+8(%rsp),$m0 # pull n0*a[0]
+ neg $carry
+ mov 8*0($nptr),%rax # pull n[0]
+ adc 8*0($tptr),%r8
+ adc 8*1($tptr),%r9
+ adc 8*2($tptr),%r10
+ adc 8*3($tptr),%r11
+ adc 8*4($tptr),%r12
+ adc 8*5($tptr),%r13
+ adc 8*6($tptr),%r14
+ adc 8*7($tptr),%r15
+ sbb $carry,$carry # top carry
+
+ mov \$8,%ecx
+ jmp .L8x_tail
+
+.align 32
+.L8x_tail_done:
+ xor %rax,%rax
+ add (%rdx),%r8 # can this overflow?
+ adc \$0,%r9
+ adc \$0,%r10
+ adc \$0,%r11
+ adc \$0,%r12
+ adc \$0,%r13
+ adc \$0,%r14
+ adc \$0,%r15
+ adc \$0,%rax
+
+ neg $carry
+.L8x_no_tail:
+ adc 8*0($tptr),%r8
+ adc 8*1($tptr),%r9
+ adc 8*2($tptr),%r10
+ adc 8*3($tptr),%r11
+ adc 8*4($tptr),%r12
+ adc 8*5($tptr),%r13
+ adc 8*6($tptr),%r14
+ adc 8*7($tptr),%r15
+ adc \$0,%rax # top-most carry
+ mov -8($nptr),%rcx # np[num-1]
+ xor $carry,$carry
+
+ movq %xmm2,$nptr # restore $nptr
+
+ mov %r8,8*0($tptr) # store top 512 bits
+ mov %r9,8*1($tptr)
+ movq %xmm3,$num # $num is %r9, can't be moved upwards
+ mov %r10,8*2($tptr)
+ mov %r11,8*3($tptr)
+ mov %r12,8*4($tptr)
+ mov %r13,8*5($tptr)
+ mov %r14,8*6($tptr)
+ mov %r15,8*7($tptr)
+ lea 8*8($tptr),$tptr
+
+ cmp %rdx,$tptr # end of t[]?
+ jb .L8x_reduction_loop
+ ret
+.cfi_endproc
+.size bn_sqr8x_internal,.-bn_sqr8x_internal
+___
+}
+##############################################################
+# Post-condition, 4x unrolled
+#
+{
+my ($tptr,$nptr)=("%rbx","%rbp");
+$code.=<<___;
+.type __bn_post4x_internal,\@abi-omnipotent
+.align 32
+__bn_post4x_internal:
+.cfi_startproc
+ mov 8*0($nptr),%r12
+ lea (%rdi,$num),$tptr # %rdi was $tptr above
+ mov $num,%rcx
+ movq %xmm1,$rptr # restore $rptr
+ neg %rax
+ movq %xmm1,$aptr # prepare for back-to-back call
+ sar \$3+2,%rcx
+ dec %r12 # so that after 'not' we get -n[0]
+ xor %r10,%r10
+ mov 8*1($nptr),%r13
+ mov 8*2($nptr),%r14
+ mov 8*3($nptr),%r15
+ jmp .Lsqr4x_sub_entry
+
+.align 16
+.Lsqr4x_sub:
+ mov 8*0($nptr),%r12
+ mov 8*1($nptr),%r13
+ mov 8*2($nptr),%r14
+ mov 8*3($nptr),%r15
+.Lsqr4x_sub_entry:
+ lea 8*4($nptr),$nptr
+ not %r12
+ not %r13
+ not %r14
+ not %r15
+ and %rax,%r12
+ and %rax,%r13
+ and %rax,%r14
+ and %rax,%r15
+
+ neg %r10 # mov %r10,%cf
+ adc 8*0($tptr),%r12
+ adc 8*1($tptr),%r13
+ adc 8*2($tptr),%r14
+ adc 8*3($tptr),%r15
+ mov %r12,8*0($rptr)
+ lea 8*4($tptr),$tptr
+ mov %r13,8*1($rptr)
+ sbb %r10,%r10 # mov %cf,%r10
+ mov %r14,8*2($rptr)
+ mov %r15,8*3($rptr)
+ lea 8*4($rptr),$rptr
+
+ inc %rcx # pass %cf
+ jnz .Lsqr4x_sub
+
+ mov $num,%r10 # prepare for back-to-back call
+ neg $num # restore $num
+ ret
+.cfi_endproc
+.size __bn_post4x_internal,.-__bn_post4x_internal
+___
+}
+}}}
+
+if ($addx) {{{
+my $bp="%rdx"; # restore original value
+
+$code.=<<___;
+.type bn_mulx4x_mont_gather5,\@function,6
+.align 32
+bn_mulx4x_mont_gather5:
+.cfi_startproc
+ mov %rsp,%rax
+.cfi_def_cfa_register %rax
+.Lmulx4x_enter:
+ push %rbx
+.cfi_push %rbx
+ push %rbp
+.cfi_push %rbp
+ push %r12
+.cfi_push %r12
+ push %r13
+.cfi_push %r13
+ push %r14
+.cfi_push %r14
+ push %r15
+.cfi_push %r15
+.Lmulx4x_prologue:
+
+ shl \$3,${num}d # convert $num to bytes
+ lea ($num,$num,2),%r10 # 3*$num in bytes
+ neg $num # -$num
+ mov ($n0),$n0 # *n0
+
+ ##############################################################
+ # Ensure that stack frame doesn't alias with $rptr+3*$num
+ # modulo 4096, which covers ret[num], am[num] and n[num]
+ # (see bn_exp.c). This is done to allow memory disambiguation
+ # logic do its magic. [Extra [num] is allocated in order
+ # to align with bn_power5's frame, which is cleansed after
+ # completing exponentiation. Extra 256 bytes is for power mask
+ # calculated from 7th argument, the index.]
+ #
+ lea -320(%rsp,$num,2),%r11
+ mov %rsp,%rbp
+ sub $rp,%r11
+ and \$4095,%r11
+ cmp %r11,%r10
+ jb .Lmulx4xsp_alt
+ sub %r11,%rbp # align with $aptr
+ lea -320(%rbp,$num,2),%rbp # future alloca(frame+2*$num*8+256)
+ jmp .Lmulx4xsp_done
+
+.Lmulx4xsp_alt:
+ lea 4096-320(,$num,2),%r10
+ lea -320(%rbp,$num,2),%rbp # future alloca(frame+2*$num*8+256)
+ sub %r10,%r11
+ mov \$0,%r10
+ cmovc %r10,%r11
+ sub %r11,%rbp
+.Lmulx4xsp_done:
+ and \$-64,%rbp # ensure alignment
+ mov %rsp,%r11
+ sub %rbp,%r11
+ and \$-4096,%r11
+ lea (%rbp,%r11),%rsp
+ mov (%rsp),%r10
+ cmp %rbp,%rsp
+ ja .Lmulx4x_page_walk
+ jmp .Lmulx4x_page_walk_done
+
+.Lmulx4x_page_walk:
+ lea -4096(%rsp),%rsp
+ mov (%rsp),%r10
+ cmp %rbp,%rsp
+ ja .Lmulx4x_page_walk
+.Lmulx4x_page_walk_done:
+
+ ##############################################################
+ # Stack layout
+ # +0 -num
+ # +8 off-loaded &b[i]
+ # +16 end of b[num]
+ # +24 inner counter
+ # +32 saved n0
+ # +40 saved %rsp
+ # +48
+ # +56 saved rp
+ # +64 tmp[num+1]
+ #
+ mov $n0, 32(%rsp) # save *n0
+ mov %rax,40(%rsp) # save original %rsp
+.cfi_cfa_expression %rsp+40,deref,+8
+.Lmulx4x_body:
+ call mulx4x_internal
+
+ mov 40(%rsp),%rsi # restore %rsp
+.cfi_def_cfa %rsi,8
+ mov \$1,%rax
+
+ mov -48(%rsi),%r15
+.cfi_restore %r15
+ mov -40(%rsi),%r14
+.cfi_restore %r14
+ mov -32(%rsi),%r13
+.cfi_restore %r13
+ mov -24(%rsi),%r12
+.cfi_restore %r12
+ mov -16(%rsi),%rbp
+.cfi_restore %rbp
+ mov -8(%rsi),%rbx
+.cfi_restore %rbx
+ lea (%rsi),%rsp
+.cfi_def_cfa_register %rsp
+.Lmulx4x_epilogue:
+ ret
+.cfi_endproc
+.size bn_mulx4x_mont_gather5,.-bn_mulx4x_mont_gather5
+
+.type mulx4x_internal,\@abi-omnipotent
+.align 32
+mulx4x_internal:
+.cfi_startproc
+ mov $num,8(%rsp) # save -$num (it was in bytes)
+ mov $num,%r10
+ neg $num # restore $num
+ shl \$5,$num
+ neg %r10 # restore $num
+ lea 128($bp,$num),%r13 # end of powers table (+size optimization)
+ shr \$5+5,$num
+ movd `($win64?56:8)`(%rax),%xmm5 # load 7th argument
+ sub \$1,$num
+ lea .Linc(%rip),%rax
+ mov %r13,16+8(%rsp) # end of b[num]
+ mov $num,24+8(%rsp) # inner counter
+ mov $rp, 56+8(%rsp) # save $rp
+___
+my ($aptr, $bptr, $nptr, $tptr, $mi, $bi, $zero, $num)=
+ ("%rsi","%rdi","%rcx","%rbx","%r8","%r9","%rbp","%rax");
+my $rptr=$bptr;
+my $STRIDE=2**5*8; # 5 is "window size"
+my $N=$STRIDE/4; # should match cache line size
+$code.=<<___;
+ movdqa 0(%rax),%xmm0 # 00000001000000010000000000000000
+ movdqa 16(%rax),%xmm1 # 00000002000000020000000200000002
+ lea 88-112(%rsp,%r10),%r10 # place the mask after tp[num+1] (+ICache optimization)
+ lea 128($bp),$bptr # size optimization
+
+ pshufd \$0,%xmm5,%xmm5 # broadcast index
+ movdqa %xmm1,%xmm4
+ .byte 0x67
+ movdqa %xmm1,%xmm2
+___
+########################################################################
+# calculate mask by comparing 0..31 to index and save result to stack
+#
+$code.=<<___;
+ .byte 0x67
+ paddd %xmm0,%xmm1
+ pcmpeqd %xmm5,%xmm0 # compare to 1,0
+ movdqa %xmm4,%xmm3
+___
+for($i=0;$i<$STRIDE/16-4;$i+=4) {
+$code.=<<___;
+ paddd %xmm1,%xmm2
+ pcmpeqd %xmm5,%xmm1 # compare to 3,2
+ movdqa %xmm0,`16*($i+0)+112`(%r10)
+ movdqa %xmm4,%xmm0
+
+ paddd %xmm2,%xmm3
+ pcmpeqd %xmm5,%xmm2 # compare to 5,4
+ movdqa %xmm1,`16*($i+1)+112`(%r10)
+ movdqa %xmm4,%xmm1
+
+ paddd %xmm3,%xmm0
+ pcmpeqd %xmm5,%xmm3 # compare to 7,6
+ movdqa %xmm2,`16*($i+2)+112`(%r10)
+ movdqa %xmm4,%xmm2
+
+ paddd %xmm0,%xmm1
+ pcmpeqd %xmm5,%xmm0
+ movdqa %xmm3,`16*($i+3)+112`(%r10)
+ movdqa %xmm4,%xmm3
+___
+}
+$code.=<<___; # last iteration can be optimized
+ .byte 0x67
+ paddd %xmm1,%xmm2
+ pcmpeqd %xmm5,%xmm1
+ movdqa %xmm0,`16*($i+0)+112`(%r10)
+
+ paddd %xmm2,%xmm3
+ pcmpeqd %xmm5,%xmm2
+ movdqa %xmm1,`16*($i+1)+112`(%r10)
+
+ pcmpeqd %xmm5,%xmm3
+ movdqa %xmm2,`16*($i+2)+112`(%r10)
+
+ pand `16*($i+0)-128`($bptr),%xmm0 # while it's still in register
+ pand `16*($i+1)-128`($bptr),%xmm1
+ pand `16*($i+2)-128`($bptr),%xmm2
+ movdqa %xmm3,`16*($i+3)+112`(%r10)
+ pand `16*($i+3)-128`($bptr),%xmm3
+ por %xmm2,%xmm0
+ por %xmm3,%xmm1
+___
+for($i=0;$i<$STRIDE/16-4;$i+=4) {
+$code.=<<___;
+ movdqa `16*($i+0)-128`($bptr),%xmm4
+ movdqa `16*($i+1)-128`($bptr),%xmm5
+ movdqa `16*($i+2)-128`($bptr),%xmm2
+ pand `16*($i+0)+112`(%r10),%xmm4
+ movdqa `16*($i+3)-128`($bptr),%xmm3
+ pand `16*($i+1)+112`(%r10),%xmm5
+ por %xmm4,%xmm0
+ pand `16*($i+2)+112`(%r10),%xmm2
+ por %xmm5,%xmm1
+ pand `16*($i+3)+112`(%r10),%xmm3
+ por %xmm2,%xmm0
+ por %xmm3,%xmm1
+___
+}
+$code.=<<___;
+ pxor %xmm1,%xmm0
+ pshufd \$0x4e,%xmm0,%xmm1
+ por %xmm1,%xmm0
+ lea $STRIDE($bptr),$bptr
+ movq %xmm0,%rdx # bp[0]
+ lea 64+8*4+8(%rsp),$tptr
+
+ mov %rdx,$bi
+ mulx 0*8($aptr),$mi,%rax # a[0]*b[0]
+ mulx 1*8($aptr),%r11,%r12 # a[1]*b[0]
+ add %rax,%r11
+ mulx 2*8($aptr),%rax,%r13 # ...
+ adc %rax,%r12
+ adc \$0,%r13
+ mulx 3*8($aptr),%rax,%r14
+
+ mov $mi,%r15
+ imulq 32+8(%rsp),$mi # "t[0]"*n0
+ xor $zero,$zero # cf=0, of=0
+ mov $mi,%rdx
+
+ mov $bptr,8+8(%rsp) # off-load &b[i]
+
+ lea 4*8($aptr),$aptr
+ adcx %rax,%r13
+ adcx $zero,%r14 # cf=0
+
+ mulx 0*8($nptr),%rax,%r10
+ adcx %rax,%r15 # discarded
+ adox %r11,%r10
+ mulx 1*8($nptr),%rax,%r11
+ adcx %rax,%r10
+ adox %r12,%r11
+ mulx 2*8($nptr),%rax,%r12
+ mov 24+8(%rsp),$bptr # counter value
+ mov %r10,-8*4($tptr)
+ adcx %rax,%r11
+ adox %r13,%r12
+ mulx 3*8($nptr),%rax,%r15
+ mov $bi,%rdx
+ mov %r11,-8*3($tptr)
+ adcx %rax,%r12
+ adox $zero,%r15 # of=0
+ lea 4*8($nptr),$nptr
+ mov %r12,-8*2($tptr)
+ jmp .Lmulx4x_1st
+
+.align 32
+.Lmulx4x_1st:
+ adcx $zero,%r15 # cf=0, modulo-scheduled
+ mulx 0*8($aptr),%r10,%rax # a[4]*b[0]
+ adcx %r14,%r10
+ mulx 1*8($aptr),%r11,%r14 # a[5]*b[0]
+ adcx %rax,%r11
+ mulx 2*8($aptr),%r12,%rax # ...
+ adcx %r14,%r12
+ mulx 3*8($aptr),%r13,%r14
+ .byte 0x67,0x67
+ mov $mi,%rdx
+ adcx %rax,%r13
+ adcx $zero,%r14 # cf=0
+ lea 4*8($aptr),$aptr
+ lea 4*8($tptr),$tptr
+
+ adox %r15,%r10
+ mulx 0*8($nptr),%rax,%r15
+ adcx %rax,%r10
+ adox %r15,%r11
+ mulx 1*8($nptr),%rax,%r15
+ adcx %rax,%r11
+ adox %r15,%r12
+ mulx 2*8($nptr),%rax,%r15
+ mov %r10,-5*8($tptr)
+ adcx %rax,%r12
+ mov %r11,-4*8($tptr)
+ adox %r15,%r13
+ mulx 3*8($nptr),%rax,%r15
+ mov $bi,%rdx
+ mov %r12,-3*8($tptr)
+ adcx %rax,%r13
+ adox $zero,%r15
+ lea 4*8($nptr),$nptr
+ mov %r13,-2*8($tptr)
+
+ dec $bptr # of=0, pass cf
+ jnz .Lmulx4x_1st
+
+ mov 8(%rsp),$num # load -num
+ adc $zero,%r15 # modulo-scheduled
+ lea ($aptr,$num),$aptr # rewind $aptr
+ add %r15,%r14
+ mov 8+8(%rsp),$bptr # re-load &b[i]
+ adc $zero,$zero # top-most carry
+ mov %r14,-1*8($tptr)
+ jmp .Lmulx4x_outer
+
+.align 32
+.Lmulx4x_outer:
+ lea 16-256($tptr),%r10 # where 256-byte mask is (+density control)
+ pxor %xmm4,%xmm4
+ .byte 0x67,0x67
+ pxor %xmm5,%xmm5
+___
+for($i=0;$i<$STRIDE/16;$i+=4) {
+$code.=<<___;
+ movdqa `16*($i+0)-128`($bptr),%xmm0
+ movdqa `16*($i+1)-128`($bptr),%xmm1
+ movdqa `16*($i+2)-128`($bptr),%xmm2
+ pand `16*($i+0)+256`(%r10),%xmm0
+ movdqa `16*($i+3)-128`($bptr),%xmm3
+ pand `16*($i+1)+256`(%r10),%xmm1
+ por %xmm0,%xmm4
+ pand `16*($i+2)+256`(%r10),%xmm2
+ por %xmm1,%xmm5
+ pand `16*($i+3)+256`(%r10),%xmm3
+ por %xmm2,%xmm4
+ por %xmm3,%xmm5
+___
+}
+$code.=<<___;
+ por %xmm5,%xmm4
+ pshufd \$0x4e,%xmm4,%xmm0
+ por %xmm4,%xmm0
+ lea $STRIDE($bptr),$bptr
+ movq %xmm0,%rdx # m0=bp[i]
+
+ mov $zero,($tptr) # save top-most carry
+ lea 4*8($tptr,$num),$tptr # rewind $tptr
+ mulx 0*8($aptr),$mi,%r11 # a[0]*b[i]
+ xor $zero,$zero # cf=0, of=0
+ mov %rdx,$bi
+ mulx 1*8($aptr),%r14,%r12 # a[1]*b[i]
+ adox -4*8($tptr),$mi # +t[0]
+ adcx %r14,%r11
+ mulx 2*8($aptr),%r15,%r13 # ...
+ adox -3*8($tptr),%r11
+ adcx %r15,%r12
+ mulx 3*8($aptr),%rdx,%r14
+ adox -2*8($tptr),%r12
+ adcx %rdx,%r13
+ lea ($nptr,$num),$nptr # rewind $nptr
+ lea 4*8($aptr),$aptr
+ adox -1*8($tptr),%r13
+ adcx $zero,%r14
+ adox $zero,%r14
+
+ mov $mi,%r15
+ imulq 32+8(%rsp),$mi # "t[0]"*n0
+
+ mov $mi,%rdx
+ xor $zero,$zero # cf=0, of=0
+ mov $bptr,8+8(%rsp) # off-load &b[i]
+
+ mulx 0*8($nptr),%rax,%r10
+ adcx %rax,%r15 # discarded
+ adox %r11,%r10
+ mulx 1*8($nptr),%rax,%r11
+ adcx %rax,%r10
+ adox %r12,%r11
+ mulx 2*8($nptr),%rax,%r12
+ adcx %rax,%r11
+ adox %r13,%r12
+ mulx 3*8($nptr),%rax,%r15
+ mov $bi,%rdx
+ mov 24+8(%rsp),$bptr # counter value
+ mov %r10,-8*4($tptr)
+ adcx %rax,%r12
+ mov %r11,-8*3($tptr)
+ adox $zero,%r15 # of=0
+ mov %r12,-8*2($tptr)
+ lea 4*8($nptr),$nptr
+ jmp .Lmulx4x_inner
+
+.align 32
+.Lmulx4x_inner:
+ mulx 0*8($aptr),%r10,%rax # a[4]*b[i]
+ adcx $zero,%r15 # cf=0, modulo-scheduled
+ adox %r14,%r10
+ mulx 1*8($aptr),%r11,%r14 # a[5]*b[i]
+ adcx 0*8($tptr),%r10
+ adox %rax,%r11
+ mulx 2*8($aptr),%r12,%rax # ...
+ adcx 1*8($tptr),%r11
+ adox %r14,%r12
+ mulx 3*8($aptr),%r13,%r14
+ mov $mi,%rdx
+ adcx 2*8($tptr),%r12
+ adox %rax,%r13
+ adcx 3*8($tptr),%r13
+ adox $zero,%r14 # of=0
+ lea 4*8($aptr),$aptr
+ lea 4*8($tptr),$tptr
+ adcx $zero,%r14 # cf=0
+
+ adox %r15,%r10
+ mulx 0*8($nptr),%rax,%r15
+ adcx %rax,%r10
+ adox %r15,%r11
+ mulx 1*8($nptr),%rax,%r15
+ adcx %rax,%r11
+ adox %r15,%r12
+ mulx 2*8($nptr),%rax,%r15
+ mov %r10,-5*8($tptr)
+ adcx %rax,%r12
+ adox %r15,%r13
+ mov %r11,-4*8($tptr)
+ mulx 3*8($nptr),%rax,%r15
+ mov $bi,%rdx
+ lea 4*8($nptr),$nptr
+ mov %r12,-3*8($tptr)
+ adcx %rax,%r13
+ adox $zero,%r15
+ mov %r13,-2*8($tptr)
+
+ dec $bptr # of=0, pass cf
+ jnz .Lmulx4x_inner
+
+ mov 0+8(%rsp),$num # load -num
+ adc $zero,%r15 # modulo-scheduled
+ sub 0*8($tptr),$bptr # pull top-most carry to %cf
+ mov 8+8(%rsp),$bptr # re-load &b[i]
+ mov 16+8(%rsp),%r10
+ adc %r15,%r14
+ lea ($aptr,$num),$aptr # rewind $aptr
+ adc $zero,$zero # top-most carry
+ mov %r14,-1*8($tptr)
+
+ cmp %r10,$bptr
+ jb .Lmulx4x_outer
+
+ mov -8($nptr),%r10
+ mov $zero,%r8
+ mov ($nptr,$num),%r12
+ lea ($nptr,$num),%rbp # rewind $nptr
+ mov $num,%rcx
+ lea ($tptr,$num),%rdi # rewind $tptr
+ xor %eax,%eax
+ xor %r15,%r15
+ sub %r14,%r10 # compare top-most words
+ adc %r15,%r15
+ or %r15,%r8
+ sar \$3+2,%rcx
+ sub %r8,%rax # %rax=-%r8
+ mov 56+8(%rsp),%rdx # restore rp
+ dec %r12 # so that after 'not' we get -n[0]
+ mov 8*1(%rbp),%r13
+ xor %r8,%r8
+ mov 8*2(%rbp),%r14
+ mov 8*3(%rbp),%r15
+ jmp .Lsqrx4x_sub_entry # common post-condition
+.cfi_endproc
+.size mulx4x_internal,.-mulx4x_internal
+___
+} {
+######################################################################
+# void bn_power5(
+my $rptr="%rdi"; # BN_ULONG *rptr,
+my $aptr="%rsi"; # const BN_ULONG *aptr,
+my $bptr="%rdx"; # const void *table,
+my $nptr="%rcx"; # const BN_ULONG *nptr,
+my $n0 ="%r8"; # const BN_ULONG *n0);
+my $num ="%r9"; # int num, has to be divisible by 8
+ # int pwr);
+
+my ($i,$j,$tptr)=("%rbp","%rcx",$rptr);
+my @A0=("%r10","%r11");
+my @A1=("%r12","%r13");
+my ($a0,$a1,$ai)=("%r14","%r15","%rbx");
+
+$code.=<<___;
+.type bn_powerx5,\@function,6
+.align 32
+bn_powerx5:
+.cfi_startproc
+ mov %rsp,%rax
+.cfi_def_cfa_register %rax
+.Lpowerx5_enter:
+ push %rbx
+.cfi_push %rbx
+ push %rbp
+.cfi_push %rbp
+ push %r12
+.cfi_push %r12
+ push %r13
+.cfi_push %r13
+ push %r14
+.cfi_push %r14
+ push %r15
+.cfi_push %r15
+.Lpowerx5_prologue:
+
+ shl \$3,${num}d # convert $num to bytes
+ lea ($num,$num,2),%r10 # 3*$num in bytes
+ neg $num
+ mov ($n0),$n0 # *n0
+
+ ##############################################################
+ # Ensure that stack frame doesn't alias with $rptr+3*$num
+ # modulo 4096, which covers ret[num], am[num] and n[num]
+ # (see bn_exp.c). This is done to allow memory disambiguation
+ # logic do its magic. [Extra 256 bytes is for power mask
+ # calculated from 7th argument, the index.]
+ #
+ lea -320(%rsp,$num,2),%r11
+ mov %rsp,%rbp
+ sub $rptr,%r11
+ and \$4095,%r11
+ cmp %r11,%r10
+ jb .Lpwrx_sp_alt
+ sub %r11,%rbp # align with $aptr
+ lea -320(%rbp,$num,2),%rbp # future alloca(frame+2*$num*8+256)
+ jmp .Lpwrx_sp_done
+
+.align 32
+.Lpwrx_sp_alt:
+ lea 4096-320(,$num,2),%r10
+ lea -320(%rbp,$num,2),%rbp # alloca(frame+2*$num*8+256)
+ sub %r10,%r11
+ mov \$0,%r10
+ cmovc %r10,%r11
+ sub %r11,%rbp
+.Lpwrx_sp_done:
+ and \$-64,%rbp
+ mov %rsp,%r11
+ sub %rbp,%r11
+ and \$-4096,%r11
+ lea (%rbp,%r11),%rsp
+ mov (%rsp),%r10
+ cmp %rbp,%rsp
+ ja .Lpwrx_page_walk
+ jmp .Lpwrx_page_walk_done
+
+.Lpwrx_page_walk:
+ lea -4096(%rsp),%rsp
+ mov (%rsp),%r10
+ cmp %rbp,%rsp
+ ja .Lpwrx_page_walk
+.Lpwrx_page_walk_done:
+
+ mov $num,%r10
+ neg $num
+
+ ##############################################################
+ # Stack layout
+ #
+ # +0 saved $num, used in reduction section
+ # +8 &t[2*$num], used in reduction section
+ # +16 intermediate carry bit
+ # +24 top-most carry bit, used in reduction section
+ # +32 saved *n0
+ # +40 saved %rsp
+ # +48 t[2*$num]
+ #
+ pxor %xmm0,%xmm0
+ movq $rptr,%xmm1 # save $rptr
+ movq $nptr,%xmm2 # save $nptr
+ movq %r10, %xmm3 # -$num
+ movq $bptr,%xmm4
+ mov $n0, 32(%rsp)
+ mov %rax, 40(%rsp) # save original %rsp
+.cfi_cfa_expression %rsp+40,deref,+8
+.Lpowerx5_body:
+
+ call __bn_sqrx8x_internal
+ call __bn_postx4x_internal
+ call __bn_sqrx8x_internal
+ call __bn_postx4x_internal
+ call __bn_sqrx8x_internal
+ call __bn_postx4x_internal
+ call __bn_sqrx8x_internal
+ call __bn_postx4x_internal
+ call __bn_sqrx8x_internal
+ call __bn_postx4x_internal
+
+ mov %r10,$num # -num
+ mov $aptr,$rptr
+ movq %xmm2,$nptr
+ movq %xmm4,$bptr
+ mov 40(%rsp),%rax
+
+ call mulx4x_internal
+
+ mov 40(%rsp),%rsi # restore %rsp
+.cfi_def_cfa %rsi,8
+ mov \$1,%rax
+
+ mov -48(%rsi),%r15
+.cfi_restore %r15
+ mov -40(%rsi),%r14
+.cfi_restore %r14
+ mov -32(%rsi),%r13
+.cfi_restore %r13
+ mov -24(%rsi),%r12
+.cfi_restore %r12
+ mov -16(%rsi),%rbp
+.cfi_restore %rbp
+ mov -8(%rsi),%rbx
+.cfi_restore %rbx
+ lea (%rsi),%rsp
+.cfi_def_cfa_register %rsp
+.Lpowerx5_epilogue:
+ ret
+.cfi_endproc
+.size bn_powerx5,.-bn_powerx5
+
+.globl bn_sqrx8x_internal
+.hidden bn_sqrx8x_internal
+.type bn_sqrx8x_internal,\@abi-omnipotent
+.align 32
+bn_sqrx8x_internal:
+__bn_sqrx8x_internal:
+.cfi_startproc
+ ##################################################################
+ # Squaring part:
+ #
+ # a) multiply-n-add everything but a[i]*a[i];
+ # b) shift result of a) by 1 to the left and accumulate
+ # a[i]*a[i] products;
+ #
+ ##################################################################
+ # a[7]a[7]a[6]a[6]a[5]a[5]a[4]a[4]a[3]a[3]a[2]a[2]a[1]a[1]a[0]a[0]
+ # a[1]a[0]
+ # a[2]a[0]
+ # a[3]a[0]
+ # a[2]a[1]
+ # a[3]a[1]
+ # a[3]a[2]
+ #
+ # a[4]a[0]
+ # a[5]a[0]
+ # a[6]a[0]
+ # a[7]a[0]
+ # a[4]a[1]
+ # a[5]a[1]
+ # a[6]a[1]
+ # a[7]a[1]
+ # a[4]a[2]
+ # a[5]a[2]
+ # a[6]a[2]
+ # a[7]a[2]
+ # a[4]a[3]
+ # a[5]a[3]
+ # a[6]a[3]
+ # a[7]a[3]
+ #
+ # a[5]a[4]
+ # a[6]a[4]
+ # a[7]a[4]
+ # a[6]a[5]
+ # a[7]a[5]
+ # a[7]a[6]
+ # a[7]a[7]a[6]a[6]a[5]a[5]a[4]a[4]a[3]a[3]a[2]a[2]a[1]a[1]a[0]a[0]
+___
+{
+my ($zero,$carry)=("%rbp","%rcx");
+my $aaptr=$zero;
+$code.=<<___;
+ lea 48+8(%rsp),$tptr
+ lea ($aptr,$num),$aaptr
+ mov $num,0+8(%rsp) # save $num
+ mov $aaptr,8+8(%rsp) # save end of $aptr
+ jmp .Lsqr8x_zero_start
+
+.align 32
+.byte 0x66,0x66,0x66,0x2e,0x0f,0x1f,0x84,0x00,0x00,0x00,0x00,0x00
+.Lsqrx8x_zero:
+ .byte 0x3e
+ movdqa %xmm0,0*8($tptr)
+ movdqa %xmm0,2*8($tptr)
+ movdqa %xmm0,4*8($tptr)
+ movdqa %xmm0,6*8($tptr)
+.Lsqr8x_zero_start: # aligned at 32
+ movdqa %xmm0,8*8($tptr)
+ movdqa %xmm0,10*8($tptr)
+ movdqa %xmm0,12*8($tptr)
+ movdqa %xmm0,14*8($tptr)
+ lea 16*8($tptr),$tptr
+ sub \$64,$num
+ jnz .Lsqrx8x_zero
+
+ mov 0*8($aptr),%rdx # a[0], modulo-scheduled
+ #xor %r9,%r9 # t[1], ex-$num, zero already
+ xor %r10,%r10
+ xor %r11,%r11
+ xor %r12,%r12
+ xor %r13,%r13
+ xor %r14,%r14
+ xor %r15,%r15
+ lea 48+8(%rsp),$tptr
+ xor $zero,$zero # cf=0, cf=0
+ jmp .Lsqrx8x_outer_loop
+
+.align 32
+.Lsqrx8x_outer_loop:
+ mulx 1*8($aptr),%r8,%rax # a[1]*a[0]
+ adcx %r9,%r8 # a[1]*a[0]+=t[1]
+ adox %rax,%r10
+ mulx 2*8($aptr),%r9,%rax # a[2]*a[0]
+ adcx %r10,%r9
+ adox %rax,%r11
+ .byte 0xc4,0xe2,0xab,0xf6,0x86,0x18,0x00,0x00,0x00 # mulx 3*8($aptr),%r10,%rax # ...
+ adcx %r11,%r10
+ adox %rax,%r12
+ .byte 0xc4,0xe2,0xa3,0xf6,0x86,0x20,0x00,0x00,0x00 # mulx 4*8($aptr),%r11,%rax
+ adcx %r12,%r11
+ adox %rax,%r13
+ mulx 5*8($aptr),%r12,%rax
+ adcx %r13,%r12
+ adox %rax,%r14
+ mulx 6*8($aptr),%r13,%rax
+ adcx %r14,%r13
+ adox %r15,%rax
+ mulx 7*8($aptr),%r14,%r15
+ mov 1*8($aptr),%rdx # a[1]
+ adcx %rax,%r14
+ adox $zero,%r15
+ adc 8*8($tptr),%r15
+ mov %r8,1*8($tptr) # t[1]
+ mov %r9,2*8($tptr) # t[2]
+ sbb $carry,$carry # mov %cf,$carry
+ xor $zero,$zero # cf=0, of=0
+
+
+ mulx 2*8($aptr),%r8,%rbx # a[2]*a[1]
+ mulx 3*8($aptr),%r9,%rax # a[3]*a[1]
+ adcx %r10,%r8
+ adox %rbx,%r9
+ mulx 4*8($aptr),%r10,%rbx # ...
+ adcx %r11,%r9
+ adox %rax,%r10
+ .byte 0xc4,0xe2,0xa3,0xf6,0x86,0x28,0x00,0x00,0x00 # mulx 5*8($aptr),%r11,%rax
+ adcx %r12,%r10
+ adox %rbx,%r11
+ .byte 0xc4,0xe2,0x9b,0xf6,0x9e,0x30,0x00,0x00,0x00 # mulx 6*8($aptr),%r12,%rbx
+ adcx %r13,%r11
+ adox %r14,%r12
+ .byte 0xc4,0x62,0x93,0xf6,0xb6,0x38,0x00,0x00,0x00 # mulx 7*8($aptr),%r13,%r14
+ mov 2*8($aptr),%rdx # a[2]
+ adcx %rax,%r12
+ adox %rbx,%r13
+ adcx %r15,%r13
+ adox $zero,%r14 # of=0
+ adcx $zero,%r14 # cf=0
+
+ mov %r8,3*8($tptr) # t[3]
+ mov %r9,4*8($tptr) # t[4]
+
+ mulx 3*8($aptr),%r8,%rbx # a[3]*a[2]
+ mulx 4*8($aptr),%r9,%rax # a[4]*a[2]
+ adcx %r10,%r8
+ adox %rbx,%r9
+ mulx 5*8($aptr),%r10,%rbx # ...
+ adcx %r11,%r9
+ adox %rax,%r10
+ .byte 0xc4,0xe2,0xa3,0xf6,0x86,0x30,0x00,0x00,0x00 # mulx 6*8($aptr),%r11,%rax
+ adcx %r12,%r10
+ adox %r13,%r11
+ .byte 0xc4,0x62,0x9b,0xf6,0xae,0x38,0x00,0x00,0x00 # mulx 7*8($aptr),%r12,%r13
+ .byte 0x3e
+ mov 3*8($aptr),%rdx # a[3]
+ adcx %rbx,%r11
+ adox %rax,%r12
+ adcx %r14,%r12
+ mov %r8,5*8($tptr) # t[5]
+ mov %r9,6*8($tptr) # t[6]
+ mulx 4*8($aptr),%r8,%rax # a[4]*a[3]
+ adox $zero,%r13 # of=0
+ adcx $zero,%r13 # cf=0
+
+ mulx 5*8($aptr),%r9,%rbx # a[5]*a[3]
+ adcx %r10,%r8
+ adox %rax,%r9
+ mulx 6*8($aptr),%r10,%rax # ...
+ adcx %r11,%r9
+ adox %r12,%r10
+ mulx 7*8($aptr),%r11,%r12
+ mov 4*8($aptr),%rdx # a[4]
+ mov 5*8($aptr),%r14 # a[5]
+ adcx %rbx,%r10
+ adox %rax,%r11
+ mov 6*8($aptr),%r15 # a[6]
+ adcx %r13,%r11
+ adox $zero,%r12 # of=0
+ adcx $zero,%r12 # cf=0
+
+ mov %r8,7*8($tptr) # t[7]
+ mov %r9,8*8($tptr) # t[8]
+
+ mulx %r14,%r9,%rax # a[5]*a[4]
+ mov 7*8($aptr),%r8 # a[7]
+ adcx %r10,%r9
+ mulx %r15,%r10,%rbx # a[6]*a[4]
+ adox %rax,%r10
+ adcx %r11,%r10
+ mulx %r8,%r11,%rax # a[7]*a[4]
+ mov %r14,%rdx # a[5]
+ adox %rbx,%r11
+ adcx %r12,%r11
+ #adox $zero,%rax # of=0
+ adcx $zero,%rax # cf=0
+
+ mulx %r15,%r14,%rbx # a[6]*a[5]
+ mulx %r8,%r12,%r13 # a[7]*a[5]
+ mov %r15,%rdx # a[6]
+ lea 8*8($aptr),$aptr
+ adcx %r14,%r11
+ adox %rbx,%r12
+ adcx %rax,%r12
+ adox $zero,%r13
+
+ .byte 0x67,0x67
+ mulx %r8,%r8,%r14 # a[7]*a[6]
+ adcx %r8,%r13
+ adcx $zero,%r14
+
+ cmp 8+8(%rsp),$aptr
+ je .Lsqrx8x_outer_break
+
+ neg $carry # mov $carry,%cf
+ mov \$-8,%rcx
+ mov $zero,%r15
+ mov 8*8($tptr),%r8
+ adcx 9*8($tptr),%r9 # +=t[9]
+ adcx 10*8($tptr),%r10 # ...
+ adcx 11*8($tptr),%r11
+ adc 12*8($tptr),%r12
+ adc 13*8($tptr),%r13
+ adc 14*8($tptr),%r14
+ adc 15*8($tptr),%r15
+ lea ($aptr),$aaptr
+ lea 2*64($tptr),$tptr
+ sbb %rax,%rax # mov %cf,$carry
+
+ mov -64($aptr),%rdx # a[0]
+ mov %rax,16+8(%rsp) # offload $carry
+ mov $tptr,24+8(%rsp)
+
+ #lea 8*8($tptr),$tptr # see 2*8*8($tptr) above
+ xor %eax,%eax # cf=0, of=0
+ jmp .Lsqrx8x_loop
+
+.align 32
+.Lsqrx8x_loop:
+ mov %r8,%rbx
+ mulx 0*8($aaptr),%rax,%r8 # a[8]*a[i]
+ adcx %rax,%rbx # +=t[8]
+ adox %r9,%r8
+
+ mulx 1*8($aaptr),%rax,%r9 # ...
+ adcx %rax,%r8
+ adox %r10,%r9
+
+ mulx 2*8($aaptr),%rax,%r10
+ adcx %rax,%r9
+ adox %r11,%r10
+
+ mulx 3*8($aaptr),%rax,%r11
+ adcx %rax,%r10
+ adox %r12,%r11
+
+ .byte 0xc4,0x62,0xfb,0xf6,0xa5,0x20,0x00,0x00,0x00 # mulx 4*8($aaptr),%rax,%r12
+ adcx %rax,%r11
+ adox %r13,%r12
+
+ mulx 5*8($aaptr),%rax,%r13
+ adcx %rax,%r12
+ adox %r14,%r13
+
+ mulx 6*8($aaptr),%rax,%r14
+ mov %rbx,($tptr,%rcx,8) # store t[8+i]
+ mov \$0,%ebx
+ adcx %rax,%r13
+ adox %r15,%r14
+
+ .byte 0xc4,0x62,0xfb,0xf6,0xbd,0x38,0x00,0x00,0x00 # mulx 7*8($aaptr),%rax,%r15
+ mov 8($aptr,%rcx,8),%rdx # a[i]
+ adcx %rax,%r14
+ adox %rbx,%r15 # %rbx is 0, of=0
+ adcx %rbx,%r15 # cf=0
+
+ .byte 0x67
+ inc %rcx # of=0
+ jnz .Lsqrx8x_loop
+
+ lea 8*8($aaptr),$aaptr
+ mov \$-8,%rcx
+ cmp 8+8(%rsp),$aaptr # done?
+ je .Lsqrx8x_break
+
+ sub 16+8(%rsp),%rbx # mov 16(%rsp),%cf
+ .byte 0x66
+ mov -64($aptr),%rdx
+ adcx 0*8($tptr),%r8
+ adcx 1*8($tptr),%r9
+ adc 2*8($tptr),%r10
+ adc 3*8($tptr),%r11
+ adc 4*8($tptr),%r12
+ adc 5*8($tptr),%r13
+ adc 6*8($tptr),%r14
+ adc 7*8($tptr),%r15
+ lea 8*8($tptr),$tptr
+ .byte 0x67
+ sbb %rax,%rax # mov %cf,%rax
+ xor %ebx,%ebx # cf=0, of=0
+ mov %rax,16+8(%rsp) # offload carry
+ jmp .Lsqrx8x_loop
+
+.align 32
+.Lsqrx8x_break:
+ xor $zero,$zero
+ sub 16+8(%rsp),%rbx # mov 16(%rsp),%cf
+ adcx $zero,%r8
+ mov 24+8(%rsp),$carry # initial $tptr, borrow $carry
+ adcx $zero,%r9
+ mov 0*8($aptr),%rdx # a[8], modulo-scheduled
+ adc \$0,%r10
+ mov %r8,0*8($tptr)
+ adc \$0,%r11
+ adc \$0,%r12
+ adc \$0,%r13
+ adc \$0,%r14
+ adc \$0,%r15
+ cmp $carry,$tptr # cf=0, of=0
+ je .Lsqrx8x_outer_loop
+
+ mov %r9,1*8($tptr)
+ mov 1*8($carry),%r9
+ mov %r10,2*8($tptr)
+ mov 2*8($carry),%r10
+ mov %r11,3*8($tptr)
+ mov 3*8($carry),%r11
+ mov %r12,4*8($tptr)
+ mov 4*8($carry),%r12
+ mov %r13,5*8($tptr)
+ mov 5*8($carry),%r13
+ mov %r14,6*8($tptr)
+ mov 6*8($carry),%r14
+ mov %r15,7*8($tptr)
+ mov 7*8($carry),%r15
+ mov $carry,$tptr
+ jmp .Lsqrx8x_outer_loop
+
+.align 32
+.Lsqrx8x_outer_break:
+ mov %r9,9*8($tptr) # t[9]
+ movq %xmm3,%rcx # -$num
+ mov %r10,10*8($tptr) # ...
+ mov %r11,11*8($tptr)
+ mov %r12,12*8($tptr)
+ mov %r13,13*8($tptr)
+ mov %r14,14*8($tptr)
+___
+} {
+my $i="%rcx";
+$code.=<<___;
+ lea 48+8(%rsp),$tptr
+ mov ($aptr,$i),%rdx # a[0]
+
+ mov 8($tptr),$A0[1] # t[1]
+ xor $A0[0],$A0[0] # t[0], of=0, cf=0
+ mov 0+8(%rsp),$num # restore $num
+ adox $A0[1],$A0[1]
+ mov 16($tptr),$A1[0] # t[2] # prefetch
+ mov 24($tptr),$A1[1] # t[3] # prefetch
+ #jmp .Lsqrx4x_shift_n_add # happens to be aligned
+
+.align 32
+.Lsqrx4x_shift_n_add:
+ mulx %rdx,%rax,%rbx
+ adox $A1[0],$A1[0]
+ adcx $A0[0],%rax
+ .byte 0x48,0x8b,0x94,0x0e,0x08,0x00,0x00,0x00 # mov 8($aptr,$i),%rdx # a[i+1] # prefetch
+ .byte 0x4c,0x8b,0x97,0x20,0x00,0x00,0x00 # mov 32($tptr),$A0[0] # t[2*i+4] # prefetch
+ adox $A1[1],$A1[1]
+ adcx $A0[1],%rbx
+ mov 40($tptr),$A0[1] # t[2*i+4+1] # prefetch
+ mov %rax,0($tptr)
+ mov %rbx,8($tptr)
+
+ mulx %rdx,%rax,%rbx
+ adox $A0[0],$A0[0]
+ adcx $A1[0],%rax
+ mov 16($aptr,$i),%rdx # a[i+2] # prefetch
+ mov 48($tptr),$A1[0] # t[2*i+6] # prefetch
+ adox $A0[1],$A0[1]
+ adcx $A1[1],%rbx
+ mov 56($tptr),$A1[1] # t[2*i+6+1] # prefetch
+ mov %rax,16($tptr)
+ mov %rbx,24($tptr)
+
+ mulx %rdx,%rax,%rbx
+ adox $A1[0],$A1[0]
+ adcx $A0[0],%rax
+ mov 24($aptr,$i),%rdx # a[i+3] # prefetch
+ lea 32($i),$i
+ mov 64($tptr),$A0[0] # t[2*i+8] # prefetch
+ adox $A1[1],$A1[1]
+ adcx $A0[1],%rbx
+ mov 72($tptr),$A0[1] # t[2*i+8+1] # prefetch
+ mov %rax,32($tptr)
+ mov %rbx,40($tptr)
+
+ mulx %rdx,%rax,%rbx
+ adox $A0[0],$A0[0]
+ adcx $A1[0],%rax
+ jrcxz .Lsqrx4x_shift_n_add_break
+ .byte 0x48,0x8b,0x94,0x0e,0x00,0x00,0x00,0x00 # mov 0($aptr,$i),%rdx # a[i+4] # prefetch
+ adox $A0[1],$A0[1]
+ adcx $A1[1],%rbx
+ mov 80($tptr),$A1[0] # t[2*i+10] # prefetch
+ mov 88($tptr),$A1[1] # t[2*i+10+1] # prefetch
+ mov %rax,48($tptr)
+ mov %rbx,56($tptr)
+ lea 64($tptr),$tptr
+ nop
+ jmp .Lsqrx4x_shift_n_add
+
+.align 32
+.Lsqrx4x_shift_n_add_break:
+ adcx $A1[1],%rbx
+ mov %rax,48($tptr)
+ mov %rbx,56($tptr)
+ lea 64($tptr),$tptr # end of t[] buffer
+___
+}
+######################################################################
+# Montgomery reduction part, "word-by-word" algorithm.
+#
+# This new path is inspired by multiple submissions from Intel, by
+# Shay Gueron, Vlad Krasnov, Erdinc Ozturk, James Guilford,
+# Vinodh Gopal...
+{
+my ($nptr,$carry,$m0)=("%rbp","%rsi","%rdx");
+
+$code.=<<___;
+ movq %xmm2,$nptr
+__bn_sqrx8x_reduction:
+ xor %eax,%eax # initial top-most carry bit
+ mov 32+8(%rsp),%rbx # n0
+ mov 48+8(%rsp),%rdx # "%r8", 8*0($tptr)
+ lea -8*8($nptr,$num),%rcx # end of n[]
+ #lea 48+8(%rsp,$num,2),$tptr # end of t[] buffer
+ mov %rcx, 0+8(%rsp) # save end of n[]
+ mov $tptr,8+8(%rsp) # save end of t[]
+
+ lea 48+8(%rsp),$tptr # initial t[] window
+ jmp .Lsqrx8x_reduction_loop
+
+.align 32
+.Lsqrx8x_reduction_loop:
+ mov 8*1($tptr),%r9
+ mov 8*2($tptr),%r10
+ mov 8*3($tptr),%r11
+ mov 8*4($tptr),%r12
+ mov %rdx,%r8
+ imulq %rbx,%rdx # n0*a[i]
+ mov 8*5($tptr),%r13
+ mov 8*6($tptr),%r14
+ mov 8*7($tptr),%r15
+ mov %rax,24+8(%rsp) # store top-most carry bit
+
+ lea 8*8($tptr),$tptr
+ xor $carry,$carry # cf=0,of=0
+ mov \$-8,%rcx
+ jmp .Lsqrx8x_reduce
+
+.align 32
+.Lsqrx8x_reduce:
+ mov %r8, %rbx
+ mulx 8*0($nptr),%rax,%r8 # n[0]
+ adcx %rbx,%rax # discarded
+ adox %r9,%r8
+
+ mulx 8*1($nptr),%rbx,%r9 # n[1]
+ adcx %rbx,%r8
+ adox %r10,%r9
+
+ mulx 8*2($nptr),%rbx,%r10
+ adcx %rbx,%r9
+ adox %r11,%r10
+
+ mulx 8*3($nptr),%rbx,%r11
+ adcx %rbx,%r10
+ adox %r12,%r11
+
+ .byte 0xc4,0x62,0xe3,0xf6,0xa5,0x20,0x00,0x00,0x00 # mulx 8*4($nptr),%rbx,%r12
+ mov %rdx,%rax
+ mov %r8,%rdx
+ adcx %rbx,%r11
+ adox %r13,%r12
+
+ mulx 32+8(%rsp),%rbx,%rdx # %rdx discarded
+ mov %rax,%rdx
+ mov %rax,64+48+8(%rsp,%rcx,8) # put aside n0*a[i]
+
+ mulx 8*5($nptr),%rax,%r13
+ adcx %rax,%r12
+ adox %r14,%r13
+
+ mulx 8*6($nptr),%rax,%r14
+ adcx %rax,%r13
+ adox %r15,%r14
+
+ mulx 8*7($nptr),%rax,%r15
+ mov %rbx,%rdx
+ adcx %rax,%r14
+ adox $carry,%r15 # $carry is 0
+ adcx $carry,%r15 # cf=0
+
+ .byte 0x67,0x67,0x67
+ inc %rcx # of=0
+ jnz .Lsqrx8x_reduce
+
+ mov $carry,%rax # xor %rax,%rax
+ cmp 0+8(%rsp),$nptr # end of n[]?
+ jae .Lsqrx8x_no_tail
+
+ mov 48+8(%rsp),%rdx # pull n0*a[0]
+ add 8*0($tptr),%r8
+ lea 8*8($nptr),$nptr
+ mov \$-8,%rcx
+ adcx 8*1($tptr),%r9
+ adcx 8*2($tptr),%r10
+ adc 8*3($tptr),%r11
+ adc 8*4($tptr),%r12
+ adc 8*5($tptr),%r13
+ adc 8*6($tptr),%r14
+ adc 8*7($tptr),%r15
+ lea 8*8($tptr),$tptr
+ sbb %rax,%rax # top carry
+
+ xor $carry,$carry # of=0, cf=0
+ mov %rax,16+8(%rsp)
+ jmp .Lsqrx8x_tail
+
+.align 32
+.Lsqrx8x_tail:
+ mov %r8,%rbx
+ mulx 8*0($nptr),%rax,%r8
+ adcx %rax,%rbx
+ adox %r9,%r8
+
+ mulx 8*1($nptr),%rax,%r9
+ adcx %rax,%r8
+ adox %r10,%r9
+
+ mulx 8*2($nptr),%rax,%r10
+ adcx %rax,%r9
+ adox %r11,%r10
+
+ mulx 8*3($nptr),%rax,%r11
+ adcx %rax,%r10
+ adox %r12,%r11
+
+ .byte 0xc4,0x62,0xfb,0xf6,0xa5,0x20,0x00,0x00,0x00 # mulx 8*4($nptr),%rax,%r12
+ adcx %rax,%r11
+ adox %r13,%r12
+
+ mulx 8*5($nptr),%rax,%r13
+ adcx %rax,%r12
+ adox %r14,%r13
+
+ mulx 8*6($nptr),%rax,%r14
+ adcx %rax,%r13
+ adox %r15,%r14
+
+ mulx 8*7($nptr),%rax,%r15
+ mov 72+48+8(%rsp,%rcx,8),%rdx # pull n0*a[i]
+ adcx %rax,%r14
+ adox $carry,%r15
+ mov %rbx,($tptr,%rcx,8) # save result
+ mov %r8,%rbx
+ adcx $carry,%r15 # cf=0
+
+ inc %rcx # of=0
+ jnz .Lsqrx8x_tail
+
+ cmp 0+8(%rsp),$nptr # end of n[]?
+ jae .Lsqrx8x_tail_done # break out of loop
+
+ sub 16+8(%rsp),$carry # mov 16(%rsp),%cf
+ mov 48+8(%rsp),%rdx # pull n0*a[0]
+ lea 8*8($nptr),$nptr
+ adc 8*0($tptr),%r8
+ adc 8*1($tptr),%r9
+ adc 8*2($tptr),%r10
+ adc 8*3($tptr),%r11
+ adc 8*4($tptr),%r12
+ adc 8*5($tptr),%r13
+ adc 8*6($tptr),%r14
+ adc 8*7($tptr),%r15
+ lea 8*8($tptr),$tptr
+ sbb %rax,%rax
+ sub \$8,%rcx # mov \$-8,%rcx
+
+ xor $carry,$carry # of=0, cf=0
+ mov %rax,16+8(%rsp)
+ jmp .Lsqrx8x_tail
+
+.align 32
+.Lsqrx8x_tail_done:
+ xor %rax,%rax
+ add 24+8(%rsp),%r8 # can this overflow?
+ adc \$0,%r9
+ adc \$0,%r10
+ adc \$0,%r11
+ adc \$0,%r12
+ adc \$0,%r13
+ adc \$0,%r14
+ adc \$0,%r15
+ adc \$0,%rax
+
+ sub 16+8(%rsp),$carry # mov 16(%rsp),%cf
+.Lsqrx8x_no_tail: # %cf is 0 if jumped here
+ adc 8*0($tptr),%r8
+ movq %xmm3,%rcx
+ adc 8*1($tptr),%r9
+ mov 8*7($nptr),$carry
+ movq %xmm2,$nptr # restore $nptr
+ adc 8*2($tptr),%r10
+ adc 8*3($tptr),%r11
+ adc 8*4($tptr),%r12
+ adc 8*5($tptr),%r13
+ adc 8*6($tptr),%r14
+ adc 8*7($tptr),%r15
+ adc \$0,%rax # top-most carry
+
+ mov 32+8(%rsp),%rbx # n0
+ mov 8*8($tptr,%rcx),%rdx # modulo-scheduled "%r8"
+
+ mov %r8,8*0($tptr) # store top 512 bits
+ lea 8*8($tptr),%r8 # borrow %r8
+ mov %r9,8*1($tptr)
+ mov %r10,8*2($tptr)
+ mov %r11,8*3($tptr)
+ mov %r12,8*4($tptr)
+ mov %r13,8*5($tptr)
+ mov %r14,8*6($tptr)
+ mov %r15,8*7($tptr)
+
+ lea 8*8($tptr,%rcx),$tptr # start of current t[] window
+ cmp 8+8(%rsp),%r8 # end of t[]?
+ jb .Lsqrx8x_reduction_loop
+ ret
+.cfi_endproc
+.size bn_sqrx8x_internal,.-bn_sqrx8x_internal
+___
+}
+##############################################################
+# Post-condition, 4x unrolled
+#
+{
+my ($rptr,$nptr)=("%rdx","%rbp");
+$code.=<<___;
+.align 32
+__bn_postx4x_internal:
+.cfi_startproc
+ mov 8*0($nptr),%r12
+ mov %rcx,%r10 # -$num
+ mov %rcx,%r9 # -$num
+ neg %rax
+ sar \$3+2,%rcx
+ #lea 48+8(%rsp,%r9),$tptr
+ movq %xmm1,$rptr # restore $rptr
+ movq %xmm1,$aptr # prepare for back-to-back call
+ dec %r12 # so that after 'not' we get -n[0]
+ mov 8*1($nptr),%r13
+ xor %r8,%r8
+ mov 8*2($nptr),%r14
+ mov 8*3($nptr),%r15
+ jmp .Lsqrx4x_sub_entry
+
+.align 16
+.Lsqrx4x_sub:
+ mov 8*0($nptr),%r12
+ mov 8*1($nptr),%r13
+ mov 8*2($nptr),%r14
+ mov 8*3($nptr),%r15
+.Lsqrx4x_sub_entry:
+ andn %rax,%r12,%r12
+ lea 8*4($nptr),$nptr
+ andn %rax,%r13,%r13
+ andn %rax,%r14,%r14
+ andn %rax,%r15,%r15
+
+ neg %r8 # mov %r8,%cf
+ adc 8*0($tptr),%r12
+ adc 8*1($tptr),%r13
+ adc 8*2($tptr),%r14
+ adc 8*3($tptr),%r15
+ mov %r12,8*0($rptr)
+ lea 8*4($tptr),$tptr
+ mov %r13,8*1($rptr)
+ sbb %r8,%r8 # mov %cf,%r8
+ mov %r14,8*2($rptr)
+ mov %r15,8*3($rptr)
+ lea 8*4($rptr),$rptr
+
+ inc %rcx
+ jnz .Lsqrx4x_sub
+
+ neg %r9 # restore $num
+
+ ret
+.cfi_endproc
+.size __bn_postx4x_internal,.-__bn_postx4x_internal
+___
+}
+}}}
+{
+my ($inp,$num,$tbl,$idx)=$win64?("%rcx","%edx","%r8", "%r9d") : # Win64 order
+ ("%rdi","%esi","%rdx","%ecx"); # Unix order
+my $out=$inp;
+my $STRIDE=2**5*8;
+my $N=$STRIDE/4;
+
+$code.=<<___;
+.globl bn_get_bits5
+.type bn_get_bits5,\@abi-omnipotent
+.align 16
+bn_get_bits5:
+.cfi_startproc
+ lea 0($inp),%r10
+ lea 1($inp),%r11
+ mov $num,%ecx
+ shr \$4,$num
+ and \$15,%ecx
+ lea -8(%ecx),%eax
+ cmp \$11,%ecx
+ cmova %r11,%r10
+ cmova %eax,%ecx
+ movzw (%r10,$num,2),%eax
+ shrl %cl,%eax
+ and \$31,%eax
+ ret
+.cfi_endproc
+.size bn_get_bits5,.-bn_get_bits5
+
+.globl bn_scatter5
+.type bn_scatter5,\@abi-omnipotent
+.align 16
+bn_scatter5:
+.cfi_startproc
+ cmp \$0, $num
+ jz .Lscatter_epilogue
+ lea ($tbl,$idx,8),$tbl
+.Lscatter:
+ mov ($inp),%rax
+ lea 8($inp),$inp
+ mov %rax,($tbl)
+ lea 32*8($tbl),$tbl
+ sub \$1,$num
+ jnz .Lscatter
+.Lscatter_epilogue:
+ ret
+.cfi_endproc
+.size bn_scatter5,.-bn_scatter5
+
+.globl bn_gather5
+.type bn_gather5,\@abi-omnipotent
+.align 32
+bn_gather5:
+.LSEH_begin_bn_gather5: # Win64 thing, but harmless in other cases
+.cfi_startproc
+ # I can't trust assembler to use specific encoding:-(
+ .byte 0x4c,0x8d,0x14,0x24 #lea (%rsp),%r10
+ .byte 0x48,0x81,0xec,0x08,0x01,0x00,0x00 #sub $0x108,%rsp
+ lea .Linc(%rip),%rax
+ and \$-16,%rsp # shouldn't be formally required
+
+ movd $idx,%xmm5
+ movdqa 0(%rax),%xmm0 # 00000001000000010000000000000000
+ movdqa 16(%rax),%xmm1 # 00000002000000020000000200000002
+ lea 128($tbl),%r11 # size optimization
+ lea 128(%rsp),%rax # size optimization
+
+ pshufd \$0,%xmm5,%xmm5 # broadcast $idx
+ movdqa %xmm1,%xmm4
+ movdqa %xmm1,%xmm2
+___
+########################################################################
+# calculate mask by comparing 0..31 to $idx and save result to stack
+#
+for($i=0;$i<$STRIDE/16;$i+=4) {
+$code.=<<___;
+ paddd %xmm0,%xmm1
+ pcmpeqd %xmm5,%xmm0 # compare to 1,0
+___
+$code.=<<___ if ($i);
+ movdqa %xmm3,`16*($i-1)-128`(%rax)
+___
+$code.=<<___;
+ movdqa %xmm4,%xmm3
+
+ paddd %xmm1,%xmm2
+ pcmpeqd %xmm5,%xmm1 # compare to 3,2
+ movdqa %xmm0,`16*($i+0)-128`(%rax)
+ movdqa %xmm4,%xmm0
+
+ paddd %xmm2,%xmm3
+ pcmpeqd %xmm5,%xmm2 # compare to 5,4
+ movdqa %xmm1,`16*($i+1)-128`(%rax)
+ movdqa %xmm4,%xmm1
+
+ paddd %xmm3,%xmm0
+ pcmpeqd %xmm5,%xmm3 # compare to 7,6
+ movdqa %xmm2,`16*($i+2)-128`(%rax)
+ movdqa %xmm4,%xmm2
+___
+}
+$code.=<<___;
+ movdqa %xmm3,`16*($i-1)-128`(%rax)
+ jmp .Lgather
+
+.align 32
+.Lgather:
+ pxor %xmm4,%xmm4
+ pxor %xmm5,%xmm5
+___
+for($i=0;$i<$STRIDE/16;$i+=4) {
+$code.=<<___;
+ movdqa `16*($i+0)-128`(%r11),%xmm0
+ movdqa `16*($i+1)-128`(%r11),%xmm1
+ movdqa `16*($i+2)-128`(%r11),%xmm2
+ pand `16*($i+0)-128`(%rax),%xmm0
+ movdqa `16*($i+3)-128`(%r11),%xmm3
+ pand `16*($i+1)-128`(%rax),%xmm1
+ por %xmm0,%xmm4
+ pand `16*($i+2)-128`(%rax),%xmm2
+ por %xmm1,%xmm5
+ pand `16*($i+3)-128`(%rax),%xmm3
+ por %xmm2,%xmm4
+ por %xmm3,%xmm5
+___
+}
+$code.=<<___;
+ por %xmm5,%xmm4
+ lea $STRIDE(%r11),%r11
+ pshufd \$0x4e,%xmm4,%xmm0
+ por %xmm4,%xmm0
+ movq %xmm0,($out) # m0=bp[0]
+ lea 8($out),$out
+ sub \$1,$num
+ jnz .Lgather
+
+ lea (%r10),%rsp
+ ret
+.LSEH_end_bn_gather5:
+.cfi_endproc
+.size bn_gather5,.-bn_gather5
+___
+}
+$code.=<<___;
+.align 64
+.Linc:
+ .long 0,0, 1,1
+ .long 2,2, 2,2
+.asciz "Montgomery Multiplication with scatter/gather for x86_64, CRYPTOGAMS by <appro\@openssl.org>"
+___
+
+# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame,
+# CONTEXT *context,DISPATCHER_CONTEXT *disp)
+if ($win64) {
+$rec="%rcx";
+$frame="%rdx";
+$context="%r8";
+$disp="%r9";
+
+$code.=<<___;
+.extern __imp_RtlVirtualUnwind
+.type mul_handler,\@abi-omnipotent
+.align 16
+mul_handler:
+ push %rsi
+ push %rdi
+ push %rbx
+ push %rbp
+ push %r12
+ push %r13
+ push %r14
+ push %r15
+ pushfq
+ sub \$64,%rsp
+
+ mov 120($context),%rax # pull context->Rax
+ mov 248($context),%rbx # pull context->Rip
+
+ mov 8($disp),%rsi # disp->ImageBase
+ mov 56($disp),%r11 # disp->HandlerData
+
+ mov 0(%r11),%r10d # HandlerData[0]
+ lea (%rsi,%r10),%r10 # end of prologue label
+ cmp %r10,%rbx # context->Rip<end of prologue label
+ jb .Lcommon_seh_tail
+
+ mov 4(%r11),%r10d # HandlerData[1]
+ lea (%rsi,%r10),%r10 # beginning of body label
+ cmp %r10,%rbx # context->Rip<body label
+ jb .Lcommon_pop_regs
+
+ mov 152($context),%rax # pull context->Rsp
+
+ mov 8(%r11),%r10d # HandlerData[2]
+ lea (%rsi,%r10),%r10 # epilogue label
+ cmp %r10,%rbx # context->Rip>=epilogue label
+ jae .Lcommon_seh_tail
+
+ lea .Lmul_epilogue(%rip),%r10
+ cmp %r10,%rbx
+ ja .Lbody_40
+
+ mov 192($context),%r10 # pull $num
+ mov 8(%rax,%r10,8),%rax # pull saved stack pointer
+
+ jmp .Lcommon_pop_regs
+
+.Lbody_40:
+ mov 40(%rax),%rax # pull saved stack pointer
+.Lcommon_pop_regs:
+ mov -8(%rax),%rbx
+ mov -16(%rax),%rbp
+ mov -24(%rax),%r12
+ mov -32(%rax),%r13
+ mov -40(%rax),%r14
+ mov -48(%rax),%r15
+ mov %rbx,144($context) # restore context->Rbx
+ mov %rbp,160($context) # restore context->Rbp
+ mov %r12,216($context) # restore context->R12
+ mov %r13,224($context) # restore context->R13
+ mov %r14,232($context) # restore context->R14
+ mov %r15,240($context) # restore context->R15
+
+.Lcommon_seh_tail:
+ mov 8(%rax),%rdi
+ mov 16(%rax),%rsi
+ mov %rax,152($context) # restore context->Rsp
+ mov %rsi,168($context) # restore context->Rsi
+ mov %rdi,176($context) # restore context->Rdi
+
+ mov 40($disp),%rdi # disp->ContextRecord
+ mov $context,%rsi # context
+ mov \$154,%ecx # sizeof(CONTEXT)
+ .long 0xa548f3fc # cld; rep movsq
+
+ mov $disp,%rsi
+ xor %rcx,%rcx # arg1, UNW_FLAG_NHANDLER
+ mov 8(%rsi),%rdx # arg2, disp->ImageBase
+ mov 0(%rsi),%r8 # arg3, disp->ControlPc
+ mov 16(%rsi),%r9 # arg4, disp->FunctionEntry
+ mov 40(%rsi),%r10 # disp->ContextRecord
+ lea 56(%rsi),%r11 # &disp->HandlerData
+ lea 24(%rsi),%r12 # &disp->EstablisherFrame
+ mov %r10,32(%rsp) # arg5
+ mov %r11,40(%rsp) # arg6
+ mov %r12,48(%rsp) # arg7
+ mov %rcx,56(%rsp) # arg8, (NULL)
+ call *__imp_RtlVirtualUnwind(%rip)
+
+ mov \$1,%eax # ExceptionContinueSearch
+ add \$64,%rsp
+ popfq
+ pop %r15
+ pop %r14
+ pop %r13
+ pop %r12
+ pop %rbp
+ pop %rbx
+ pop %rdi
+ pop %rsi
+ ret
+.size mul_handler,.-mul_handler
+
+.section .pdata
+.align 4
+ .rva .LSEH_begin_bn_mul_mont_gather5
+ .rva .LSEH_end_bn_mul_mont_gather5
+ .rva .LSEH_info_bn_mul_mont_gather5
+
+ .rva .LSEH_begin_bn_mul4x_mont_gather5
+ .rva .LSEH_end_bn_mul4x_mont_gather5
+ .rva .LSEH_info_bn_mul4x_mont_gather5
+
+ .rva .LSEH_begin_bn_power5
+ .rva .LSEH_end_bn_power5
+ .rva .LSEH_info_bn_power5
+___
+$code.=<<___ if ($addx);
+ .rva .LSEH_begin_bn_mulx4x_mont_gather5
+ .rva .LSEH_end_bn_mulx4x_mont_gather5
+ .rva .LSEH_info_bn_mulx4x_mont_gather5
+
+ .rva .LSEH_begin_bn_powerx5
+ .rva .LSEH_end_bn_powerx5
+ .rva .LSEH_info_bn_powerx5
+___
+$code.=<<___;
+ .rva .LSEH_begin_bn_gather5
+ .rva .LSEH_end_bn_gather5
+ .rva .LSEH_info_bn_gather5
+
+.section .xdata
+.align 8
+.LSEH_info_bn_mul_mont_gather5:
+ .byte 9,0,0,0
+ .rva mul_handler
+ .rva .Lmul_body,.Lmul_body,.Lmul_epilogue # HandlerData[]
+.align 8
+.LSEH_info_bn_mul4x_mont_gather5:
+ .byte 9,0,0,0
+ .rva mul_handler
+ .rva .Lmul4x_prologue,.Lmul4x_body,.Lmul4x_epilogue # HandlerData[]
+.align 8
+.LSEH_info_bn_power5:
+ .byte 9,0,0,0
+ .rva mul_handler
+ .rva .Lpower5_prologue,.Lpower5_body,.Lpower5_epilogue # HandlerData[]
+___
+$code.=<<___ if ($addx);
+.align 8
+.LSEH_info_bn_mulx4x_mont_gather5:
+ .byte 9,0,0,0
+ .rva mul_handler
+ .rva .Lmulx4x_prologue,.Lmulx4x_body,.Lmulx4x_epilogue # HandlerData[]
+.align 8
+.LSEH_info_bn_powerx5:
+ .byte 9,0,0,0
+ .rva mul_handler
+ .rva .Lpowerx5_prologue,.Lpowerx5_body,.Lpowerx5_epilogue # HandlerData[]
+___
+$code.=<<___;
+.align 8
+.LSEH_info_bn_gather5:
+ .byte 0x01,0x0b,0x03,0x0a
+ .byte 0x0b,0x01,0x21,0x00 # sub rsp,0x108
+ .byte 0x04,0xa3,0x00,0x00 # lea r10,(rsp)
+.align 8
+___
+}
+
+$code =~ s/\`([^\`]*)\`/eval($1)/gem;
+
+print $code;
+close STDOUT or die "error closing STDOUT: $!";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_add.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_add.c
new file mode 100644
index 0000000..8ffe496
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_add.c
@@ -0,0 +1,171 @@
+/*
+ * Copyright 1995-2018 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include "internal/cryptlib.h"
+#include "bn_local.h"
+
+/* signed add of b to a. */
+int BN_add(BIGNUM *r, const BIGNUM *a, const BIGNUM *b)
+{
+ int ret, r_neg, cmp_res;
+
+ bn_check_top(a);
+ bn_check_top(b);
+
+ if (a->neg == b->neg) {
+ r_neg = a->neg;
+ ret = BN_uadd(r, a, b);
+ } else {
+ cmp_res = BN_ucmp(a, b);
+ if (cmp_res > 0) {
+ r_neg = a->neg;
+ ret = BN_usub(r, a, b);
+ } else if (cmp_res < 0) {
+ r_neg = b->neg;
+ ret = BN_usub(r, b, a);
+ } else {
+ r_neg = 0;
+ BN_zero(r);
+ ret = 1;
+ }
+ }
+
+ r->neg = r_neg;
+ bn_check_top(r);
+ return ret;
+}
+
+/* signed sub of b from a. */
+int BN_sub(BIGNUM *r, const BIGNUM *a, const BIGNUM *b)
+{
+ int ret, r_neg, cmp_res;
+
+ bn_check_top(a);
+ bn_check_top(b);
+
+ if (a->neg != b->neg) {
+ r_neg = a->neg;
+ ret = BN_uadd(r, a, b);
+ } else {
+ cmp_res = BN_ucmp(a, b);
+ if (cmp_res > 0) {
+ r_neg = a->neg;
+ ret = BN_usub(r, a, b);
+ } else if (cmp_res < 0) {
+ r_neg = !b->neg;
+ ret = BN_usub(r, b, a);
+ } else {
+ r_neg = 0;
+ BN_zero(r);
+ ret = 1;
+ }
+ }
+
+ r->neg = r_neg;
+ bn_check_top(r);
+ return ret;
+}
+
+/* unsigned add of b to a, r can be equal to a or b. */
+int BN_uadd(BIGNUM *r, const BIGNUM *a, const BIGNUM *b)
+{
+ int max, min, dif;
+ const BN_ULONG *ap, *bp;
+ BN_ULONG *rp, carry, t1, t2;
+
+ bn_check_top(a);
+ bn_check_top(b);
+
+ if (a->top < b->top) {
+ const BIGNUM *tmp;
+
+ tmp = a;
+ a = b;
+ b = tmp;
+ }
+ max = a->top;
+ min = b->top;
+ dif = max - min;
+
+ if (bn_wexpand(r, max + 1) == NULL)
+ return 0;
+
+ r->top = max;
+
+ ap = a->d;
+ bp = b->d;
+ rp = r->d;
+
+ carry = bn_add_words(rp, ap, bp, min);
+ rp += min;
+ ap += min;
+
+ while (dif) {
+ dif--;
+ t1 = *(ap++);
+ t2 = (t1 + carry) & BN_MASK2;
+ *(rp++) = t2;
+ carry &= (t2 == 0);
+ }
+ *rp = carry;
+ r->top += carry;
+
+ r->neg = 0;
+ bn_check_top(r);
+ return 1;
+}
+
+/* unsigned subtraction of b from a, a must be larger than b. */
+int BN_usub(BIGNUM *r, const BIGNUM *a, const BIGNUM *b)
+{
+ int max, min, dif;
+ BN_ULONG t1, t2, borrow, *rp;
+ const BN_ULONG *ap, *bp;
+
+ bn_check_top(a);
+ bn_check_top(b);
+
+ max = a->top;
+ min = b->top;
+ dif = max - min;
+
+ if (dif < 0) { /* hmm... should not be happening */
+ BNerr(BN_F_BN_USUB, BN_R_ARG2_LT_ARG3);
+ return 0;
+ }
+
+ if (bn_wexpand(r, max) == NULL)
+ return 0;
+
+ ap = a->d;
+ bp = b->d;
+ rp = r->d;
+
+ borrow = bn_sub_words(rp, ap, bp, min);
+ ap += min;
+ rp += min;
+
+ while (dif) {
+ dif--;
+ t1 = *(ap++);
+ t2 = (t1 - borrow) & BN_MASK2;
+ *(rp++) = t2;
+ borrow &= (t1 == 0);
+ }
+
+ while (max && *--rp == 0)
+ max--;
+
+ r->top = max;
+ r->neg = 0;
+ bn_pollute(r);
+
+ return 1;
+}
+
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_asm.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_asm.c
new file mode 100644
index 0000000..4d83a8c
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_asm.c
@@ -0,0 +1,1039 @@
+/*
+ * Copyright 1995-2016 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include <assert.h>
+#include <openssl/crypto.h>
+#include "internal/cryptlib.h"
+#include "bn_local.h"
+
+#if defined(BN_LLONG) || defined(BN_UMULT_HIGH)
+
+BN_ULONG bn_mul_add_words(BN_ULONG *rp, const BN_ULONG *ap, int num,
+ BN_ULONG w)
+{
+ BN_ULONG c1 = 0;
+
+ assert(num >= 0);
+ if (num <= 0)
+ return c1;
+
+# ifndef OPENSSL_SMALL_FOOTPRINT
+ while (num & ~3) {
+ mul_add(rp[0], ap[0], w, c1);
+ mul_add(rp[1], ap[1], w, c1);
+ mul_add(rp[2], ap[2], w, c1);
+ mul_add(rp[3], ap[3], w, c1);
+ ap += 4;
+ rp += 4;
+ num -= 4;
+ }
+# endif
+ while (num) {
+ mul_add(rp[0], ap[0], w, c1);
+ ap++;
+ rp++;
+ num--;
+ }
+
+ return c1;
+}
+
+BN_ULONG bn_mul_words(BN_ULONG *rp, const BN_ULONG *ap, int num, BN_ULONG w)
+{
+ BN_ULONG c1 = 0;
+
+ assert(num >= 0);
+ if (num <= 0)
+ return c1;
+
+# ifndef OPENSSL_SMALL_FOOTPRINT
+ while (num & ~3) {
+ mul(rp[0], ap[0], w, c1);
+ mul(rp[1], ap[1], w, c1);
+ mul(rp[2], ap[2], w, c1);
+ mul(rp[3], ap[3], w, c1);
+ ap += 4;
+ rp += 4;
+ num -= 4;
+ }
+# endif
+ while (num) {
+ mul(rp[0], ap[0], w, c1);
+ ap++;
+ rp++;
+ num--;
+ }
+ return c1;
+}
+
+void bn_sqr_words(BN_ULONG *r, const BN_ULONG *a, int n)
+{
+ assert(n >= 0);
+ if (n <= 0)
+ return;
+
+# ifndef OPENSSL_SMALL_FOOTPRINT
+ while (n & ~3) {
+ sqr(r[0], r[1], a[0]);
+ sqr(r[2], r[3], a[1]);
+ sqr(r[4], r[5], a[2]);
+ sqr(r[6], r[7], a[3]);
+ a += 4;
+ r += 8;
+ n -= 4;
+ }
+# endif
+ while (n) {
+ sqr(r[0], r[1], a[0]);
+ a++;
+ r += 2;
+ n--;
+ }
+}
+
+#else /* !(defined(BN_LLONG) ||
+ * defined(BN_UMULT_HIGH)) */
+
+BN_ULONG bn_mul_add_words(BN_ULONG *rp, const BN_ULONG *ap, int num,
+ BN_ULONG w)
+{
+ BN_ULONG c = 0;
+ BN_ULONG bl, bh;
+
+ assert(num >= 0);
+ if (num <= 0)
+ return (BN_ULONG)0;
+
+ bl = LBITS(w);
+ bh = HBITS(w);
+
+# ifndef OPENSSL_SMALL_FOOTPRINT
+ while (num & ~3) {
+ mul_add(rp[0], ap[0], bl, bh, c);
+ mul_add(rp[1], ap[1], bl, bh, c);
+ mul_add(rp[2], ap[2], bl, bh, c);
+ mul_add(rp[3], ap[3], bl, bh, c);
+ ap += 4;
+ rp += 4;
+ num -= 4;
+ }
+# endif
+ while (num) {
+ mul_add(rp[0], ap[0], bl, bh, c);
+ ap++;
+ rp++;
+ num--;
+ }
+ return c;
+}
+
+BN_ULONG bn_mul_words(BN_ULONG *rp, const BN_ULONG *ap, int num, BN_ULONG w)
+{
+ BN_ULONG carry = 0;
+ BN_ULONG bl, bh;
+
+ assert(num >= 0);
+ if (num <= 0)
+ return (BN_ULONG)0;
+
+ bl = LBITS(w);
+ bh = HBITS(w);
+
+# ifndef OPENSSL_SMALL_FOOTPRINT
+ while (num & ~3) {
+ mul(rp[0], ap[0], bl, bh, carry);
+ mul(rp[1], ap[1], bl, bh, carry);
+ mul(rp[2], ap[2], bl, bh, carry);
+ mul(rp[3], ap[3], bl, bh, carry);
+ ap += 4;
+ rp += 4;
+ num -= 4;
+ }
+# endif
+ while (num) {
+ mul(rp[0], ap[0], bl, bh, carry);
+ ap++;
+ rp++;
+ num--;
+ }
+ return carry;
+}
+
+void bn_sqr_words(BN_ULONG *r, const BN_ULONG *a, int n)
+{
+ assert(n >= 0);
+ if (n <= 0)
+ return;
+
+# ifndef OPENSSL_SMALL_FOOTPRINT
+ while (n & ~3) {
+ sqr64(r[0], r[1], a[0]);
+ sqr64(r[2], r[3], a[1]);
+ sqr64(r[4], r[5], a[2]);
+ sqr64(r[6], r[7], a[3]);
+ a += 4;
+ r += 8;
+ n -= 4;
+ }
+# endif
+ while (n) {
+ sqr64(r[0], r[1], a[0]);
+ a++;
+ r += 2;
+ n--;
+ }
+}
+
+#endif /* !(defined(BN_LLONG) ||
+ * defined(BN_UMULT_HIGH)) */
+
+#if defined(BN_LLONG) && defined(BN_DIV2W)
+
+BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d)
+{
+ return ((BN_ULONG)(((((BN_ULLONG) h) << BN_BITS2) | l) / (BN_ULLONG) d));
+}
+
+#else
+
+/* Divide h,l by d and return the result. */
+/* I need to test this some more :-( */
+BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d)
+{
+ BN_ULONG dh, dl, q, ret = 0, th, tl, t;
+ int i, count = 2;
+
+ if (d == 0)
+ return BN_MASK2;
+
+ i = BN_num_bits_word(d);
+ assert((i == BN_BITS2) || (h <= (BN_ULONG)1 << i));
+
+ i = BN_BITS2 - i;
+ if (h >= d)
+ h -= d;
+
+ if (i) {
+ d <<= i;
+ h = (h << i) | (l >> (BN_BITS2 - i));
+ l <<= i;
+ }
+ dh = (d & BN_MASK2h) >> BN_BITS4;
+ dl = (d & BN_MASK2l);
+ for (;;) {
+ if ((h >> BN_BITS4) == dh)
+ q = BN_MASK2l;
+ else
+ q = h / dh;
+
+ th = q * dh;
+ tl = dl * q;
+ for (;;) {
+ t = h - th;
+ if ((t & BN_MASK2h) ||
+ ((tl) <= ((t << BN_BITS4) | ((l & BN_MASK2h) >> BN_BITS4))))
+ break;
+ q--;
+ th -= dh;
+ tl -= dl;
+ }
+ t = (tl >> BN_BITS4);
+ tl = (tl << BN_BITS4) & BN_MASK2h;
+ th += t;
+
+ if (l < tl)
+ th++;
+ l -= tl;
+ if (h < th) {
+ h += d;
+ q--;
+ }
+ h -= th;
+
+ if (--count == 0)
+ break;
+
+ ret = q << BN_BITS4;
+ h = ((h << BN_BITS4) | (l >> BN_BITS4)) & BN_MASK2;
+ l = (l & BN_MASK2l) << BN_BITS4;
+ }
+ ret |= q;
+ return ret;
+}
+#endif /* !defined(BN_LLONG) && defined(BN_DIV2W) */
+
+#ifdef BN_LLONG
+BN_ULONG bn_add_words(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b,
+ int n)
+{
+ BN_ULLONG ll = 0;
+
+ assert(n >= 0);
+ if (n <= 0)
+ return (BN_ULONG)0;
+
+# ifndef OPENSSL_SMALL_FOOTPRINT
+ while (n & ~3) {
+ ll += (BN_ULLONG) a[0] + b[0];
+ r[0] = (BN_ULONG)ll & BN_MASK2;
+ ll >>= BN_BITS2;
+ ll += (BN_ULLONG) a[1] + b[1];
+ r[1] = (BN_ULONG)ll & BN_MASK2;
+ ll >>= BN_BITS2;
+ ll += (BN_ULLONG) a[2] + b[2];
+ r[2] = (BN_ULONG)ll & BN_MASK2;
+ ll >>= BN_BITS2;
+ ll += (BN_ULLONG) a[3] + b[3];
+ r[3] = (BN_ULONG)ll & BN_MASK2;
+ ll >>= BN_BITS2;
+ a += 4;
+ b += 4;
+ r += 4;
+ n -= 4;
+ }
+# endif
+ while (n) {
+ ll += (BN_ULLONG) a[0] + b[0];
+ r[0] = (BN_ULONG)ll & BN_MASK2;
+ ll >>= BN_BITS2;
+ a++;
+ b++;
+ r++;
+ n--;
+ }
+ return (BN_ULONG)ll;
+}
+#else /* !BN_LLONG */
+BN_ULONG bn_add_words(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b,
+ int n)
+{
+ BN_ULONG c, l, t;
+
+ assert(n >= 0);
+ if (n <= 0)
+ return (BN_ULONG)0;
+
+ c = 0;
+# ifndef OPENSSL_SMALL_FOOTPRINT
+ while (n & ~3) {
+ t = a[0];
+ t = (t + c) & BN_MASK2;
+ c = (t < c);
+ l = (t + b[0]) & BN_MASK2;
+ c += (l < t);
+ r[0] = l;
+ t = a[1];
+ t = (t + c) & BN_MASK2;
+ c = (t < c);
+ l = (t + b[1]) & BN_MASK2;
+ c += (l < t);
+ r[1] = l;
+ t = a[2];
+ t = (t + c) & BN_MASK2;
+ c = (t < c);
+ l = (t + b[2]) & BN_MASK2;
+ c += (l < t);
+ r[2] = l;
+ t = a[3];
+ t = (t + c) & BN_MASK2;
+ c = (t < c);
+ l = (t + b[3]) & BN_MASK2;
+ c += (l < t);
+ r[3] = l;
+ a += 4;
+ b += 4;
+ r += 4;
+ n -= 4;
+ }
+# endif
+ while (n) {
+ t = a[0];
+ t = (t + c) & BN_MASK2;
+ c = (t < c);
+ l = (t + b[0]) & BN_MASK2;
+ c += (l < t);
+ r[0] = l;
+ a++;
+ b++;
+ r++;
+ n--;
+ }
+ return (BN_ULONG)c;
+}
+#endif /* !BN_LLONG */
+
+BN_ULONG bn_sub_words(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b,
+ int n)
+{
+ BN_ULONG t1, t2;
+ int c = 0;
+
+ assert(n >= 0);
+ if (n <= 0)
+ return (BN_ULONG)0;
+
+#ifndef OPENSSL_SMALL_FOOTPRINT
+ while (n & ~3) {
+ t1 = a[0];
+ t2 = b[0];
+ r[0] = (t1 - t2 - c) & BN_MASK2;
+ if (t1 != t2)
+ c = (t1 < t2);
+ t1 = a[1];
+ t2 = b[1];
+ r[1] = (t1 - t2 - c) & BN_MASK2;
+ if (t1 != t2)
+ c = (t1 < t2);
+ t1 = a[2];
+ t2 = b[2];
+ r[2] = (t1 - t2 - c) & BN_MASK2;
+ if (t1 != t2)
+ c = (t1 < t2);
+ t1 = a[3];
+ t2 = b[3];
+ r[3] = (t1 - t2 - c) & BN_MASK2;
+ if (t1 != t2)
+ c = (t1 < t2);
+ a += 4;
+ b += 4;
+ r += 4;
+ n -= 4;
+ }
+#endif
+ while (n) {
+ t1 = a[0];
+ t2 = b[0];
+ r[0] = (t1 - t2 - c) & BN_MASK2;
+ if (t1 != t2)
+ c = (t1 < t2);
+ a++;
+ b++;
+ r++;
+ n--;
+ }
+ return c;
+}
+
+#if defined(BN_MUL_COMBA) && !defined(OPENSSL_SMALL_FOOTPRINT)
+
+# undef bn_mul_comba8
+# undef bn_mul_comba4
+# undef bn_sqr_comba8
+# undef bn_sqr_comba4
+
+/* mul_add_c(a,b,c0,c1,c2) -- c+=a*b for three word number c=(c2,c1,c0) */
+/* mul_add_c2(a,b,c0,c1,c2) -- c+=2*a*b for three word number c=(c2,c1,c0) */
+/* sqr_add_c(a,i,c0,c1,c2) -- c+=a[i]^2 for three word number c=(c2,c1,c0) */
+/*
+ * sqr_add_c2(a,i,c0,c1,c2) -- c+=2*a[i]*a[j] for three word number
+ * c=(c2,c1,c0)
+ */
+
+# ifdef BN_LLONG
+/*
+ * Keep in mind that additions to multiplication result can not
+ * overflow, because its high half cannot be all-ones.
+ */
+# define mul_add_c(a,b,c0,c1,c2) do { \
+ BN_ULONG hi; \
+ BN_ULLONG t = (BN_ULLONG)(a)*(b); \
+ t += c0; /* no carry */ \
+ c0 = (BN_ULONG)Lw(t); \
+ hi = (BN_ULONG)Hw(t); \
+ c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
+ } while(0)
+
+# define mul_add_c2(a,b,c0,c1,c2) do { \
+ BN_ULONG hi; \
+ BN_ULLONG t = (BN_ULLONG)(a)*(b); \
+ BN_ULLONG tt = t+c0; /* no carry */ \
+ c0 = (BN_ULONG)Lw(tt); \
+ hi = (BN_ULONG)Hw(tt); \
+ c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
+ t += c0; /* no carry */ \
+ c0 = (BN_ULONG)Lw(t); \
+ hi = (BN_ULONG)Hw(t); \
+ c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
+ } while(0)
+
+# define sqr_add_c(a,i,c0,c1,c2) do { \
+ BN_ULONG hi; \
+ BN_ULLONG t = (BN_ULLONG)a[i]*a[i]; \
+ t += c0; /* no carry */ \
+ c0 = (BN_ULONG)Lw(t); \
+ hi = (BN_ULONG)Hw(t); \
+ c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
+ } while(0)
+
+# define sqr_add_c2(a,i,j,c0,c1,c2) \
+ mul_add_c2((a)[i],(a)[j],c0,c1,c2)
+
+# elif defined(BN_UMULT_LOHI)
+/*
+ * Keep in mind that additions to hi can not overflow, because
+ * the high word of a multiplication result cannot be all-ones.
+ */
+# define mul_add_c(a,b,c0,c1,c2) do { \
+ BN_ULONG ta = (a), tb = (b); \
+ BN_ULONG lo, hi; \
+ BN_UMULT_LOHI(lo,hi,ta,tb); \
+ c0 += lo; hi += (c0<lo)?1:0; \
+ c1 += hi; c2 += (c1<hi)?1:0; \
+ } while(0)
+
+# define mul_add_c2(a,b,c0,c1,c2) do { \
+ BN_ULONG ta = (a), tb = (b); \
+ BN_ULONG lo, hi, tt; \
+ BN_UMULT_LOHI(lo,hi,ta,tb); \
+ c0 += lo; tt = hi+((c0<lo)?1:0); \
+ c1 += tt; c2 += (c1<tt)?1:0; \
+ c0 += lo; hi += (c0<lo)?1:0; \
+ c1 += hi; c2 += (c1<hi)?1:0; \
+ } while(0)
+
+# define sqr_add_c(a,i,c0,c1,c2) do { \
+ BN_ULONG ta = (a)[i]; \
+ BN_ULONG lo, hi; \
+ BN_UMULT_LOHI(lo,hi,ta,ta); \
+ c0 += lo; hi += (c0<lo)?1:0; \
+ c1 += hi; c2 += (c1<hi)?1:0; \
+ } while(0)
+
+# define sqr_add_c2(a,i,j,c0,c1,c2) \
+ mul_add_c2((a)[i],(a)[j],c0,c1,c2)
+
+# elif defined(BN_UMULT_HIGH)
+/*
+ * Keep in mind that additions to hi can not overflow, because
+ * the high word of a multiplication result cannot be all-ones.
+ */
+# define mul_add_c(a,b,c0,c1,c2) do { \
+ BN_ULONG ta = (a), tb = (b); \
+ BN_ULONG lo = ta * tb; \
+ BN_ULONG hi = BN_UMULT_HIGH(ta,tb); \
+ c0 += lo; hi += (c0<lo)?1:0; \
+ c1 += hi; c2 += (c1<hi)?1:0; \
+ } while(0)
+
+# define mul_add_c2(a,b,c0,c1,c2) do { \
+ BN_ULONG ta = (a), tb = (b), tt; \
+ BN_ULONG lo = ta * tb; \
+ BN_ULONG hi = BN_UMULT_HIGH(ta,tb); \
+ c0 += lo; tt = hi + ((c0<lo)?1:0); \
+ c1 += tt; c2 += (c1<tt)?1:0; \
+ c0 += lo; hi += (c0<lo)?1:0; \
+ c1 += hi; c2 += (c1<hi)?1:0; \
+ } while(0)
+
+# define sqr_add_c(a,i,c0,c1,c2) do { \
+ BN_ULONG ta = (a)[i]; \
+ BN_ULONG lo = ta * ta; \
+ BN_ULONG hi = BN_UMULT_HIGH(ta,ta); \
+ c0 += lo; hi += (c0<lo)?1:0; \
+ c1 += hi; c2 += (c1<hi)?1:0; \
+ } while(0)
+
+# define sqr_add_c2(a,i,j,c0,c1,c2) \
+ mul_add_c2((a)[i],(a)[j],c0,c1,c2)
+
+# else /* !BN_LLONG */
+/*
+ * Keep in mind that additions to hi can not overflow, because
+ * the high word of a multiplication result cannot be all-ones.
+ */
+# define mul_add_c(a,b,c0,c1,c2) do { \
+ BN_ULONG lo = LBITS(a), hi = HBITS(a); \
+ BN_ULONG bl = LBITS(b), bh = HBITS(b); \
+ mul64(lo,hi,bl,bh); \
+ c0 = (c0+lo)&BN_MASK2; if (c0<lo) hi++; \
+ c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
+ } while(0)
+
+# define mul_add_c2(a,b,c0,c1,c2) do { \
+ BN_ULONG tt; \
+ BN_ULONG lo = LBITS(a), hi = HBITS(a); \
+ BN_ULONG bl = LBITS(b), bh = HBITS(b); \
+ mul64(lo,hi,bl,bh); \
+ tt = hi; \
+ c0 = (c0+lo)&BN_MASK2; if (c0<lo) tt++; \
+ c1 = (c1+tt)&BN_MASK2; if (c1<tt) c2++; \
+ c0 = (c0+lo)&BN_MASK2; if (c0<lo) hi++; \
+ c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
+ } while(0)
+
+# define sqr_add_c(a,i,c0,c1,c2) do { \
+ BN_ULONG lo, hi; \
+ sqr64(lo,hi,(a)[i]); \
+ c0 = (c0+lo)&BN_MASK2; if (c0<lo) hi++; \
+ c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
+ } while(0)
+
+# define sqr_add_c2(a,i,j,c0,c1,c2) \
+ mul_add_c2((a)[i],(a)[j],c0,c1,c2)
+# endif /* !BN_LLONG */
+
+void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
+{
+ BN_ULONG c1, c2, c3;
+
+ c1 = 0;
+ c2 = 0;
+ c3 = 0;
+ mul_add_c(a[0], b[0], c1, c2, c3);
+ r[0] = c1;
+ c1 = 0;
+ mul_add_c(a[0], b[1], c2, c3, c1);
+ mul_add_c(a[1], b[0], c2, c3, c1);
+ r[1] = c2;
+ c2 = 0;
+ mul_add_c(a[2], b[0], c3, c1, c2);
+ mul_add_c(a[1], b[1], c3, c1, c2);
+ mul_add_c(a[0], b[2], c3, c1, c2);
+ r[2] = c3;
+ c3 = 0;
+ mul_add_c(a[0], b[3], c1, c2, c3);
+ mul_add_c(a[1], b[2], c1, c2, c3);
+ mul_add_c(a[2], b[1], c1, c2, c3);
+ mul_add_c(a[3], b[0], c1, c2, c3);
+ r[3] = c1;
+ c1 = 0;
+ mul_add_c(a[4], b[0], c2, c3, c1);
+ mul_add_c(a[3], b[1], c2, c3, c1);
+ mul_add_c(a[2], b[2], c2, c3, c1);
+ mul_add_c(a[1], b[3], c2, c3, c1);
+ mul_add_c(a[0], b[4], c2, c3, c1);
+ r[4] = c2;
+ c2 = 0;
+ mul_add_c(a[0], b[5], c3, c1, c2);
+ mul_add_c(a[1], b[4], c3, c1, c2);
+ mul_add_c(a[2], b[3], c3, c1, c2);
+ mul_add_c(a[3], b[2], c3, c1, c2);
+ mul_add_c(a[4], b[1], c3, c1, c2);
+ mul_add_c(a[5], b[0], c3, c1, c2);
+ r[5] = c3;
+ c3 = 0;
+ mul_add_c(a[6], b[0], c1, c2, c3);
+ mul_add_c(a[5], b[1], c1, c2, c3);
+ mul_add_c(a[4], b[2], c1, c2, c3);
+ mul_add_c(a[3], b[3], c1, c2, c3);
+ mul_add_c(a[2], b[4], c1, c2, c3);
+ mul_add_c(a[1], b[5], c1, c2, c3);
+ mul_add_c(a[0], b[6], c1, c2, c3);
+ r[6] = c1;
+ c1 = 0;
+ mul_add_c(a[0], b[7], c2, c3, c1);
+ mul_add_c(a[1], b[6], c2, c3, c1);
+ mul_add_c(a[2], b[5], c2, c3, c1);
+ mul_add_c(a[3], b[4], c2, c3, c1);
+ mul_add_c(a[4], b[3], c2, c3, c1);
+ mul_add_c(a[5], b[2], c2, c3, c1);
+ mul_add_c(a[6], b[1], c2, c3, c1);
+ mul_add_c(a[7], b[0], c2, c3, c1);
+ r[7] = c2;
+ c2 = 0;
+ mul_add_c(a[7], b[1], c3, c1, c2);
+ mul_add_c(a[6], b[2], c3, c1, c2);
+ mul_add_c(a[5], b[3], c3, c1, c2);
+ mul_add_c(a[4], b[4], c3, c1, c2);
+ mul_add_c(a[3], b[5], c3, c1, c2);
+ mul_add_c(a[2], b[6], c3, c1, c2);
+ mul_add_c(a[1], b[7], c3, c1, c2);
+ r[8] = c3;
+ c3 = 0;
+ mul_add_c(a[2], b[7], c1, c2, c3);
+ mul_add_c(a[3], b[6], c1, c2, c3);
+ mul_add_c(a[4], b[5], c1, c2, c3);
+ mul_add_c(a[5], b[4], c1, c2, c3);
+ mul_add_c(a[6], b[3], c1, c2, c3);
+ mul_add_c(a[7], b[2], c1, c2, c3);
+ r[9] = c1;
+ c1 = 0;
+ mul_add_c(a[7], b[3], c2, c3, c1);
+ mul_add_c(a[6], b[4], c2, c3, c1);
+ mul_add_c(a[5], b[5], c2, c3, c1);
+ mul_add_c(a[4], b[6], c2, c3, c1);
+ mul_add_c(a[3], b[7], c2, c3, c1);
+ r[10] = c2;
+ c2 = 0;
+ mul_add_c(a[4], b[7], c3, c1, c2);
+ mul_add_c(a[5], b[6], c3, c1, c2);
+ mul_add_c(a[6], b[5], c3, c1, c2);
+ mul_add_c(a[7], b[4], c3, c1, c2);
+ r[11] = c3;
+ c3 = 0;
+ mul_add_c(a[7], b[5], c1, c2, c3);
+ mul_add_c(a[6], b[6], c1, c2, c3);
+ mul_add_c(a[5], b[7], c1, c2, c3);
+ r[12] = c1;
+ c1 = 0;
+ mul_add_c(a[6], b[7], c2, c3, c1);
+ mul_add_c(a[7], b[6], c2, c3, c1);
+ r[13] = c2;
+ c2 = 0;
+ mul_add_c(a[7], b[7], c3, c1, c2);
+ r[14] = c3;
+ r[15] = c1;
+}
+
+void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
+{
+ BN_ULONG c1, c2, c3;
+
+ c1 = 0;
+ c2 = 0;
+ c3 = 0;
+ mul_add_c(a[0], b[0], c1, c2, c3);
+ r[0] = c1;
+ c1 = 0;
+ mul_add_c(a[0], b[1], c2, c3, c1);
+ mul_add_c(a[1], b[0], c2, c3, c1);
+ r[1] = c2;
+ c2 = 0;
+ mul_add_c(a[2], b[0], c3, c1, c2);
+ mul_add_c(a[1], b[1], c3, c1, c2);
+ mul_add_c(a[0], b[2], c3, c1, c2);
+ r[2] = c3;
+ c3 = 0;
+ mul_add_c(a[0], b[3], c1, c2, c3);
+ mul_add_c(a[1], b[2], c1, c2, c3);
+ mul_add_c(a[2], b[1], c1, c2, c3);
+ mul_add_c(a[3], b[0], c1, c2, c3);
+ r[3] = c1;
+ c1 = 0;
+ mul_add_c(a[3], b[1], c2, c3, c1);
+ mul_add_c(a[2], b[2], c2, c3, c1);
+ mul_add_c(a[1], b[3], c2, c3, c1);
+ r[4] = c2;
+ c2 = 0;
+ mul_add_c(a[2], b[3], c3, c1, c2);
+ mul_add_c(a[3], b[2], c3, c1, c2);
+ r[5] = c3;
+ c3 = 0;
+ mul_add_c(a[3], b[3], c1, c2, c3);
+ r[6] = c1;
+ r[7] = c2;
+}
+
+void bn_sqr_comba8(BN_ULONG *r, const BN_ULONG *a)
+{
+ BN_ULONG c1, c2, c3;
+
+ c1 = 0;
+ c2 = 0;
+ c3 = 0;
+ sqr_add_c(a, 0, c1, c2, c3);
+ r[0] = c1;
+ c1 = 0;
+ sqr_add_c2(a, 1, 0, c2, c3, c1);
+ r[1] = c2;
+ c2 = 0;
+ sqr_add_c(a, 1, c3, c1, c2);
+ sqr_add_c2(a, 2, 0, c3, c1, c2);
+ r[2] = c3;
+ c3 = 0;
+ sqr_add_c2(a, 3, 0, c1, c2, c3);
+ sqr_add_c2(a, 2, 1, c1, c2, c3);
+ r[3] = c1;
+ c1 = 0;
+ sqr_add_c(a, 2, c2, c3, c1);
+ sqr_add_c2(a, 3, 1, c2, c3, c1);
+ sqr_add_c2(a, 4, 0, c2, c3, c1);
+ r[4] = c2;
+ c2 = 0;
+ sqr_add_c2(a, 5, 0, c3, c1, c2);
+ sqr_add_c2(a, 4, 1, c3, c1, c2);
+ sqr_add_c2(a, 3, 2, c3, c1, c2);
+ r[5] = c3;
+ c3 = 0;
+ sqr_add_c(a, 3, c1, c2, c3);
+ sqr_add_c2(a, 4, 2, c1, c2, c3);
+ sqr_add_c2(a, 5, 1, c1, c2, c3);
+ sqr_add_c2(a, 6, 0, c1, c2, c3);
+ r[6] = c1;
+ c1 = 0;
+ sqr_add_c2(a, 7, 0, c2, c3, c1);
+ sqr_add_c2(a, 6, 1, c2, c3, c1);
+ sqr_add_c2(a, 5, 2, c2, c3, c1);
+ sqr_add_c2(a, 4, 3, c2, c3, c1);
+ r[7] = c2;
+ c2 = 0;
+ sqr_add_c(a, 4, c3, c1, c2);
+ sqr_add_c2(a, 5, 3, c3, c1, c2);
+ sqr_add_c2(a, 6, 2, c3, c1, c2);
+ sqr_add_c2(a, 7, 1, c3, c1, c2);
+ r[8] = c3;
+ c3 = 0;
+ sqr_add_c2(a, 7, 2, c1, c2, c3);
+ sqr_add_c2(a, 6, 3, c1, c2, c3);
+ sqr_add_c2(a, 5, 4, c1, c2, c3);
+ r[9] = c1;
+ c1 = 0;
+ sqr_add_c(a, 5, c2, c3, c1);
+ sqr_add_c2(a, 6, 4, c2, c3, c1);
+ sqr_add_c2(a, 7, 3, c2, c3, c1);
+ r[10] = c2;
+ c2 = 0;
+ sqr_add_c2(a, 7, 4, c3, c1, c2);
+ sqr_add_c2(a, 6, 5, c3, c1, c2);
+ r[11] = c3;
+ c3 = 0;
+ sqr_add_c(a, 6, c1, c2, c3);
+ sqr_add_c2(a, 7, 5, c1, c2, c3);
+ r[12] = c1;
+ c1 = 0;
+ sqr_add_c2(a, 7, 6, c2, c3, c1);
+ r[13] = c2;
+ c2 = 0;
+ sqr_add_c(a, 7, c3, c1, c2);
+ r[14] = c3;
+ r[15] = c1;
+}
+
+void bn_sqr_comba4(BN_ULONG *r, const BN_ULONG *a)
+{
+ BN_ULONG c1, c2, c3;
+
+ c1 = 0;
+ c2 = 0;
+ c3 = 0;
+ sqr_add_c(a, 0, c1, c2, c3);
+ r[0] = c1;
+ c1 = 0;
+ sqr_add_c2(a, 1, 0, c2, c3, c1);
+ r[1] = c2;
+ c2 = 0;
+ sqr_add_c(a, 1, c3, c1, c2);
+ sqr_add_c2(a, 2, 0, c3, c1, c2);
+ r[2] = c3;
+ c3 = 0;
+ sqr_add_c2(a, 3, 0, c1, c2, c3);
+ sqr_add_c2(a, 2, 1, c1, c2, c3);
+ r[3] = c1;
+ c1 = 0;
+ sqr_add_c(a, 2, c2, c3, c1);
+ sqr_add_c2(a, 3, 1, c2, c3, c1);
+ r[4] = c2;
+ c2 = 0;
+ sqr_add_c2(a, 3, 2, c3, c1, c2);
+ r[5] = c3;
+ c3 = 0;
+ sqr_add_c(a, 3, c1, c2, c3);
+ r[6] = c1;
+ r[7] = c2;
+}
+
+# ifdef OPENSSL_NO_ASM
+# ifdef OPENSSL_BN_ASM_MONT
+# include <alloca.h>
+/*
+ * This is essentially reference implementation, which may or may not
+ * result in performance improvement. E.g. on IA-32 this routine was
+ * observed to give 40% faster rsa1024 private key operations and 10%
+ * faster rsa4096 ones, while on AMD64 it improves rsa1024 sign only
+ * by 10% and *worsens* rsa4096 sign by 15%. Once again, it's a
+ * reference implementation, one to be used as starting point for
+ * platform-specific assembler. Mentioned numbers apply to compiler
+ * generated code compiled with and without -DOPENSSL_BN_ASM_MONT and
+ * can vary not only from platform to platform, but even for compiler
+ * versions. Assembler vs. assembler improvement coefficients can
+ * [and are known to] differ and are to be documented elsewhere.
+ */
+int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
+ const BN_ULONG *np, const BN_ULONG *n0p, int num)
+{
+ BN_ULONG c0, c1, ml, *tp, n0;
+# ifdef mul64
+ BN_ULONG mh;
+# endif
+ volatile BN_ULONG *vp;
+ int i = 0, j;
+
+# if 0 /* template for platform-specific
+ * implementation */
+ if (ap == bp)
+ return bn_sqr_mont(rp, ap, np, n0p, num);
+# endif
+ vp = tp = alloca((num + 2) * sizeof(BN_ULONG));
+
+ n0 = *n0p;
+
+ c0 = 0;
+ ml = bp[0];
+# ifdef mul64
+ mh = HBITS(ml);
+ ml = LBITS(ml);
+ for (j = 0; j < num; ++j)
+ mul(tp[j], ap[j], ml, mh, c0);
+# else
+ for (j = 0; j < num; ++j)
+ mul(tp[j], ap[j], ml, c0);
+# endif
+
+ tp[num] = c0;
+ tp[num + 1] = 0;
+ goto enter;
+
+ for (i = 0; i < num; i++) {
+ c0 = 0;
+ ml = bp[i];
+# ifdef mul64
+ mh = HBITS(ml);
+ ml = LBITS(ml);
+ for (j = 0; j < num; ++j)
+ mul_add(tp[j], ap[j], ml, mh, c0);
+# else
+ for (j = 0; j < num; ++j)
+ mul_add(tp[j], ap[j], ml, c0);
+# endif
+ c1 = (tp[num] + c0) & BN_MASK2;
+ tp[num] = c1;
+ tp[num + 1] = (c1 < c0 ? 1 : 0);
+ enter:
+ c1 = tp[0];
+ ml = (c1 * n0) & BN_MASK2;
+ c0 = 0;
+# ifdef mul64
+ mh = HBITS(ml);
+ ml = LBITS(ml);
+ mul_add(c1, np[0], ml, mh, c0);
+# else
+ mul_add(c1, ml, np[0], c0);
+# endif
+ for (j = 1; j < num; j++) {
+ c1 = tp[j];
+# ifdef mul64
+ mul_add(c1, np[j], ml, mh, c0);
+# else
+ mul_add(c1, ml, np[j], c0);
+# endif
+ tp[j - 1] = c1 & BN_MASK2;
+ }
+ c1 = (tp[num] + c0) & BN_MASK2;
+ tp[num - 1] = c1;
+ tp[num] = tp[num + 1] + (c1 < c0 ? 1 : 0);
+ }
+
+ if (tp[num] != 0 || tp[num - 1] >= np[num - 1]) {
+ c0 = bn_sub_words(rp, tp, np, num);
+ if (tp[num] != 0 || c0 == 0) {
+ for (i = 0; i < num + 2; i++)
+ vp[i] = 0;
+ return 1;
+ }
+ }
+ for (i = 0; i < num; i++)
+ rp[i] = tp[i], vp[i] = 0;
+ vp[num] = 0;
+ vp[num + 1] = 0;
+ return 1;
+}
+# else
+/*
+ * Return value of 0 indicates that multiplication/convolution was not
+ * performed to signal the caller to fall down to alternative/original
+ * code-path.
+ */
+int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
+ const BN_ULONG *np, const BN_ULONG *n0, int num)
+{
+ return 0;
+}
+# endif /* OPENSSL_BN_ASM_MONT */
+# endif
+
+#else /* !BN_MUL_COMBA */
+
+/* hmm... is it faster just to do a multiply? */
+# undef bn_sqr_comba4
+# undef bn_sqr_comba8
+void bn_sqr_comba4(BN_ULONG *r, const BN_ULONG *a)
+{
+ BN_ULONG t[8];
+ bn_sqr_normal(r, a, 4, t);
+}
+
+void bn_sqr_comba8(BN_ULONG *r, const BN_ULONG *a)
+{
+ BN_ULONG t[16];
+ bn_sqr_normal(r, a, 8, t);
+}
+
+void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
+{
+ r[4] = bn_mul_words(&(r[0]), a, 4, b[0]);
+ r[5] = bn_mul_add_words(&(r[1]), a, 4, b[1]);
+ r[6] = bn_mul_add_words(&(r[2]), a, 4, b[2]);
+ r[7] = bn_mul_add_words(&(r[3]), a, 4, b[3]);
+}
+
+void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
+{
+ r[8] = bn_mul_words(&(r[0]), a, 8, b[0]);
+ r[9] = bn_mul_add_words(&(r[1]), a, 8, b[1]);
+ r[10] = bn_mul_add_words(&(r[2]), a, 8, b[2]);
+ r[11] = bn_mul_add_words(&(r[3]), a, 8, b[3]);
+ r[12] = bn_mul_add_words(&(r[4]), a, 8, b[4]);
+ r[13] = bn_mul_add_words(&(r[5]), a, 8, b[5]);
+ r[14] = bn_mul_add_words(&(r[6]), a, 8, b[6]);
+ r[15] = bn_mul_add_words(&(r[7]), a, 8, b[7]);
+}
+
+# ifdef OPENSSL_NO_ASM
+# ifdef OPENSSL_BN_ASM_MONT
+# include <alloca.h>
+int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
+ const BN_ULONG *np, const BN_ULONG *n0p, int num)
+{
+ BN_ULONG c0, c1, *tp, n0 = *n0p;
+ volatile BN_ULONG *vp;
+ int i = 0, j;
+
+ vp = tp = alloca((num + 2) * sizeof(BN_ULONG));
+
+ for (i = 0; i <= num; i++)
+ tp[i] = 0;
+
+ for (i = 0; i < num; i++) {
+ c0 = bn_mul_add_words(tp, ap, num, bp[i]);
+ c1 = (tp[num] + c0) & BN_MASK2;
+ tp[num] = c1;
+ tp[num + 1] = (c1 < c0 ? 1 : 0);
+
+ c0 = bn_mul_add_words(tp, np, num, tp[0] * n0);
+ c1 = (tp[num] + c0) & BN_MASK2;
+ tp[num] = c1;
+ tp[num + 1] += (c1 < c0 ? 1 : 0);
+ for (j = 0; j <= num; j++)
+ tp[j] = tp[j + 1];
+ }
+
+ if (tp[num] != 0 || tp[num - 1] >= np[num - 1]) {
+ c0 = bn_sub_words(rp, tp, np, num);
+ if (tp[num] != 0 || c0 == 0) {
+ for (i = 0; i < num + 2; i++)
+ vp[i] = 0;
+ return 1;
+ }
+ }
+ for (i = 0; i < num; i++)
+ rp[i] = tp[i], vp[i] = 0;
+ vp[num] = 0;
+ vp[num + 1] = 0;
+ return 1;
+}
+# else
+int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
+ const BN_ULONG *np, const BN_ULONG *n0, int num)
+{
+ return 0;
+}
+# endif /* OPENSSL_BN_ASM_MONT */
+# endif
+
+#endif /* !BN_MUL_COMBA */
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_blind.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_blind.c
new file mode 100644
index 0000000..76fc7eb
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_blind.c
@@ -0,0 +1,312 @@
+/*
+ * Copyright 1998-2018 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include <openssl/opensslconf.h>
+#include "internal/cryptlib.h"
+#include "bn_local.h"
+
+#define BN_BLINDING_COUNTER 32
+
+struct bn_blinding_st {
+ BIGNUM *A;
+ BIGNUM *Ai;
+ BIGNUM *e;
+ BIGNUM *mod; /* just a reference */
+ CRYPTO_THREAD_ID tid;
+ int counter;
+ unsigned long flags;
+ BN_MONT_CTX *m_ctx;
+ int (*bn_mod_exp) (BIGNUM *r, const BIGNUM *a, const BIGNUM *p,
+ const BIGNUM *m, BN_CTX *ctx, BN_MONT_CTX *m_ctx);
+ CRYPTO_RWLOCK *lock;
+};
+
+BN_BLINDING *BN_BLINDING_new(const BIGNUM *A, const BIGNUM *Ai, BIGNUM *mod)
+{
+ BN_BLINDING *ret = NULL;
+
+ bn_check_top(mod);
+
+ if ((ret = OPENSSL_zalloc(sizeof(*ret))) == NULL) {
+ BNerr(BN_F_BN_BLINDING_NEW, ERR_R_MALLOC_FAILURE);
+ return NULL;
+ }
+
+ ret->lock = CRYPTO_THREAD_lock_new();
+ if (ret->lock == NULL) {
+ BNerr(BN_F_BN_BLINDING_NEW, ERR_R_MALLOC_FAILURE);
+ OPENSSL_free(ret);
+ return NULL;
+ }
+
+ BN_BLINDING_set_current_thread(ret);
+
+ if (A != NULL) {
+ if ((ret->A = BN_dup(A)) == NULL)
+ goto err;
+ }
+
+ if (Ai != NULL) {
+ if ((ret->Ai = BN_dup(Ai)) == NULL)
+ goto err;
+ }
+
+ /* save a copy of mod in the BN_BLINDING structure */
+ if ((ret->mod = BN_dup(mod)) == NULL)
+ goto err;
+
+ if (BN_get_flags(mod, BN_FLG_CONSTTIME) != 0)
+ BN_set_flags(ret->mod, BN_FLG_CONSTTIME);
+
+ /*
+ * Set the counter to the special value -1 to indicate that this is
+ * never-used fresh blinding that does not need updating before first
+ * use.
+ */
+ ret->counter = -1;
+
+ return ret;
+
+ err:
+ BN_BLINDING_free(ret);
+ return NULL;
+}
+
+void BN_BLINDING_free(BN_BLINDING *r)
+{
+ if (r == NULL)
+ return;
+ BN_free(r->A);
+ BN_free(r->Ai);
+ BN_free(r->e);
+ BN_free(r->mod);
+ CRYPTO_THREAD_lock_free(r->lock);
+ OPENSSL_free(r);
+}
+
+int BN_BLINDING_update(BN_BLINDING *b, BN_CTX *ctx)
+{
+ int ret = 0;
+
+ if ((b->A == NULL) || (b->Ai == NULL)) {
+ BNerr(BN_F_BN_BLINDING_UPDATE, BN_R_NOT_INITIALIZED);
+ goto err;
+ }
+
+ if (b->counter == -1)
+ b->counter = 0;
+
+ if (++b->counter == BN_BLINDING_COUNTER && b->e != NULL &&
+ !(b->flags & BN_BLINDING_NO_RECREATE)) {
+ /* re-create blinding parameters */
+ if (!BN_BLINDING_create_param(b, NULL, NULL, ctx, NULL, NULL))
+ goto err;
+ } else if (!(b->flags & BN_BLINDING_NO_UPDATE)) {
+ if (b->m_ctx != NULL) {
+ if (!bn_mul_mont_fixed_top(b->Ai, b->Ai, b->Ai, b->m_ctx, ctx)
+ || !bn_mul_mont_fixed_top(b->A, b->A, b->A, b->m_ctx, ctx))
+ goto err;
+ } else {
+ if (!BN_mod_mul(b->Ai, b->Ai, b->Ai, b->mod, ctx)
+ || !BN_mod_mul(b->A, b->A, b->A, b->mod, ctx))
+ goto err;
+ }
+ }
+
+ ret = 1;
+ err:
+ if (b->counter == BN_BLINDING_COUNTER)
+ b->counter = 0;
+ return ret;
+}
+
+int BN_BLINDING_convert(BIGNUM *n, BN_BLINDING *b, BN_CTX *ctx)
+{
+ return BN_BLINDING_convert_ex(n, NULL, b, ctx);
+}
+
+int BN_BLINDING_convert_ex(BIGNUM *n, BIGNUM *r, BN_BLINDING *b, BN_CTX *ctx)
+{
+ int ret = 1;
+
+ bn_check_top(n);
+
+ if ((b->A == NULL) || (b->Ai == NULL)) {
+ BNerr(BN_F_BN_BLINDING_CONVERT_EX, BN_R_NOT_INITIALIZED);
+ return 0;
+ }
+
+ if (b->counter == -1)
+ /* Fresh blinding, doesn't need updating. */
+ b->counter = 0;
+ else if (!BN_BLINDING_update(b, ctx))
+ return 0;
+
+ if (r != NULL && (BN_copy(r, b->Ai) == NULL))
+ return 0;
+
+ if (b->m_ctx != NULL)
+ ret = BN_mod_mul_montgomery(n, n, b->A, b->m_ctx, ctx);
+ else
+ ret = BN_mod_mul(n, n, b->A, b->mod, ctx);
+
+ return ret;
+}
+
+int BN_BLINDING_invert(BIGNUM *n, BN_BLINDING *b, BN_CTX *ctx)
+{
+ return BN_BLINDING_invert_ex(n, NULL, b, ctx);
+}
+
+int BN_BLINDING_invert_ex(BIGNUM *n, const BIGNUM *r, BN_BLINDING *b,
+ BN_CTX *ctx)
+{
+ int ret;
+
+ bn_check_top(n);
+
+ if (r == NULL && (r = b->Ai) == NULL) {
+ BNerr(BN_F_BN_BLINDING_INVERT_EX, BN_R_NOT_INITIALIZED);
+ return 0;
+ }
+
+ if (b->m_ctx != NULL) {
+ /* ensure that BN_mod_mul_montgomery takes pre-defined path */
+ if (n->dmax >= r->top) {
+ size_t i, rtop = r->top, ntop = n->top;
+ BN_ULONG mask;
+
+ for (i = 0; i < rtop; i++) {
+ mask = (BN_ULONG)0 - ((i - ntop) >> (8 * sizeof(i) - 1));
+ n->d[i] &= mask;
+ }
+ mask = (BN_ULONG)0 - ((rtop - ntop) >> (8 * sizeof(ntop) - 1));
+ /* always true, if (rtop >= ntop) n->top = r->top; */
+ n->top = (int)(rtop & ~mask) | (ntop & mask);
+ n->flags |= (BN_FLG_FIXED_TOP & ~mask);
+ }
+ ret = BN_mod_mul_montgomery(n, n, r, b->m_ctx, ctx);
+ } else {
+ ret = BN_mod_mul(n, n, r, b->mod, ctx);
+ }
+
+ bn_check_top(n);
+ return ret;
+}
+
+int BN_BLINDING_is_current_thread(BN_BLINDING *b)
+{
+ return CRYPTO_THREAD_compare_id(CRYPTO_THREAD_get_current_id(), b->tid);
+}
+
+void BN_BLINDING_set_current_thread(BN_BLINDING *b)
+{
+ b->tid = CRYPTO_THREAD_get_current_id();
+}
+
+int BN_BLINDING_lock(BN_BLINDING *b)
+{
+ return CRYPTO_THREAD_write_lock(b->lock);
+}
+
+int BN_BLINDING_unlock(BN_BLINDING *b)
+{
+ return CRYPTO_THREAD_unlock(b->lock);
+}
+
+unsigned long BN_BLINDING_get_flags(const BN_BLINDING *b)
+{
+ return b->flags;
+}
+
+void BN_BLINDING_set_flags(BN_BLINDING *b, unsigned long flags)
+{
+ b->flags = flags;
+}
+
+BN_BLINDING *BN_BLINDING_create_param(BN_BLINDING *b,
+ const BIGNUM *e, BIGNUM *m, BN_CTX *ctx,
+ int (*bn_mod_exp) (BIGNUM *r,
+ const BIGNUM *a,
+ const BIGNUM *p,
+ const BIGNUM *m,
+ BN_CTX *ctx,
+ BN_MONT_CTX *m_ctx),
+ BN_MONT_CTX *m_ctx)
+{
+ int retry_counter = 32;
+ BN_BLINDING *ret = NULL;
+
+ if (b == NULL)
+ ret = BN_BLINDING_new(NULL, NULL, m);
+ else
+ ret = b;
+
+ if (ret == NULL)
+ goto err;
+
+ if (ret->A == NULL && (ret->A = BN_new()) == NULL)
+ goto err;
+ if (ret->Ai == NULL && (ret->Ai = BN_new()) == NULL)
+ goto err;
+
+ if (e != NULL) {
+ BN_free(ret->e);
+ ret->e = BN_dup(e);
+ }
+ if (ret->e == NULL)
+ goto err;
+
+ if (bn_mod_exp != NULL)
+ ret->bn_mod_exp = bn_mod_exp;
+ if (m_ctx != NULL)
+ ret->m_ctx = m_ctx;
+
+ do {
+ int rv;
+ if (!BN_priv_rand_range(ret->A, ret->mod))
+ goto err;
+ if (int_bn_mod_inverse(ret->Ai, ret->A, ret->mod, ctx, &rv))
+ break;
+
+ /*
+ * this should almost never happen for good RSA keys
+ */
+ if (!rv)
+ goto err;
+
+ if (retry_counter-- == 0) {
+ BNerr(BN_F_BN_BLINDING_CREATE_PARAM, BN_R_TOO_MANY_ITERATIONS);
+ goto err;
+ }
+ } while (1);
+
+ if (ret->bn_mod_exp != NULL && ret->m_ctx != NULL) {
+ if (!ret->bn_mod_exp(ret->A, ret->A, ret->e, ret->mod, ctx, ret->m_ctx))
+ goto err;
+ } else {
+ if (!BN_mod_exp(ret->A, ret->A, ret->e, ret->mod, ctx))
+ goto err;
+ }
+
+ if (ret->m_ctx != NULL) {
+ if (!bn_to_mont_fixed_top(ret->Ai, ret->Ai, ret->m_ctx, ctx)
+ || !bn_to_mont_fixed_top(ret->A, ret->A, ret->m_ctx, ctx))
+ goto err;
+ }
+
+ return ret;
+ err:
+ if (b == NULL) {
+ BN_BLINDING_free(ret);
+ ret = NULL;
+ }
+
+ return ret;
+}
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_const.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_const.c
new file mode 100644
index 0000000..39dd612
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_const.c
@@ -0,0 +1,553 @@
+/*
+ * Copyright 2005-2016 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include <openssl/bn.h>
+
+/*-
+ * "First Oakley Default Group" from RFC2409, section 6.1.
+ *
+ * The prime is: 2^768 - 2 ^704 - 1 + 2^64 * { [2^638 pi] + 149686 }
+ *
+ * RFC2409 specifies a generator of 2.
+ * RFC2412 specifies a generator of of 22.
+ */
+
+BIGNUM *BN_get_rfc2409_prime_768(BIGNUM *bn)
+{
+ static const unsigned char RFC2409_PRIME_768[] = {
+ 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
+ 0xC9, 0x0F, 0xDA, 0xA2, 0x21, 0x68, 0xC2, 0x34,
+ 0xC4, 0xC6, 0x62, 0x8B, 0x80, 0xDC, 0x1C, 0xD1,
+ 0x29, 0x02, 0x4E, 0x08, 0x8A, 0x67, 0xCC, 0x74,
+ 0x02, 0x0B, 0xBE, 0xA6, 0x3B, 0x13, 0x9B, 0x22,
+ 0x51, 0x4A, 0x08, 0x79, 0x8E, 0x34, 0x04, 0xDD,
+ 0xEF, 0x95, 0x19, 0xB3, 0xCD, 0x3A, 0x43, 0x1B,
+ 0x30, 0x2B, 0x0A, 0x6D, 0xF2, 0x5F, 0x14, 0x37,
+ 0x4F, 0xE1, 0x35, 0x6D, 0x6D, 0x51, 0xC2, 0x45,
+ 0xE4, 0x85, 0xB5, 0x76, 0x62, 0x5E, 0x7E, 0xC6,
+ 0xF4, 0x4C, 0x42, 0xE9, 0xA6, 0x3A, 0x36, 0x20,
+ 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
+ };
+ return BN_bin2bn(RFC2409_PRIME_768, sizeof(RFC2409_PRIME_768), bn);
+}
+
+/*-
+ * "Second Oakley Default Group" from RFC2409, section 6.2.
+ *
+ * The prime is: 2^1024 - 2^960 - 1 + 2^64 * { [2^894 pi] + 129093 }.
+ *
+ * RFC2409 specifies a generator of 2.
+ * RFC2412 specifies a generator of 22.
+ */
+
+BIGNUM *BN_get_rfc2409_prime_1024(BIGNUM *bn)
+{
+ static const unsigned char RFC2409_PRIME_1024[] = {
+ 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
+ 0xC9, 0x0F, 0xDA, 0xA2, 0x21, 0x68, 0xC2, 0x34,
+ 0xC4, 0xC6, 0x62, 0x8B, 0x80, 0xDC, 0x1C, 0xD1,
+ 0x29, 0x02, 0x4E, 0x08, 0x8A, 0x67, 0xCC, 0x74,
+ 0x02, 0x0B, 0xBE, 0xA6, 0x3B, 0x13, 0x9B, 0x22,
+ 0x51, 0x4A, 0x08, 0x79, 0x8E, 0x34, 0x04, 0xDD,
+ 0xEF, 0x95, 0x19, 0xB3, 0xCD, 0x3A, 0x43, 0x1B,
+ 0x30, 0x2B, 0x0A, 0x6D, 0xF2, 0x5F, 0x14, 0x37,
+ 0x4F, 0xE1, 0x35, 0x6D, 0x6D, 0x51, 0xC2, 0x45,
+ 0xE4, 0x85, 0xB5, 0x76, 0x62, 0x5E, 0x7E, 0xC6,
+ 0xF4, 0x4C, 0x42, 0xE9, 0xA6, 0x37, 0xED, 0x6B,
+ 0x0B, 0xFF, 0x5C, 0xB6, 0xF4, 0x06, 0xB7, 0xED,
+ 0xEE, 0x38, 0x6B, 0xFB, 0x5A, 0x89, 0x9F, 0xA5,
+ 0xAE, 0x9F, 0x24, 0x11, 0x7C, 0x4B, 0x1F, 0xE6,
+ 0x49, 0x28, 0x66, 0x51, 0xEC, 0xE6, 0x53, 0x81,
+ 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
+ };
+ return BN_bin2bn(RFC2409_PRIME_1024, sizeof(RFC2409_PRIME_1024), bn);
+}
+
+/*-
+ * "1536-bit MODP Group" from RFC3526, Section 2.
+ *
+ * The prime is: 2^1536 - 2^1472 - 1 + 2^64 * { [2^1406 pi] + 741804 }
+ *
+ * RFC3526 specifies a generator of 2.
+ * RFC2312 specifies a generator of 22.
+ */
+
+BIGNUM *BN_get_rfc3526_prime_1536(BIGNUM *bn)
+{
+ static const unsigned char RFC3526_PRIME_1536[] = {
+ 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
+ 0xC9, 0x0F, 0xDA, 0xA2, 0x21, 0x68, 0xC2, 0x34,
+ 0xC4, 0xC6, 0x62, 0x8B, 0x80, 0xDC, 0x1C, 0xD1,
+ 0x29, 0x02, 0x4E, 0x08, 0x8A, 0x67, 0xCC, 0x74,
+ 0x02, 0x0B, 0xBE, 0xA6, 0x3B, 0x13, 0x9B, 0x22,
+ 0x51, 0x4A, 0x08, 0x79, 0x8E, 0x34, 0x04, 0xDD,
+ 0xEF, 0x95, 0x19, 0xB3, 0xCD, 0x3A, 0x43, 0x1B,
+ 0x30, 0x2B, 0x0A, 0x6D, 0xF2, 0x5F, 0x14, 0x37,
+ 0x4F, 0xE1, 0x35, 0x6D, 0x6D, 0x51, 0xC2, 0x45,
+ 0xE4, 0x85, 0xB5, 0x76, 0x62, 0x5E, 0x7E, 0xC6,
+ 0xF4, 0x4C, 0x42, 0xE9, 0xA6, 0x37, 0xED, 0x6B,
+ 0x0B, 0xFF, 0x5C, 0xB6, 0xF4, 0x06, 0xB7, 0xED,
+ 0xEE, 0x38, 0x6B, 0xFB, 0x5A, 0x89, 0x9F, 0xA5,
+ 0xAE, 0x9F, 0x24, 0x11, 0x7C, 0x4B, 0x1F, 0xE6,
+ 0x49, 0x28, 0x66, 0x51, 0xEC, 0xE4, 0x5B, 0x3D,
+ 0xC2, 0x00, 0x7C, 0xB8, 0xA1, 0x63, 0xBF, 0x05,
+ 0x98, 0xDA, 0x48, 0x36, 0x1C, 0x55, 0xD3, 0x9A,
+ 0x69, 0x16, 0x3F, 0xA8, 0xFD, 0x24, 0xCF, 0x5F,
+ 0x83, 0x65, 0x5D, 0x23, 0xDC, 0xA3, 0xAD, 0x96,
+ 0x1C, 0x62, 0xF3, 0x56, 0x20, 0x85, 0x52, 0xBB,
+ 0x9E, 0xD5, 0x29, 0x07, 0x70, 0x96, 0x96, 0x6D,
+ 0x67, 0x0C, 0x35, 0x4E, 0x4A, 0xBC, 0x98, 0x04,
+ 0xF1, 0x74, 0x6C, 0x08, 0xCA, 0x23, 0x73, 0x27,
+ 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
+ };
+ return BN_bin2bn(RFC3526_PRIME_1536, sizeof(RFC3526_PRIME_1536), bn);
+}
+
+/*-
+ * "2048-bit MODP Group" from RFC3526, Section 3.
+ *
+ * The prime is: 2^2048 - 2^1984 - 1 + 2^64 * { [2^1918 pi] + 124476 }
+ *
+ * RFC3526 specifies a generator of 2.
+ */
+
+BIGNUM *BN_get_rfc3526_prime_2048(BIGNUM *bn)
+{
+ static const unsigned char RFC3526_PRIME_2048[] = {
+ 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
+ 0xC9, 0x0F, 0xDA, 0xA2, 0x21, 0x68, 0xC2, 0x34,
+ 0xC4, 0xC6, 0x62, 0x8B, 0x80, 0xDC, 0x1C, 0xD1,
+ 0x29, 0x02, 0x4E, 0x08, 0x8A, 0x67, 0xCC, 0x74,
+ 0x02, 0x0B, 0xBE, 0xA6, 0x3B, 0x13, 0x9B, 0x22,
+ 0x51, 0x4A, 0x08, 0x79, 0x8E, 0x34, 0x04, 0xDD,
+ 0xEF, 0x95, 0x19, 0xB3, 0xCD, 0x3A, 0x43, 0x1B,
+ 0x30, 0x2B, 0x0A, 0x6D, 0xF2, 0x5F, 0x14, 0x37,
+ 0x4F, 0xE1, 0x35, 0x6D, 0x6D, 0x51, 0xC2, 0x45,
+ 0xE4, 0x85, 0xB5, 0x76, 0x62, 0x5E, 0x7E, 0xC6,
+ 0xF4, 0x4C, 0x42, 0xE9, 0xA6, 0x37, 0xED, 0x6B,
+ 0x0B, 0xFF, 0x5C, 0xB6, 0xF4, 0x06, 0xB7, 0xED,
+ 0xEE, 0x38, 0x6B, 0xFB, 0x5A, 0x89, 0x9F, 0xA5,
+ 0xAE, 0x9F, 0x24, 0x11, 0x7C, 0x4B, 0x1F, 0xE6,
+ 0x49, 0x28, 0x66, 0x51, 0xEC, 0xE4, 0x5B, 0x3D,
+ 0xC2, 0x00, 0x7C, 0xB8, 0xA1, 0x63, 0xBF, 0x05,
+ 0x98, 0xDA, 0x48, 0x36, 0x1C, 0x55, 0xD3, 0x9A,
+ 0x69, 0x16, 0x3F, 0xA8, 0xFD, 0x24, 0xCF, 0x5F,
+ 0x83, 0x65, 0x5D, 0x23, 0xDC, 0xA3, 0xAD, 0x96,
+ 0x1C, 0x62, 0xF3, 0x56, 0x20, 0x85, 0x52, 0xBB,
+ 0x9E, 0xD5, 0x29, 0x07, 0x70, 0x96, 0x96, 0x6D,
+ 0x67, 0x0C, 0x35, 0x4E, 0x4A, 0xBC, 0x98, 0x04,
+ 0xF1, 0x74, 0x6C, 0x08, 0xCA, 0x18, 0x21, 0x7C,
+ 0x32, 0x90, 0x5E, 0x46, 0x2E, 0x36, 0xCE, 0x3B,
+ 0xE3, 0x9E, 0x77, 0x2C, 0x18, 0x0E, 0x86, 0x03,
+ 0x9B, 0x27, 0x83, 0xA2, 0xEC, 0x07, 0xA2, 0x8F,
+ 0xB5, 0xC5, 0x5D, 0xF0, 0x6F, 0x4C, 0x52, 0xC9,
+ 0xDE, 0x2B, 0xCB, 0xF6, 0x95, 0x58, 0x17, 0x18,
+ 0x39, 0x95, 0x49, 0x7C, 0xEA, 0x95, 0x6A, 0xE5,
+ 0x15, 0xD2, 0x26, 0x18, 0x98, 0xFA, 0x05, 0x10,
+ 0x15, 0x72, 0x8E, 0x5A, 0x8A, 0xAC, 0xAA, 0x68,
+ 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
+ };
+ return BN_bin2bn(RFC3526_PRIME_2048, sizeof(RFC3526_PRIME_2048), bn);
+}
+
+/*-
+ * "3072-bit MODP Group" from RFC3526, Section 4.
+ *
+ * The prime is: 2^3072 - 2^3008 - 1 + 2^64 * { [2^2942 pi] + 1690314 }
+ *
+ * RFC3526 specifies a generator of 2.
+ */
+
+BIGNUM *BN_get_rfc3526_prime_3072(BIGNUM *bn)
+{
+ static const unsigned char RFC3526_PRIME_3072[] = {
+ 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
+ 0xC9, 0x0F, 0xDA, 0xA2, 0x21, 0x68, 0xC2, 0x34,
+ 0xC4, 0xC6, 0x62, 0x8B, 0x80, 0xDC, 0x1C, 0xD1,
+ 0x29, 0x02, 0x4E, 0x08, 0x8A, 0x67, 0xCC, 0x74,
+ 0x02, 0x0B, 0xBE, 0xA6, 0x3B, 0x13, 0x9B, 0x22,
+ 0x51, 0x4A, 0x08, 0x79, 0x8E, 0x34, 0x04, 0xDD,
+ 0xEF, 0x95, 0x19, 0xB3, 0xCD, 0x3A, 0x43, 0x1B,
+ 0x30, 0x2B, 0x0A, 0x6D, 0xF2, 0x5F, 0x14, 0x37,
+ 0x4F, 0xE1, 0x35, 0x6D, 0x6D, 0x51, 0xC2, 0x45,
+ 0xE4, 0x85, 0xB5, 0x76, 0x62, 0x5E, 0x7E, 0xC6,
+ 0xF4, 0x4C, 0x42, 0xE9, 0xA6, 0x37, 0xED, 0x6B,
+ 0x0B, 0xFF, 0x5C, 0xB6, 0xF4, 0x06, 0xB7, 0xED,
+ 0xEE, 0x38, 0x6B, 0xFB, 0x5A, 0x89, 0x9F, 0xA5,
+ 0xAE, 0x9F, 0x24, 0x11, 0x7C, 0x4B, 0x1F, 0xE6,
+ 0x49, 0x28, 0x66, 0x51, 0xEC, 0xE4, 0x5B, 0x3D,
+ 0xC2, 0x00, 0x7C, 0xB8, 0xA1, 0x63, 0xBF, 0x05,
+ 0x98, 0xDA, 0x48, 0x36, 0x1C, 0x55, 0xD3, 0x9A,
+ 0x69, 0x16, 0x3F, 0xA8, 0xFD, 0x24, 0xCF, 0x5F,
+ 0x83, 0x65, 0x5D, 0x23, 0xDC, 0xA3, 0xAD, 0x96,
+ 0x1C, 0x62, 0xF3, 0x56, 0x20, 0x85, 0x52, 0xBB,
+ 0x9E, 0xD5, 0x29, 0x07, 0x70, 0x96, 0x96, 0x6D,
+ 0x67, 0x0C, 0x35, 0x4E, 0x4A, 0xBC, 0x98, 0x04,
+ 0xF1, 0x74, 0x6C, 0x08, 0xCA, 0x18, 0x21, 0x7C,
+ 0x32, 0x90, 0x5E, 0x46, 0x2E, 0x36, 0xCE, 0x3B,
+ 0xE3, 0x9E, 0x77, 0x2C, 0x18, 0x0E, 0x86, 0x03,
+ 0x9B, 0x27, 0x83, 0xA2, 0xEC, 0x07, 0xA2, 0x8F,
+ 0xB5, 0xC5, 0x5D, 0xF0, 0x6F, 0x4C, 0x52, 0xC9,
+ 0xDE, 0x2B, 0xCB, 0xF6, 0x95, 0x58, 0x17, 0x18,
+ 0x39, 0x95, 0x49, 0x7C, 0xEA, 0x95, 0x6A, 0xE5,
+ 0x15, 0xD2, 0x26, 0x18, 0x98, 0xFA, 0x05, 0x10,
+ 0x15, 0x72, 0x8E, 0x5A, 0x8A, 0xAA, 0xC4, 0x2D,
+ 0xAD, 0x33, 0x17, 0x0D, 0x04, 0x50, 0x7A, 0x33,
+ 0xA8, 0x55, 0x21, 0xAB, 0xDF, 0x1C, 0xBA, 0x64,
+ 0xEC, 0xFB, 0x85, 0x04, 0x58, 0xDB, 0xEF, 0x0A,
+ 0x8A, 0xEA, 0x71, 0x57, 0x5D, 0x06, 0x0C, 0x7D,
+ 0xB3, 0x97, 0x0F, 0x85, 0xA6, 0xE1, 0xE4, 0xC7,
+ 0xAB, 0xF5, 0xAE, 0x8C, 0xDB, 0x09, 0x33, 0xD7,
+ 0x1E, 0x8C, 0x94, 0xE0, 0x4A, 0x25, 0x61, 0x9D,
+ 0xCE, 0xE3, 0xD2, 0x26, 0x1A, 0xD2, 0xEE, 0x6B,
+ 0xF1, 0x2F, 0xFA, 0x06, 0xD9, 0x8A, 0x08, 0x64,
+ 0xD8, 0x76, 0x02, 0x73, 0x3E, 0xC8, 0x6A, 0x64,
+ 0x52, 0x1F, 0x2B, 0x18, 0x17, 0x7B, 0x20, 0x0C,
+ 0xBB, 0xE1, 0x17, 0x57, 0x7A, 0x61, 0x5D, 0x6C,
+ 0x77, 0x09, 0x88, 0xC0, 0xBA, 0xD9, 0x46, 0xE2,
+ 0x08, 0xE2, 0x4F, 0xA0, 0x74, 0xE5, 0xAB, 0x31,
+ 0x43, 0xDB, 0x5B, 0xFC, 0xE0, 0xFD, 0x10, 0x8E,
+ 0x4B, 0x82, 0xD1, 0x20, 0xA9, 0x3A, 0xD2, 0xCA,
+ 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
+ };
+ return BN_bin2bn(RFC3526_PRIME_3072, sizeof(RFC3526_PRIME_3072), bn);
+}
+
+/*-
+ * "4096-bit MODP Group" from RFC3526, Section 5.
+ *
+ * The prime is: 2^4096 - 2^4032 - 1 + 2^64 * { [2^3966 pi] + 240904 }
+ *
+ * RFC3526 specifies a generator of 2.
+ */
+
+BIGNUM *BN_get_rfc3526_prime_4096(BIGNUM *bn)
+{
+ static const unsigned char RFC3526_PRIME_4096[] = {
+ 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
+ 0xC9, 0x0F, 0xDA, 0xA2, 0x21, 0x68, 0xC2, 0x34,
+ 0xC4, 0xC6, 0x62, 0x8B, 0x80, 0xDC, 0x1C, 0xD1,
+ 0x29, 0x02, 0x4E, 0x08, 0x8A, 0x67, 0xCC, 0x74,
+ 0x02, 0x0B, 0xBE, 0xA6, 0x3B, 0x13, 0x9B, 0x22,
+ 0x51, 0x4A, 0x08, 0x79, 0x8E, 0x34, 0x04, 0xDD,
+ 0xEF, 0x95, 0x19, 0xB3, 0xCD, 0x3A, 0x43, 0x1B,
+ 0x30, 0x2B, 0x0A, 0x6D, 0xF2, 0x5F, 0x14, 0x37,
+ 0x4F, 0xE1, 0x35, 0x6D, 0x6D, 0x51, 0xC2, 0x45,
+ 0xE4, 0x85, 0xB5, 0x76, 0x62, 0x5E, 0x7E, 0xC6,
+ 0xF4, 0x4C, 0x42, 0xE9, 0xA6, 0x37, 0xED, 0x6B,
+ 0x0B, 0xFF, 0x5C, 0xB6, 0xF4, 0x06, 0xB7, 0xED,
+ 0xEE, 0x38, 0x6B, 0xFB, 0x5A, 0x89, 0x9F, 0xA5,
+ 0xAE, 0x9F, 0x24, 0x11, 0x7C, 0x4B, 0x1F, 0xE6,
+ 0x49, 0x28, 0x66, 0x51, 0xEC, 0xE4, 0x5B, 0x3D,
+ 0xC2, 0x00, 0x7C, 0xB8, 0xA1, 0x63, 0xBF, 0x05,
+ 0x98, 0xDA, 0x48, 0x36, 0x1C, 0x55, 0xD3, 0x9A,
+ 0x69, 0x16, 0x3F, 0xA8, 0xFD, 0x24, 0xCF, 0x5F,
+ 0x83, 0x65, 0x5D, 0x23, 0xDC, 0xA3, 0xAD, 0x96,
+ 0x1C, 0x62, 0xF3, 0x56, 0x20, 0x85, 0x52, 0xBB,
+ 0x9E, 0xD5, 0x29, 0x07, 0x70, 0x96, 0x96, 0x6D,
+ 0x67, 0x0C, 0x35, 0x4E, 0x4A, 0xBC, 0x98, 0x04,
+ 0xF1, 0x74, 0x6C, 0x08, 0xCA, 0x18, 0x21, 0x7C,
+ 0x32, 0x90, 0x5E, 0x46, 0x2E, 0x36, 0xCE, 0x3B,
+ 0xE3, 0x9E, 0x77, 0x2C, 0x18, 0x0E, 0x86, 0x03,
+ 0x9B, 0x27, 0x83, 0xA2, 0xEC, 0x07, 0xA2, 0x8F,
+ 0xB5, 0xC5, 0x5D, 0xF0, 0x6F, 0x4C, 0x52, 0xC9,
+ 0xDE, 0x2B, 0xCB, 0xF6, 0x95, 0x58, 0x17, 0x18,
+ 0x39, 0x95, 0x49, 0x7C, 0xEA, 0x95, 0x6A, 0xE5,
+ 0x15, 0xD2, 0x26, 0x18, 0x98, 0xFA, 0x05, 0x10,
+ 0x15, 0x72, 0x8E, 0x5A, 0x8A, 0xAA, 0xC4, 0x2D,
+ 0xAD, 0x33, 0x17, 0x0D, 0x04, 0x50, 0x7A, 0x33,
+ 0xA8, 0x55, 0x21, 0xAB, 0xDF, 0x1C, 0xBA, 0x64,
+ 0xEC, 0xFB, 0x85, 0x04, 0x58, 0xDB, 0xEF, 0x0A,
+ 0x8A, 0xEA, 0x71, 0x57, 0x5D, 0x06, 0x0C, 0x7D,
+ 0xB3, 0x97, 0x0F, 0x85, 0xA6, 0xE1, 0xE4, 0xC7,
+ 0xAB, 0xF5, 0xAE, 0x8C, 0xDB, 0x09, 0x33, 0xD7,
+ 0x1E, 0x8C, 0x94, 0xE0, 0x4A, 0x25, 0x61, 0x9D,
+ 0xCE, 0xE3, 0xD2, 0x26, 0x1A, 0xD2, 0xEE, 0x6B,
+ 0xF1, 0x2F, 0xFA, 0x06, 0xD9, 0x8A, 0x08, 0x64,
+ 0xD8, 0x76, 0x02, 0x73, 0x3E, 0xC8, 0x6A, 0x64,
+ 0x52, 0x1F, 0x2B, 0x18, 0x17, 0x7B, 0x20, 0x0C,
+ 0xBB, 0xE1, 0x17, 0x57, 0x7A, 0x61, 0x5D, 0x6C,
+ 0x77, 0x09, 0x88, 0xC0, 0xBA, 0xD9, 0x46, 0xE2,
+ 0x08, 0xE2, 0x4F, 0xA0, 0x74, 0xE5, 0xAB, 0x31,
+ 0x43, 0xDB, 0x5B, 0xFC, 0xE0, 0xFD, 0x10, 0x8E,
+ 0x4B, 0x82, 0xD1, 0x20, 0xA9, 0x21, 0x08, 0x01,
+ 0x1A, 0x72, 0x3C, 0x12, 0xA7, 0x87, 0xE6, 0xD7,
+ 0x88, 0x71, 0x9A, 0x10, 0xBD, 0xBA, 0x5B, 0x26,
+ 0x99, 0xC3, 0x27, 0x18, 0x6A, 0xF4, 0xE2, 0x3C,
+ 0x1A, 0x94, 0x68, 0x34, 0xB6, 0x15, 0x0B, 0xDA,
+ 0x25, 0x83, 0xE9, 0xCA, 0x2A, 0xD4, 0x4C, 0xE8,
+ 0xDB, 0xBB, 0xC2, 0xDB, 0x04, 0xDE, 0x8E, 0xF9,
+ 0x2E, 0x8E, 0xFC, 0x14, 0x1F, 0xBE, 0xCA, 0xA6,
+ 0x28, 0x7C, 0x59, 0x47, 0x4E, 0x6B, 0xC0, 0x5D,
+ 0x99, 0xB2, 0x96, 0x4F, 0xA0, 0x90, 0xC3, 0xA2,
+ 0x23, 0x3B, 0xA1, 0x86, 0x51, 0x5B, 0xE7, 0xED,
+ 0x1F, 0x61, 0x29, 0x70, 0xCE, 0xE2, 0xD7, 0xAF,
+ 0xB8, 0x1B, 0xDD, 0x76, 0x21, 0x70, 0x48, 0x1C,
+ 0xD0, 0x06, 0x91, 0x27, 0xD5, 0xB0, 0x5A, 0xA9,
+ 0x93, 0xB4, 0xEA, 0x98, 0x8D, 0x8F, 0xDD, 0xC1,
+ 0x86, 0xFF, 0xB7, 0xDC, 0x90, 0xA6, 0xC0, 0x8F,
+ 0x4D, 0xF4, 0x35, 0xC9, 0x34, 0x06, 0x31, 0x99,
+ 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
+ };
+ return BN_bin2bn(RFC3526_PRIME_4096, sizeof(RFC3526_PRIME_4096), bn);
+}
+
+/*-
+ * "6144-bit MODP Group" from RFC3526, Section 6.
+ *
+ * The prime is: 2^6144 - 2^6080 - 1 + 2^64 * { [2^6014 pi] + 929484 }
+ *
+ * RFC3526 specifies a generator of 2.
+ */
+
+BIGNUM *BN_get_rfc3526_prime_6144(BIGNUM *bn)
+{
+ static const unsigned char RFC3526_PRIME_6144[] = {
+ 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
+ 0xC9, 0x0F, 0xDA, 0xA2, 0x21, 0x68, 0xC2, 0x34,
+ 0xC4, 0xC6, 0x62, 0x8B, 0x80, 0xDC, 0x1C, 0xD1,
+ 0x29, 0x02, 0x4E, 0x08, 0x8A, 0x67, 0xCC, 0x74,
+ 0x02, 0x0B, 0xBE, 0xA6, 0x3B, 0x13, 0x9B, 0x22,
+ 0x51, 0x4A, 0x08, 0x79, 0x8E, 0x34, 0x04, 0xDD,
+ 0xEF, 0x95, 0x19, 0xB3, 0xCD, 0x3A, 0x43, 0x1B,
+ 0x30, 0x2B, 0x0A, 0x6D, 0xF2, 0x5F, 0x14, 0x37,
+ 0x4F, 0xE1, 0x35, 0x6D, 0x6D, 0x51, 0xC2, 0x45,
+ 0xE4, 0x85, 0xB5, 0x76, 0x62, 0x5E, 0x7E, 0xC6,
+ 0xF4, 0x4C, 0x42, 0xE9, 0xA6, 0x37, 0xED, 0x6B,
+ 0x0B, 0xFF, 0x5C, 0xB6, 0xF4, 0x06, 0xB7, 0xED,
+ 0xEE, 0x38, 0x6B, 0xFB, 0x5A, 0x89, 0x9F, 0xA5,
+ 0xAE, 0x9F, 0x24, 0x11, 0x7C, 0x4B, 0x1F, 0xE6,
+ 0x49, 0x28, 0x66, 0x51, 0xEC, 0xE4, 0x5B, 0x3D,
+ 0xC2, 0x00, 0x7C, 0xB8, 0xA1, 0x63, 0xBF, 0x05,
+ 0x98, 0xDA, 0x48, 0x36, 0x1C, 0x55, 0xD3, 0x9A,
+ 0x69, 0x16, 0x3F, 0xA8, 0xFD, 0x24, 0xCF, 0x5F,
+ 0x83, 0x65, 0x5D, 0x23, 0xDC, 0xA3, 0xAD, 0x96,
+ 0x1C, 0x62, 0xF3, 0x56, 0x20, 0x85, 0x52, 0xBB,
+ 0x9E, 0xD5, 0x29, 0x07, 0x70, 0x96, 0x96, 0x6D,
+ 0x67, 0x0C, 0x35, 0x4E, 0x4A, 0xBC, 0x98, 0x04,
+ 0xF1, 0x74, 0x6C, 0x08, 0xCA, 0x18, 0x21, 0x7C,
+ 0x32, 0x90, 0x5E, 0x46, 0x2E, 0x36, 0xCE, 0x3B,
+ 0xE3, 0x9E, 0x77, 0x2C, 0x18, 0x0E, 0x86, 0x03,
+ 0x9B, 0x27, 0x83, 0xA2, 0xEC, 0x07, 0xA2, 0x8F,
+ 0xB5, 0xC5, 0x5D, 0xF0, 0x6F, 0x4C, 0x52, 0xC9,
+ 0xDE, 0x2B, 0xCB, 0xF6, 0x95, 0x58, 0x17, 0x18,
+ 0x39, 0x95, 0x49, 0x7C, 0xEA, 0x95, 0x6A, 0xE5,
+ 0x15, 0xD2, 0x26, 0x18, 0x98, 0xFA, 0x05, 0x10,
+ 0x15, 0x72, 0x8E, 0x5A, 0x8A, 0xAA, 0xC4, 0x2D,
+ 0xAD, 0x33, 0x17, 0x0D, 0x04, 0x50, 0x7A, 0x33,
+ 0xA8, 0x55, 0x21, 0xAB, 0xDF, 0x1C, 0xBA, 0x64,
+ 0xEC, 0xFB, 0x85, 0x04, 0x58, 0xDB, 0xEF, 0x0A,
+ 0x8A, 0xEA, 0x71, 0x57, 0x5D, 0x06, 0x0C, 0x7D,
+ 0xB3, 0x97, 0x0F, 0x85, 0xA6, 0xE1, 0xE4, 0xC7,
+ 0xAB, 0xF5, 0xAE, 0x8C, 0xDB, 0x09, 0x33, 0xD7,
+ 0x1E, 0x8C, 0x94, 0xE0, 0x4A, 0x25, 0x61, 0x9D,
+ 0xCE, 0xE3, 0xD2, 0x26, 0x1A, 0xD2, 0xEE, 0x6B,
+ 0xF1, 0x2F, 0xFA, 0x06, 0xD9, 0x8A, 0x08, 0x64,
+ 0xD8, 0x76, 0x02, 0x73, 0x3E, 0xC8, 0x6A, 0x64,
+ 0x52, 0x1F, 0x2B, 0x18, 0x17, 0x7B, 0x20, 0x0C,
+ 0xBB, 0xE1, 0x17, 0x57, 0x7A, 0x61, 0x5D, 0x6C,
+ 0x77, 0x09, 0x88, 0xC0, 0xBA, 0xD9, 0x46, 0xE2,
+ 0x08, 0xE2, 0x4F, 0xA0, 0x74, 0xE5, 0xAB, 0x31,
+ 0x43, 0xDB, 0x5B, 0xFC, 0xE0, 0xFD, 0x10, 0x8E,
+ 0x4B, 0x82, 0xD1, 0x20, 0xA9, 0x21, 0x08, 0x01,
+ 0x1A, 0x72, 0x3C, 0x12, 0xA7, 0x87, 0xE6, 0xD7,
+ 0x88, 0x71, 0x9A, 0x10, 0xBD, 0xBA, 0x5B, 0x26,
+ 0x99, 0xC3, 0x27, 0x18, 0x6A, 0xF4, 0xE2, 0x3C,
+ 0x1A, 0x94, 0x68, 0x34, 0xB6, 0x15, 0x0B, 0xDA,
+ 0x25, 0x83, 0xE9, 0xCA, 0x2A, 0xD4, 0x4C, 0xE8,
+ 0xDB, 0xBB, 0xC2, 0xDB, 0x04, 0xDE, 0x8E, 0xF9,
+ 0x2E, 0x8E, 0xFC, 0x14, 0x1F, 0xBE, 0xCA, 0xA6,
+ 0x28, 0x7C, 0x59, 0x47, 0x4E, 0x6B, 0xC0, 0x5D,
+ 0x99, 0xB2, 0x96, 0x4F, 0xA0, 0x90, 0xC3, 0xA2,
+ 0x23, 0x3B, 0xA1, 0x86, 0x51, 0x5B, 0xE7, 0xED,
+ 0x1F, 0x61, 0x29, 0x70, 0xCE, 0xE2, 0xD7, 0xAF,
+ 0xB8, 0x1B, 0xDD, 0x76, 0x21, 0x70, 0x48, 0x1C,
+ 0xD0, 0x06, 0x91, 0x27, 0xD5, 0xB0, 0x5A, 0xA9,
+ 0x93, 0xB4, 0xEA, 0x98, 0x8D, 0x8F, 0xDD, 0xC1,
+ 0x86, 0xFF, 0xB7, 0xDC, 0x90, 0xA6, 0xC0, 0x8F,
+ 0x4D, 0xF4, 0x35, 0xC9, 0x34, 0x02, 0x84, 0x92,
+ 0x36, 0xC3, 0xFA, 0xB4, 0xD2, 0x7C, 0x70, 0x26,
+ 0xC1, 0xD4, 0xDC, 0xB2, 0x60, 0x26, 0x46, 0xDE,
+ 0xC9, 0x75, 0x1E, 0x76, 0x3D, 0xBA, 0x37, 0xBD,
+ 0xF8, 0xFF, 0x94, 0x06, 0xAD, 0x9E, 0x53, 0x0E,
+ 0xE5, 0xDB, 0x38, 0x2F, 0x41, 0x30, 0x01, 0xAE,
+ 0xB0, 0x6A, 0x53, 0xED, 0x90, 0x27, 0xD8, 0x31,
+ 0x17, 0x97, 0x27, 0xB0, 0x86, 0x5A, 0x89, 0x18,
+ 0xDA, 0x3E, 0xDB, 0xEB, 0xCF, 0x9B, 0x14, 0xED,
+ 0x44, 0xCE, 0x6C, 0xBA, 0xCE, 0xD4, 0xBB, 0x1B,
+ 0xDB, 0x7F, 0x14, 0x47, 0xE6, 0xCC, 0x25, 0x4B,
+ 0x33, 0x20, 0x51, 0x51, 0x2B, 0xD7, 0xAF, 0x42,
+ 0x6F, 0xB8, 0xF4, 0x01, 0x37, 0x8C, 0xD2, 0xBF,
+ 0x59, 0x83, 0xCA, 0x01, 0xC6, 0x4B, 0x92, 0xEC,
+ 0xF0, 0x32, 0xEA, 0x15, 0xD1, 0x72, 0x1D, 0x03,
+ 0xF4, 0x82, 0xD7, 0xCE, 0x6E, 0x74, 0xFE, 0xF6,
+ 0xD5, 0x5E, 0x70, 0x2F, 0x46, 0x98, 0x0C, 0x82,
+ 0xB5, 0xA8, 0x40, 0x31, 0x90, 0x0B, 0x1C, 0x9E,
+ 0x59, 0xE7, 0xC9, 0x7F, 0xBE, 0xC7, 0xE8, 0xF3,
+ 0x23, 0xA9, 0x7A, 0x7E, 0x36, 0xCC, 0x88, 0xBE,
+ 0x0F, 0x1D, 0x45, 0xB7, 0xFF, 0x58, 0x5A, 0xC5,
+ 0x4B, 0xD4, 0x07, 0xB2, 0x2B, 0x41, 0x54, 0xAA,
+ 0xCC, 0x8F, 0x6D, 0x7E, 0xBF, 0x48, 0xE1, 0xD8,
+ 0x14, 0xCC, 0x5E, 0xD2, 0x0F, 0x80, 0x37, 0xE0,
+ 0xA7, 0x97, 0x15, 0xEE, 0xF2, 0x9B, 0xE3, 0x28,
+ 0x06, 0xA1, 0xD5, 0x8B, 0xB7, 0xC5, 0xDA, 0x76,
+ 0xF5, 0x50, 0xAA, 0x3D, 0x8A, 0x1F, 0xBF, 0xF0,
+ 0xEB, 0x19, 0xCC, 0xB1, 0xA3, 0x13, 0xD5, 0x5C,
+ 0xDA, 0x56, 0xC9, 0xEC, 0x2E, 0xF2, 0x96, 0x32,
+ 0x38, 0x7F, 0xE8, 0xD7, 0x6E, 0x3C, 0x04, 0x68,
+ 0x04, 0x3E, 0x8F, 0x66, 0x3F, 0x48, 0x60, 0xEE,
+ 0x12, 0xBF, 0x2D, 0x5B, 0x0B, 0x74, 0x74, 0xD6,
+ 0xE6, 0x94, 0xF9, 0x1E, 0x6D, 0xCC, 0x40, 0x24,
+ 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
+ };
+ return BN_bin2bn(RFC3526_PRIME_6144, sizeof(RFC3526_PRIME_6144), bn);
+}
+
+/*-
+ * "8192-bit MODP Group" from RFC3526, Section 7.
+ *
+ * The prime is: 2^8192 - 2^8128 - 1 + 2^64 * { [2^8062 pi] + 4743158 }
+ *
+ * RFC3526 specifies a generator of 2.
+ */
+
+BIGNUM *BN_get_rfc3526_prime_8192(BIGNUM *bn)
+{
+ static const unsigned char RFC3526_PRIME_8192[] = {
+ 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
+ 0xC9, 0x0F, 0xDA, 0xA2, 0x21, 0x68, 0xC2, 0x34,
+ 0xC4, 0xC6, 0x62, 0x8B, 0x80, 0xDC, 0x1C, 0xD1,
+ 0x29, 0x02, 0x4E, 0x08, 0x8A, 0x67, 0xCC, 0x74,
+ 0x02, 0x0B, 0xBE, 0xA6, 0x3B, 0x13, 0x9B, 0x22,
+ 0x51, 0x4A, 0x08, 0x79, 0x8E, 0x34, 0x04, 0xDD,
+ 0xEF, 0x95, 0x19, 0xB3, 0xCD, 0x3A, 0x43, 0x1B,
+ 0x30, 0x2B, 0x0A, 0x6D, 0xF2, 0x5F, 0x14, 0x37,
+ 0x4F, 0xE1, 0x35, 0x6D, 0x6D, 0x51, 0xC2, 0x45,
+ 0xE4, 0x85, 0xB5, 0x76, 0x62, 0x5E, 0x7E, 0xC6,
+ 0xF4, 0x4C, 0x42, 0xE9, 0xA6, 0x37, 0xED, 0x6B,
+ 0x0B, 0xFF, 0x5C, 0xB6, 0xF4, 0x06, 0xB7, 0xED,
+ 0xEE, 0x38, 0x6B, 0xFB, 0x5A, 0x89, 0x9F, 0xA5,
+ 0xAE, 0x9F, 0x24, 0x11, 0x7C, 0x4B, 0x1F, 0xE6,
+ 0x49, 0x28, 0x66, 0x51, 0xEC, 0xE4, 0x5B, 0x3D,
+ 0xC2, 0x00, 0x7C, 0xB8, 0xA1, 0x63, 0xBF, 0x05,
+ 0x98, 0xDA, 0x48, 0x36, 0x1C, 0x55, 0xD3, 0x9A,
+ 0x69, 0x16, 0x3F, 0xA8, 0xFD, 0x24, 0xCF, 0x5F,
+ 0x83, 0x65, 0x5D, 0x23, 0xDC, 0xA3, 0xAD, 0x96,
+ 0x1C, 0x62, 0xF3, 0x56, 0x20, 0x85, 0x52, 0xBB,
+ 0x9E, 0xD5, 0x29, 0x07, 0x70, 0x96, 0x96, 0x6D,
+ 0x67, 0x0C, 0x35, 0x4E, 0x4A, 0xBC, 0x98, 0x04,
+ 0xF1, 0x74, 0x6C, 0x08, 0xCA, 0x18, 0x21, 0x7C,
+ 0x32, 0x90, 0x5E, 0x46, 0x2E, 0x36, 0xCE, 0x3B,
+ 0xE3, 0x9E, 0x77, 0x2C, 0x18, 0x0E, 0x86, 0x03,
+ 0x9B, 0x27, 0x83, 0xA2, 0xEC, 0x07, 0xA2, 0x8F,
+ 0xB5, 0xC5, 0x5D, 0xF0, 0x6F, 0x4C, 0x52, 0xC9,
+ 0xDE, 0x2B, 0xCB, 0xF6, 0x95, 0x58, 0x17, 0x18,
+ 0x39, 0x95, 0x49, 0x7C, 0xEA, 0x95, 0x6A, 0xE5,
+ 0x15, 0xD2, 0x26, 0x18, 0x98, 0xFA, 0x05, 0x10,
+ 0x15, 0x72, 0x8E, 0x5A, 0x8A, 0xAA, 0xC4, 0x2D,
+ 0xAD, 0x33, 0x17, 0x0D, 0x04, 0x50, 0x7A, 0x33,
+ 0xA8, 0x55, 0x21, 0xAB, 0xDF, 0x1C, 0xBA, 0x64,
+ 0xEC, 0xFB, 0x85, 0x04, 0x58, 0xDB, 0xEF, 0x0A,
+ 0x8A, 0xEA, 0x71, 0x57, 0x5D, 0x06, 0x0C, 0x7D,
+ 0xB3, 0x97, 0x0F, 0x85, 0xA6, 0xE1, 0xE4, 0xC7,
+ 0xAB, 0xF5, 0xAE, 0x8C, 0xDB, 0x09, 0x33, 0xD7,
+ 0x1E, 0x8C, 0x94, 0xE0, 0x4A, 0x25, 0x61, 0x9D,
+ 0xCE, 0xE3, 0xD2, 0x26, 0x1A, 0xD2, 0xEE, 0x6B,
+ 0xF1, 0x2F, 0xFA, 0x06, 0xD9, 0x8A, 0x08, 0x64,
+ 0xD8, 0x76, 0x02, 0x73, 0x3E, 0xC8, 0x6A, 0x64,
+ 0x52, 0x1F, 0x2B, 0x18, 0x17, 0x7B, 0x20, 0x0C,
+ 0xBB, 0xE1, 0x17, 0x57, 0x7A, 0x61, 0x5D, 0x6C,
+ 0x77, 0x09, 0x88, 0xC0, 0xBA, 0xD9, 0x46, 0xE2,
+ 0x08, 0xE2, 0x4F, 0xA0, 0x74, 0xE5, 0xAB, 0x31,
+ 0x43, 0xDB, 0x5B, 0xFC, 0xE0, 0xFD, 0x10, 0x8E,
+ 0x4B, 0x82, 0xD1, 0x20, 0xA9, 0x21, 0x08, 0x01,
+ 0x1A, 0x72, 0x3C, 0x12, 0xA7, 0x87, 0xE6, 0xD7,
+ 0x88, 0x71, 0x9A, 0x10, 0xBD, 0xBA, 0x5B, 0x26,
+ 0x99, 0xC3, 0x27, 0x18, 0x6A, 0xF4, 0xE2, 0x3C,
+ 0x1A, 0x94, 0x68, 0x34, 0xB6, 0x15, 0x0B, 0xDA,
+ 0x25, 0x83, 0xE9, 0xCA, 0x2A, 0xD4, 0x4C, 0xE8,
+ 0xDB, 0xBB, 0xC2, 0xDB, 0x04, 0xDE, 0x8E, 0xF9,
+ 0x2E, 0x8E, 0xFC, 0x14, 0x1F, 0xBE, 0xCA, 0xA6,
+ 0x28, 0x7C, 0x59, 0x47, 0x4E, 0x6B, 0xC0, 0x5D,
+ 0x99, 0xB2, 0x96, 0x4F, 0xA0, 0x90, 0xC3, 0xA2,
+ 0x23, 0x3B, 0xA1, 0x86, 0x51, 0x5B, 0xE7, 0xED,
+ 0x1F, 0x61, 0x29, 0x70, 0xCE, 0xE2, 0xD7, 0xAF,
+ 0xB8, 0x1B, 0xDD, 0x76, 0x21, 0x70, 0x48, 0x1C,
+ 0xD0, 0x06, 0x91, 0x27, 0xD5, 0xB0, 0x5A, 0xA9,
+ 0x93, 0xB4, 0xEA, 0x98, 0x8D, 0x8F, 0xDD, 0xC1,
+ 0x86, 0xFF, 0xB7, 0xDC, 0x90, 0xA6, 0xC0, 0x8F,
+ 0x4D, 0xF4, 0x35, 0xC9, 0x34, 0x02, 0x84, 0x92,
+ 0x36, 0xC3, 0xFA, 0xB4, 0xD2, 0x7C, 0x70, 0x26,
+ 0xC1, 0xD4, 0xDC, 0xB2, 0x60, 0x26, 0x46, 0xDE,
+ 0xC9, 0x75, 0x1E, 0x76, 0x3D, 0xBA, 0x37, 0xBD,
+ 0xF8, 0xFF, 0x94, 0x06, 0xAD, 0x9E, 0x53, 0x0E,
+ 0xE5, 0xDB, 0x38, 0x2F, 0x41, 0x30, 0x01, 0xAE,
+ 0xB0, 0x6A, 0x53, 0xED, 0x90, 0x27, 0xD8, 0x31,
+ 0x17, 0x97, 0x27, 0xB0, 0x86, 0x5A, 0x89, 0x18,
+ 0xDA, 0x3E, 0xDB, 0xEB, 0xCF, 0x9B, 0x14, 0xED,
+ 0x44, 0xCE, 0x6C, 0xBA, 0xCE, 0xD4, 0xBB, 0x1B,
+ 0xDB, 0x7F, 0x14, 0x47, 0xE6, 0xCC, 0x25, 0x4B,
+ 0x33, 0x20, 0x51, 0x51, 0x2B, 0xD7, 0xAF, 0x42,
+ 0x6F, 0xB8, 0xF4, 0x01, 0x37, 0x8C, 0xD2, 0xBF,
+ 0x59, 0x83, 0xCA, 0x01, 0xC6, 0x4B, 0x92, 0xEC,
+ 0xF0, 0x32, 0xEA, 0x15, 0xD1, 0x72, 0x1D, 0x03,
+ 0xF4, 0x82, 0xD7, 0xCE, 0x6E, 0x74, 0xFE, 0xF6,
+ 0xD5, 0x5E, 0x70, 0x2F, 0x46, 0x98, 0x0C, 0x82,
+ 0xB5, 0xA8, 0x40, 0x31, 0x90, 0x0B, 0x1C, 0x9E,
+ 0x59, 0xE7, 0xC9, 0x7F, 0xBE, 0xC7, 0xE8, 0xF3,
+ 0x23, 0xA9, 0x7A, 0x7E, 0x36, 0xCC, 0x88, 0xBE,
+ 0x0F, 0x1D, 0x45, 0xB7, 0xFF, 0x58, 0x5A, 0xC5,
+ 0x4B, 0xD4, 0x07, 0xB2, 0x2B, 0x41, 0x54, 0xAA,
+ 0xCC, 0x8F, 0x6D, 0x7E, 0xBF, 0x48, 0xE1, 0xD8,
+ 0x14, 0xCC, 0x5E, 0xD2, 0x0F, 0x80, 0x37, 0xE0,
+ 0xA7, 0x97, 0x15, 0xEE, 0xF2, 0x9B, 0xE3, 0x28,
+ 0x06, 0xA1, 0xD5, 0x8B, 0xB7, 0xC5, 0xDA, 0x76,
+ 0xF5, 0x50, 0xAA, 0x3D, 0x8A, 0x1F, 0xBF, 0xF0,
+ 0xEB, 0x19, 0xCC, 0xB1, 0xA3, 0x13, 0xD5, 0x5C,
+ 0xDA, 0x56, 0xC9, 0xEC, 0x2E, 0xF2, 0x96, 0x32,
+ 0x38, 0x7F, 0xE8, 0xD7, 0x6E, 0x3C, 0x04, 0x68,
+ 0x04, 0x3E, 0x8F, 0x66, 0x3F, 0x48, 0x60, 0xEE,
+ 0x12, 0xBF, 0x2D, 0x5B, 0x0B, 0x74, 0x74, 0xD6,
+ 0xE6, 0x94, 0xF9, 0x1E, 0x6D, 0xBE, 0x11, 0x59,
+ 0x74, 0xA3, 0x92, 0x6F, 0x12, 0xFE, 0xE5, 0xE4,
+ 0x38, 0x77, 0x7C, 0xB6, 0xA9, 0x32, 0xDF, 0x8C,
+ 0xD8, 0xBE, 0xC4, 0xD0, 0x73, 0xB9, 0x31, 0xBA,
+ 0x3B, 0xC8, 0x32, 0xB6, 0x8D, 0x9D, 0xD3, 0x00,
+ 0x74, 0x1F, 0xA7, 0xBF, 0x8A, 0xFC, 0x47, 0xED,
+ 0x25, 0x76, 0xF6, 0x93, 0x6B, 0xA4, 0x24, 0x66,
+ 0x3A, 0xAB, 0x63, 0x9C, 0x5A, 0xE4, 0xF5, 0x68,
+ 0x34, 0x23, 0xB4, 0x74, 0x2B, 0xF1, 0xC9, 0x78,
+ 0x23, 0x8F, 0x16, 0xCB, 0xE3, 0x9D, 0x65, 0x2D,
+ 0xE3, 0xFD, 0xB8, 0xBE, 0xFC, 0x84, 0x8A, 0xD9,
+ 0x22, 0x22, 0x2E, 0x04, 0xA4, 0x03, 0x7C, 0x07,
+ 0x13, 0xEB, 0x57, 0xA8, 0x1A, 0x23, 0xF0, 0xC7,
+ 0x34, 0x73, 0xFC, 0x64, 0x6C, 0xEA, 0x30, 0x6B,
+ 0x4B, 0xCB, 0xC8, 0x86, 0x2F, 0x83, 0x85, 0xDD,
+ 0xFA, 0x9D, 0x4B, 0x7F, 0xA2, 0xC0, 0x87, 0xE8,
+ 0x79, 0x68, 0x33, 0x03, 0xED, 0x5B, 0xDD, 0x3A,
+ 0x06, 0x2B, 0x3C, 0xF5, 0xB3, 0xA2, 0x78, 0xA6,
+ 0x6D, 0x2A, 0x13, 0xF8, 0x3F, 0x44, 0xF8, 0x2D,
+ 0xDF, 0x31, 0x0E, 0xE0, 0x74, 0xAB, 0x6A, 0x36,
+ 0x45, 0x97, 0xE8, 0x99, 0xA0, 0x25, 0x5D, 0xC1,
+ 0x64, 0xF3, 0x1C, 0xC5, 0x08, 0x46, 0x85, 0x1D,
+ 0xF9, 0xAB, 0x48, 0x19, 0x5D, 0xED, 0x7E, 0xA1,
+ 0xB1, 0xD5, 0x10, 0xBD, 0x7E, 0xE7, 0x4D, 0x73,
+ 0xFA, 0xF3, 0x6B, 0xC3, 0x1E, 0xCF, 0xA2, 0x68,
+ 0x35, 0x90, 0x46, 0xF4, 0xEB, 0x87, 0x9F, 0x92,
+ 0x40, 0x09, 0x43, 0x8B, 0x48, 0x1C, 0x6C, 0xD7,
+ 0x88, 0x9A, 0x00, 0x2E, 0xD5, 0xEE, 0x38, 0x2B,
+ 0xC9, 0x19, 0x0D, 0xA6, 0xFC, 0x02, 0x6E, 0x47,
+ 0x95, 0x58, 0xE4, 0x47, 0x56, 0x77, 0xE9, 0xAA,
+ 0x9E, 0x30, 0x50, 0xE2, 0x76, 0x56, 0x94, 0xDF,
+ 0xC8, 0x1F, 0x56, 0xE8, 0x80, 0xB9, 0x6E, 0x71,
+ 0x60, 0xC9, 0x80, 0xDD, 0x98, 0xED, 0xD3, 0xDF,
+ 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
+ };
+ return BN_bin2bn(RFC3526_PRIME_8192, sizeof(RFC3526_PRIME_8192), bn);
+}
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_ctx.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_ctx.c
new file mode 100644
index 0000000..042cb24
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_ctx.c
@@ -0,0 +1,363 @@
+/*
+ * Copyright 2000-2019 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include "internal/cryptlib.h"
+#include "bn_local.h"
+
+/*-
+ * TODO list
+ *
+ * 1. Check a bunch of "(words+1)" type hacks in various bignum functions and
+ * check they can be safely removed.
+ * - Check +1 and other ugliness in BN_from_montgomery()
+ *
+ * 2. Consider allowing a BN_new_ex() that, at least, lets you specify an
+ * appropriate 'block' size that will be honoured by bn_expand_internal() to
+ * prevent piddly little reallocations. OTOH, profiling bignum expansions in
+ * BN_CTX doesn't show this to be a big issue.
+ */
+
+/* How many bignums are in each "pool item"; */
+#define BN_CTX_POOL_SIZE 16
+/* The stack frame info is resizing, set a first-time expansion size; */
+#define BN_CTX_START_FRAMES 32
+
+/***********/
+/* BN_POOL */
+/***********/
+
+/* A bundle of bignums that can be linked with other bundles */
+typedef struct bignum_pool_item {
+ /* The bignum values */
+ BIGNUM vals[BN_CTX_POOL_SIZE];
+ /* Linked-list admin */
+ struct bignum_pool_item *prev, *next;
+} BN_POOL_ITEM;
+/* A linked-list of bignums grouped in bundles */
+typedef struct bignum_pool {
+ /* Linked-list admin */
+ BN_POOL_ITEM *head, *current, *tail;
+ /* Stack depth and allocation size */
+ unsigned used, size;
+} BN_POOL;
+static void BN_POOL_init(BN_POOL *);
+static void BN_POOL_finish(BN_POOL *);
+static BIGNUM *BN_POOL_get(BN_POOL *, int);
+static void BN_POOL_release(BN_POOL *, unsigned int);
+
+/************/
+/* BN_STACK */
+/************/
+
+/* A wrapper to manage the "stack frames" */
+typedef struct bignum_ctx_stack {
+ /* Array of indexes into the bignum stack */
+ unsigned int *indexes;
+ /* Number of stack frames, and the size of the allocated array */
+ unsigned int depth, size;
+} BN_STACK;
+static void BN_STACK_init(BN_STACK *);
+static void BN_STACK_finish(BN_STACK *);
+static int BN_STACK_push(BN_STACK *, unsigned int);
+static unsigned int BN_STACK_pop(BN_STACK *);
+
+/**********/
+/* BN_CTX */
+/**********/
+
+/* The opaque BN_CTX type */
+struct bignum_ctx {
+ /* The bignum bundles */
+ BN_POOL pool;
+ /* The "stack frames", if you will */
+ BN_STACK stack;
+ /* The number of bignums currently assigned */
+ unsigned int used;
+ /* Depth of stack overflow */
+ int err_stack;
+ /* Block "gets" until an "end" (compatibility behaviour) */
+ int too_many;
+ /* Flags. */
+ int flags;
+};
+
+/* Enable this to find BN_CTX bugs */
+#ifdef BN_CTX_DEBUG
+static const char *ctxdbg_cur = NULL;
+static void ctxdbg(BN_CTX *ctx)
+{
+ unsigned int bnidx = 0, fpidx = 0;
+ BN_POOL_ITEM *item = ctx->pool.head;
+ BN_STACK *stack = &ctx->stack;
+ fprintf(stderr, "(%16p): ", ctx);
+ while (bnidx < ctx->used) {
+ fprintf(stderr, "%03x ", item->vals[bnidx++ % BN_CTX_POOL_SIZE].dmax);
+ if (!(bnidx % BN_CTX_POOL_SIZE))
+ item = item->next;
+ }
+ fprintf(stderr, "\n");
+ bnidx = 0;
+ fprintf(stderr, " : ");
+ while (fpidx < stack->depth) {
+ while (bnidx++ < stack->indexes[fpidx])
+ fprintf(stderr, " ");
+ fprintf(stderr, "^^^ ");
+ bnidx++;
+ fpidx++;
+ }
+ fprintf(stderr, "\n");
+}
+
+# define CTXDBG_ENTRY(str, ctx) do { \
+ ctxdbg_cur = (str); \
+ fprintf(stderr,"Starting %s\n", ctxdbg_cur); \
+ ctxdbg(ctx); \
+ } while(0)
+# define CTXDBG_EXIT(ctx) do { \
+ fprintf(stderr,"Ending %s\n", ctxdbg_cur); \
+ ctxdbg(ctx); \
+ } while(0)
+# define CTXDBG_RET(ctx,ret)
+#else
+# define CTXDBG_ENTRY(str, ctx)
+# define CTXDBG_EXIT(ctx)
+# define CTXDBG_RET(ctx,ret)
+#endif
+
+
+BN_CTX *BN_CTX_new(void)
+{
+ BN_CTX *ret;
+
+ if ((ret = OPENSSL_zalloc(sizeof(*ret))) == NULL) {
+ BNerr(BN_F_BN_CTX_NEW, ERR_R_MALLOC_FAILURE);
+ return NULL;
+ }
+ /* Initialise the structure */
+ BN_POOL_init(&ret->pool);
+ BN_STACK_init(&ret->stack);
+ return ret;
+}
+
+BN_CTX *BN_CTX_secure_new(void)
+{
+ BN_CTX *ret = BN_CTX_new();
+
+ if (ret != NULL)
+ ret->flags = BN_FLG_SECURE;
+ return ret;
+}
+
+void BN_CTX_free(BN_CTX *ctx)
+{
+ if (ctx == NULL)
+ return;
+#ifdef BN_CTX_DEBUG
+ {
+ BN_POOL_ITEM *pool = ctx->pool.head;
+ fprintf(stderr, "BN_CTX_free, stack-size=%d, pool-bignums=%d\n",
+ ctx->stack.size, ctx->pool.size);
+ fprintf(stderr, "dmaxs: ");
+ while (pool) {
+ unsigned loop = 0;
+ while (loop < BN_CTX_POOL_SIZE)
+ fprintf(stderr, "%02x ", pool->vals[loop++].dmax);
+ pool = pool->next;
+ }
+ fprintf(stderr, "\n");
+ }
+#endif
+ BN_STACK_finish(&ctx->stack);
+ BN_POOL_finish(&ctx->pool);
+ OPENSSL_free(ctx);
+}
+
+void BN_CTX_start(BN_CTX *ctx)
+{
+ CTXDBG_ENTRY("BN_CTX_start", ctx);
+ /* If we're already overflowing ... */
+ if (ctx->err_stack || ctx->too_many)
+ ctx->err_stack++;
+ /* (Try to) get a new frame pointer */
+ else if (!BN_STACK_push(&ctx->stack, ctx->used)) {
+ BNerr(BN_F_BN_CTX_START, BN_R_TOO_MANY_TEMPORARY_VARIABLES);
+ ctx->err_stack++;
+ }
+ CTXDBG_EXIT(ctx);
+}
+
+void BN_CTX_end(BN_CTX *ctx)
+{
+ if (ctx == NULL)
+ return;
+ CTXDBG_ENTRY("BN_CTX_end", ctx);
+ if (ctx->err_stack)
+ ctx->err_stack--;
+ else {
+ unsigned int fp = BN_STACK_pop(&ctx->stack);
+ /* Does this stack frame have anything to release? */
+ if (fp < ctx->used)
+ BN_POOL_release(&ctx->pool, ctx->used - fp);
+ ctx->used = fp;
+ /* Unjam "too_many" in case "get" had failed */
+ ctx->too_many = 0;
+ }
+ CTXDBG_EXIT(ctx);
+}
+
+BIGNUM *BN_CTX_get(BN_CTX *ctx)
+{
+ BIGNUM *ret;
+
+ CTXDBG_ENTRY("BN_CTX_get", ctx);
+ if (ctx->err_stack || ctx->too_many)
+ return NULL;
+ if ((ret = BN_POOL_get(&ctx->pool, ctx->flags)) == NULL) {
+ /*
+ * Setting too_many prevents repeated "get" attempts from cluttering
+ * the error stack.
+ */
+ ctx->too_many = 1;
+ BNerr(BN_F_BN_CTX_GET, BN_R_TOO_MANY_TEMPORARY_VARIABLES);
+ return NULL;
+ }
+ /* OK, make sure the returned bignum is "zero" */
+ BN_zero(ret);
+ /* clear BN_FLG_CONSTTIME if leaked from previous frames */
+ ret->flags &= (~BN_FLG_CONSTTIME);
+ ctx->used++;
+ CTXDBG_RET(ctx, ret);
+ return ret;
+}
+
+/************/
+/* BN_STACK */
+/************/
+
+static void BN_STACK_init(BN_STACK *st)
+{
+ st->indexes = NULL;
+ st->depth = st->size = 0;
+}
+
+static void BN_STACK_finish(BN_STACK *st)
+{
+ OPENSSL_free(st->indexes);
+ st->indexes = NULL;
+}
+
+
+static int BN_STACK_push(BN_STACK *st, unsigned int idx)
+{
+ if (st->depth == st->size) {
+ /* Need to expand */
+ unsigned int newsize =
+ st->size ? (st->size * 3 / 2) : BN_CTX_START_FRAMES;
+ unsigned int *newitems;
+
+ if ((newitems = OPENSSL_malloc(sizeof(*newitems) * newsize)) == NULL) {
+ BNerr(BN_F_BN_STACK_PUSH, ERR_R_MALLOC_FAILURE);
+ return 0;
+ }
+ if (st->depth)
+ memcpy(newitems, st->indexes, sizeof(*newitems) * st->depth);
+ OPENSSL_free(st->indexes);
+ st->indexes = newitems;
+ st->size = newsize;
+ }
+ st->indexes[(st->depth)++] = idx;
+ return 1;
+}
+
+static unsigned int BN_STACK_pop(BN_STACK *st)
+{
+ return st->indexes[--(st->depth)];
+}
+
+/***********/
+/* BN_POOL */
+/***********/
+
+static void BN_POOL_init(BN_POOL *p)
+{
+ p->head = p->current = p->tail = NULL;
+ p->used = p->size = 0;
+}
+
+static void BN_POOL_finish(BN_POOL *p)
+{
+ unsigned int loop;
+ BIGNUM *bn;
+
+ while (p->head) {
+ for (loop = 0, bn = p->head->vals; loop++ < BN_CTX_POOL_SIZE; bn++)
+ if (bn->d)
+ BN_clear_free(bn);
+ p->current = p->head->next;
+ OPENSSL_free(p->head);
+ p->head = p->current;
+ }
+}
+
+
+static BIGNUM *BN_POOL_get(BN_POOL *p, int flag)
+{
+ BIGNUM *bn;
+ unsigned int loop;
+
+ /* Full; allocate a new pool item and link it in. */
+ if (p->used == p->size) {
+ BN_POOL_ITEM *item;
+
+ if ((item = OPENSSL_malloc(sizeof(*item))) == NULL) {
+ BNerr(BN_F_BN_POOL_GET, ERR_R_MALLOC_FAILURE);
+ return NULL;
+ }
+ for (loop = 0, bn = item->vals; loop++ < BN_CTX_POOL_SIZE; bn++) {
+ bn_init(bn);
+ if ((flag & BN_FLG_SECURE) != 0)
+ BN_set_flags(bn, BN_FLG_SECURE);
+ }
+ item->prev = p->tail;
+ item->next = NULL;
+
+ if (p->head == NULL)
+ p->head = p->current = p->tail = item;
+ else {
+ p->tail->next = item;
+ p->tail = item;
+ p->current = item;
+ }
+ p->size += BN_CTX_POOL_SIZE;
+ p->used++;
+ /* Return the first bignum from the new pool */
+ return item->vals;
+ }
+
+ if (!p->used)
+ p->current = p->head;
+ else if ((p->used % BN_CTX_POOL_SIZE) == 0)
+ p->current = p->current->next;
+ return p->current->vals + ((p->used++) % BN_CTX_POOL_SIZE);
+}
+
+static void BN_POOL_release(BN_POOL *p, unsigned int num)
+{
+ unsigned int offset = (p->used - 1) % BN_CTX_POOL_SIZE;
+
+ p->used -= num;
+ while (num--) {
+ bn_check_top(p->current->vals + offset);
+ if (offset == 0) {
+ offset = BN_CTX_POOL_SIZE - 1;
+ p->current = p->current->prev;
+ } else
+ offset--;
+ }
+}
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_depr.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_depr.c
new file mode 100644
index 0000000..b60269c
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_depr.c
@@ -0,0 +1,68 @@
+/*
+ * Copyright 2002-2019 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+/*
+ * Support for deprecated functions goes here - static linkage will only
+ * slurp this code if applications are using them directly.
+ */
+
+#include <openssl/opensslconf.h>
+#if OPENSSL_API_COMPAT >= 0x00908000L
+NON_EMPTY_TRANSLATION_UNIT
+#else
+
+# include <stdio.h>
+# include <time.h>
+# include "internal/cryptlib.h"
+# include "bn_local.h"
+
+BIGNUM *BN_generate_prime(BIGNUM *ret, int bits, int safe,
+ const BIGNUM *add, const BIGNUM *rem,
+ void (*callback) (int, int, void *), void *cb_arg)
+{
+ BN_GENCB cb;
+ BIGNUM *rnd = NULL;
+
+ BN_GENCB_set_old(&cb, callback, cb_arg);
+
+ if (ret == NULL) {
+ if ((rnd = BN_new()) == NULL)
+ goto err;
+ } else
+ rnd = ret;
+ if (!BN_generate_prime_ex(rnd, bits, safe, add, rem, &cb))
+ goto err;
+
+ /* we have a prime :-) */
+ return rnd;
+ err:
+ BN_free(rnd);
+ return NULL;
+}
+
+int BN_is_prime(const BIGNUM *a, int checks,
+ void (*callback) (int, int, void *), BN_CTX *ctx_passed,
+ void *cb_arg)
+{
+ BN_GENCB cb;
+ BN_GENCB_set_old(&cb, callback, cb_arg);
+ return BN_is_prime_ex(a, checks, ctx_passed, &cb);
+}
+
+int BN_is_prime_fasttest(const BIGNUM *a, int checks,
+ void (*callback) (int, int, void *),
+ BN_CTX *ctx_passed, void *cb_arg,
+ int do_trial_division)
+{
+ BN_GENCB cb;
+ BN_GENCB_set_old(&cb, callback, cb_arg);
+ return BN_is_prime_fasttest_ex(a, checks, ctx_passed,
+ do_trial_division, &cb);
+}
+#endif
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_dh.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_dh.c
new file mode 100644
index 0000000..58c44f0
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_dh.c
@@ -0,0 +1,512 @@
+/*
+ * Copyright 2014-2017 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include "bn_local.h"
+#include "internal/nelem.h"
+
+#ifndef OPENSSL_NO_DH
+#include <openssl/dh.h>
+#include "crypto/bn_dh.h"
+/* DH parameters from RFC5114 */
+
+# if BN_BITS2 == 64
+static const BN_ULONG dh1024_160_p[] = {
+ 0xDF1FB2BC2E4A4371ULL, 0xE68CFDA76D4DA708ULL, 0x45BF37DF365C1A65ULL,
+ 0xA151AF5F0DC8B4BDULL, 0xFAA31A4FF55BCCC0ULL, 0x4EFFD6FAE5644738ULL,
+ 0x98488E9C219A7372ULL, 0xACCBDD7D90C4BD70ULL, 0x24975C3CD49B83BFULL,
+ 0x13ECB4AEA9061123ULL, 0x9838EF1E2EE652C0ULL, 0x6073E28675A23D18ULL,
+ 0x9A6A9DCA52D23B61ULL, 0x52C99FBCFB06A3C6ULL, 0xDE92DE5EAE5D54ECULL,
+ 0xB10B8F96A080E01DULL
+};
+
+static const BN_ULONG dh1024_160_g[] = {
+ 0x855E6EEB22B3B2E5ULL, 0x858F4DCEF97C2A24ULL, 0x2D779D5918D08BC8ULL,
+ 0xD662A4D18E73AFA3ULL, 0x1DBF0A0169B6A28AULL, 0xA6A24C087A091F53ULL,
+ 0x909D0D2263F80A76ULL, 0xD7FBD7D3B9A92EE1ULL, 0x5E91547F9E2749F4ULL,
+ 0x160217B4B01B886AULL, 0x777E690F5504F213ULL, 0x266FEA1E5C41564BULL,
+ 0xD6406CFF14266D31ULL, 0xF8104DD258AC507FULL, 0x6765A442EFB99905ULL,
+ 0xA4D1CBD5C3FD3412ULL
+};
+
+static const BN_ULONG dh1024_160_q[] = {
+ 0x64B7CB9D49462353ULL, 0x81A8DF278ABA4E7DULL, 0x00000000F518AA87ULL
+};
+
+static const BN_ULONG dh2048_224_p[] = {
+ 0x0AC4DFFE0C10E64FULL, 0xCF9DE5384E71B81CULL, 0x7EF363E2FFA31F71ULL,
+ 0xE3FB73C16B8E75B9ULL, 0xC9B53DCF4BA80A29ULL, 0x23F10B0E16E79763ULL,
+ 0xC52172E413042E9BULL, 0xBE60E69CC928B2B9ULL, 0x80CD86A1B9E587E8ULL,
+ 0x315D75E198C641A4ULL, 0xCDF93ACC44328387ULL, 0x15987D9ADC0A486DULL,
+ 0x7310F7121FD5A074ULL, 0x278273C7DE31EFDCULL, 0x1602E714415D9330ULL,
+ 0x81286130BC8985DBULL, 0xB3BF8A3170918836ULL, 0x6A00E0A0B9C49708ULL,
+ 0xC6BA0B2C8BBC27BEULL, 0xC9F98D11ED34DBF6ULL, 0x7AD5B7D0B6C12207ULL,
+ 0xD91E8FEF55B7394BULL, 0x9037C9EDEFDA4DF8ULL, 0x6D3F8152AD6AC212ULL,
+ 0x1DE6B85A1274A0A6ULL, 0xEB3D688A309C180EULL, 0xAF9A3C407BA1DF15ULL,
+ 0xE6FA141DF95A56DBULL, 0xB54B1597B61D0A75ULL, 0xA20D64E5683B9FD1ULL,
+ 0xD660FAA79559C51FULL, 0xAD107E1E9123A9D0ULL
+};
+
+static const BN_ULONG dh2048_224_g[] = {
+ 0x84B890D3191F2BFAULL, 0x81BC087F2A7065B3ULL, 0x19C418E1F6EC0179ULL,
+ 0x7B5A0F1C71CFFF4CULL, 0xEDFE72FE9B6AA4BDULL, 0x81E1BCFE94B30269ULL,
+ 0x566AFBB48D6C0191ULL, 0xB539CCE3409D13CDULL, 0x6AA21E7F5F2FF381ULL,
+ 0xD9E263E4770589EFULL, 0x10E183EDD19963DDULL, 0xB70A8137150B8EEBULL,
+ 0x051AE3D428C8F8ACULL, 0xBB77A86F0C1AB15BULL, 0x6E3025E316A330EFULL,
+ 0x19529A45D6F83456ULL, 0xF180EB34118E98D1ULL, 0xB5F6C6B250717CBEULL,
+ 0x09939D54DA7460CDULL, 0xE247150422EA1ED4ULL, 0xB8A762D0521BC98AULL,
+ 0xF4D027275AC1348BULL, 0xC17669101999024AULL, 0xBE5E9001A8D66AD7ULL,
+ 0xC57DB17C620A8652ULL, 0xAB739D7700C29F52ULL, 0xDD921F01A70C4AFAULL,
+ 0xA6824A4E10B9A6F0ULL, 0x74866A08CFE4FFE3ULL, 0x6CDEBE7B89998CAFULL,
+ 0x9DF30B5C8FFDAC50ULL, 0xAC4032EF4F2D9AE3ULL
+};
+
+static const BN_ULONG dh2048_224_q[] = {
+ 0xBF389A99B36371EBULL, 0x1F80535A4738CEBCULL, 0xC58D93FE99717710ULL,
+ 0x00000000801C0D34ULL
+};
+
+static const BN_ULONG dh2048_256_p[] = {
+ 0xDB094AE91E1A1597ULL, 0x693877FAD7EF09CAULL, 0x6116D2276E11715FULL,
+ 0xA4B54330C198AF12ULL, 0x75F26375D7014103ULL, 0xC3A3960A54E710C3ULL,
+ 0xDED4010ABD0BE621ULL, 0xC0B857F689962856ULL, 0xB3CA3F7971506026ULL,
+ 0x1CCACB83E6B486F6ULL, 0x67E144E514056425ULL, 0xF6A167B5A41825D9ULL,
+ 0x3AD8347796524D8EULL, 0xF13C6D9A51BFA4ABULL, 0x2D52526735488A0EULL,
+ 0xB63ACAE1CAA6B790ULL, 0x4FDB70C581B23F76ULL, 0xBC39A0BF12307F5CULL,
+ 0xB941F54EB1E59BB8ULL, 0x6C5BFC11D45F9088ULL, 0x22E0B1EF4275BF7BULL,
+ 0x91F9E6725B4758C0ULL, 0x5A8A9D306BCF67EDULL, 0x209E0C6497517ABDULL,
+ 0x3BF4296D830E9A7CULL, 0x16C3D91134096FAAULL, 0xFAF7DF4561B2AA30ULL,
+ 0xE00DF8F1D61957D4ULL, 0x5D2CEED4435E3B00ULL, 0x8CEEF608660DD0F2ULL,
+ 0xFFBBD19C65195999ULL, 0x87A8E61DB4B6663CULL
+};
+
+static const BN_ULONG dh2048_256_g[] = {
+ 0x664B4C0F6CC41659ULL, 0x5E2327CFEF98C582ULL, 0xD647D148D4795451ULL,
+ 0x2F63078490F00EF8ULL, 0x184B523D1DB246C3ULL, 0xC7891428CDC67EB6ULL,
+ 0x7FD028370DF92B52ULL, 0xB3353BBB64E0EC37ULL, 0xECD06E1557CD0915ULL,
+ 0xB7D2BBD2DF016199ULL, 0xC8484B1E052588B9ULL, 0xDB2A3B7313D3FE14ULL,
+ 0xD052B985D182EA0AULL, 0xA4BD1BFFE83B9C80ULL, 0xDFC967C1FB3F2E55ULL,
+ 0xB5045AF2767164E1ULL, 0x1D14348F6F2F9193ULL, 0x64E67982428EBC83ULL,
+ 0x8AC376D282D6ED38ULL, 0x777DE62AAAB8A862ULL, 0xDDF463E5E9EC144BULL,
+ 0x0196F931C77A57F2ULL, 0xA55AE31341000A65ULL, 0x901228F8C28CBB18ULL,
+ 0xBC3773BF7E8C6F62ULL, 0xBE3A6C1B0C6B47B1ULL, 0xFF4FED4AAC0BB555ULL,
+ 0x10DBC15077BE463FULL, 0x07F4793A1A0BA125ULL, 0x4CA7B18F21EF2054ULL,
+ 0x2E77506660EDBD48ULL, 0x3FB32C9B73134D0BULL
+};
+
+static const BN_ULONG dh2048_256_q[] = {
+ 0xA308B0FE64F5FBD3ULL, 0x99B1A47D1EB3750BULL, 0xB447997640129DA2ULL,
+ 0x8CF83642A709A097ULL
+};
+
+/* Primes from RFC 7919 */
+static const BN_ULONG ffdhe2048_p[] = {
+ 0xFFFFFFFFFFFFFFFFULL, 0x886B423861285C97ULL, 0xC6F34A26C1B2EFFAULL,
+ 0xC58EF1837D1683B2ULL, 0x3BB5FCBC2EC22005ULL, 0xC3FE3B1B4C6FAD73ULL,
+ 0x8E4F1232EEF28183ULL, 0x9172FE9CE98583FFULL, 0xC03404CD28342F61ULL,
+ 0x9E02FCE1CDF7E2ECULL, 0x0B07A7C8EE0A6D70ULL, 0xAE56EDE76372BB19ULL,
+ 0x1D4F42A3DE394DF4ULL, 0xB96ADAB760D7F468ULL, 0xD108A94BB2C8E3FBULL,
+ 0xBC0AB182B324FB61ULL, 0x30ACCA4F483A797AULL, 0x1DF158A136ADE735ULL,
+ 0xE2A689DAF3EFE872ULL, 0x984F0C70E0E68B77ULL, 0xB557135E7F57C935ULL,
+ 0x856365553DED1AF3ULL, 0x2433F51F5F066ED0ULL, 0xD3DF1ED5D5FD6561ULL,
+ 0xF681B202AEC4617AULL, 0x7D2FE363630C75D8ULL, 0xCC939DCE249B3EF9ULL,
+ 0xA9E13641146433FBULL, 0xD8B9C583CE2D3695ULL, 0xAFDC5620273D3CF1ULL,
+ 0xADF85458A2BB4A9AULL, 0xFFFFFFFFFFFFFFFFULL
+};
+
+static const BN_ULONG ffdhe3072_p[] = {
+ 0xFFFFFFFFFFFFFFFFULL, 0x25E41D2B66C62E37ULL, 0x3C1B20EE3FD59D7CULL,
+ 0x0ABCD06BFA53DDEFULL, 0x1DBF9A42D5C4484EULL, 0xABC521979B0DEADAULL,
+ 0xE86D2BC522363A0DULL, 0x5CAE82AB9C9DF69EULL, 0x64F2E21E71F54BFFULL,
+ 0xF4FD4452E2D74DD3ULL, 0xB4130C93BC437944ULL, 0xAEFE130985139270ULL,
+ 0x598CB0FAC186D91CULL, 0x7AD91D2691F7F7EEULL, 0x61B46FC9D6E6C907ULL,
+ 0xBC34F4DEF99C0238ULL, 0xDE355B3B6519035BULL, 0x886B4238611FCFDCULL,
+ 0xC6F34A26C1B2EFFAULL, 0xC58EF1837D1683B2ULL, 0x3BB5FCBC2EC22005ULL,
+ 0xC3FE3B1B4C6FAD73ULL, 0x8E4F1232EEF28183ULL, 0x9172FE9CE98583FFULL,
+ 0xC03404CD28342F61ULL, 0x9E02FCE1CDF7E2ECULL, 0x0B07A7C8EE0A6D70ULL,
+ 0xAE56EDE76372BB19ULL, 0x1D4F42A3DE394DF4ULL, 0xB96ADAB760D7F468ULL,
+ 0xD108A94BB2C8E3FBULL, 0xBC0AB182B324FB61ULL, 0x30ACCA4F483A797AULL,
+ 0x1DF158A136ADE735ULL, 0xE2A689DAF3EFE872ULL, 0x984F0C70E0E68B77ULL,
+ 0xB557135E7F57C935ULL, 0x856365553DED1AF3ULL, 0x2433F51F5F066ED0ULL,
+ 0xD3DF1ED5D5FD6561ULL, 0xF681B202AEC4617AULL, 0x7D2FE363630C75D8ULL,
+ 0xCC939DCE249B3EF9ULL, 0xA9E13641146433FBULL, 0xD8B9C583CE2D3695ULL,
+ 0xAFDC5620273D3CF1ULL, 0xADF85458A2BB4A9AULL, 0xFFFFFFFFFFFFFFFFULL
+};
+
+static const BN_ULONG ffdhe4096_p[] = {
+ 0xFFFFFFFFFFFFFFFFULL, 0xC68A007E5E655F6AULL, 0x4DB5A851F44182E1ULL,
+ 0x8EC9B55A7F88A46BULL, 0x0A8291CDCEC97DCFULL, 0x2A4ECEA9F98D0ACCULL,
+ 0x1A1DB93D7140003CULL, 0x092999A333CB8B7AULL, 0x6DC778F971AD0038ULL,
+ 0xA907600A918130C4ULL, 0xED6A1E012D9E6832ULL, 0x7135C886EFB4318AULL,
+ 0x87F55BA57E31CC7AULL, 0x7763CF1D55034004ULL, 0xAC7D5F42D69F6D18ULL,
+ 0x7930E9E4E58857B6ULL, 0x6E6F52C3164DF4FBULL, 0x25E41D2B669E1EF1ULL,
+ 0x3C1B20EE3FD59D7CULL, 0x0ABCD06BFA53DDEFULL, 0x1DBF9A42D5C4484EULL,
+ 0xABC521979B0DEADAULL, 0xE86D2BC522363A0DULL, 0x5CAE82AB9C9DF69EULL,
+ 0x64F2E21E71F54BFFULL, 0xF4FD4452E2D74DD3ULL, 0xB4130C93BC437944ULL,
+ 0xAEFE130985139270ULL, 0x598CB0FAC186D91CULL, 0x7AD91D2691F7F7EEULL,
+ 0x61B46FC9D6E6C907ULL, 0xBC34F4DEF99C0238ULL, 0xDE355B3B6519035BULL,
+ 0x886B4238611FCFDCULL, 0xC6F34A26C1B2EFFAULL, 0xC58EF1837D1683B2ULL,
+ 0x3BB5FCBC2EC22005ULL, 0xC3FE3B1B4C6FAD73ULL, 0x8E4F1232EEF28183ULL,
+ 0x9172FE9CE98583FFULL, 0xC03404CD28342F61ULL, 0x9E02FCE1CDF7E2ECULL,
+ 0x0B07A7C8EE0A6D70ULL, 0xAE56EDE76372BB19ULL, 0x1D4F42A3DE394DF4ULL,
+ 0xB96ADAB760D7F468ULL, 0xD108A94BB2C8E3FBULL, 0xBC0AB182B324FB61ULL,
+ 0x30ACCA4F483A797AULL, 0x1DF158A136ADE735ULL, 0xE2A689DAF3EFE872ULL,
+ 0x984F0C70E0E68B77ULL, 0xB557135E7F57C935ULL, 0x856365553DED1AF3ULL,
+ 0x2433F51F5F066ED0ULL, 0xD3DF1ED5D5FD6561ULL, 0xF681B202AEC4617AULL,
+ 0x7D2FE363630C75D8ULL, 0xCC939DCE249B3EF9ULL, 0xA9E13641146433FBULL,
+ 0xD8B9C583CE2D3695ULL, 0xAFDC5620273D3CF1ULL, 0xADF85458A2BB4A9AULL,
+ 0xFFFFFFFFFFFFFFFFULL
+};
+
+static const BN_ULONG ffdhe6144_p[] = {
+ 0xFFFFFFFFFFFFFFFFULL, 0xA40E329CD0E40E65ULL, 0xA41D570D7938DAD4ULL,
+ 0x62A69526D43161C1ULL, 0x3FDD4A8E9ADB1E69ULL, 0x5B3B71F9DC6B80D6ULL,
+ 0xEC9D1810C6272B04ULL, 0x8CCF2DD5CACEF403ULL, 0xE49F5235C95B9117ULL,
+ 0x505DC82DB854338AULL, 0x62292C311562A846ULL, 0xD72B03746AE77F5EULL,
+ 0xF9C9091B462D538CULL, 0x0AE8DB5847A67CBEULL, 0xB3A739C122611682ULL,
+ 0xEEAAC0232A281BF6ULL, 0x94C6651E77CAF992ULL, 0x763E4E4B94B2BBC1ULL,
+ 0x587E38DA0077D9B4ULL, 0x7FB29F8C183023C3ULL, 0x0ABEC1FFF9E3A26EULL,
+ 0xA00EF092350511E3ULL, 0xB855322EDB6340D8ULL, 0xA52471F7A9A96910ULL,
+ 0x388147FB4CFDB477ULL, 0x9B1F5C3E4E46041FULL, 0xCDAD0657FCCFEC71ULL,
+ 0xB38E8C334C701C3AULL, 0x917BDD64B1C0FD4CULL, 0x3BB454329B7624C8ULL,
+ 0x23BA4442CAF53EA6ULL, 0x4E677D2C38532A3AULL, 0x0BFD64B645036C7AULL,
+ 0xC68A007E5E0DD902ULL, 0x4DB5A851F44182E1ULL, 0x8EC9B55A7F88A46BULL,
+ 0x0A8291CDCEC97DCFULL, 0x2A4ECEA9F98D0ACCULL, 0x1A1DB93D7140003CULL,
+ 0x092999A333CB8B7AULL, 0x6DC778F971AD0038ULL, 0xA907600A918130C4ULL,
+ 0xED6A1E012D9E6832ULL, 0x7135C886EFB4318AULL, 0x87F55BA57E31CC7AULL,
+ 0x7763CF1D55034004ULL, 0xAC7D5F42D69F6D18ULL, 0x7930E9E4E58857B6ULL,
+ 0x6E6F52C3164DF4FBULL, 0x25E41D2B669E1EF1ULL, 0x3C1B20EE3FD59D7CULL,
+ 0x0ABCD06BFA53DDEFULL, 0x1DBF9A42D5C4484EULL, 0xABC521979B0DEADAULL,
+ 0xE86D2BC522363A0DULL, 0x5CAE82AB9C9DF69EULL, 0x64F2E21E71F54BFFULL,
+ 0xF4FD4452E2D74DD3ULL, 0xB4130C93BC437944ULL, 0xAEFE130985139270ULL,
+ 0x598CB0FAC186D91CULL, 0x7AD91D2691F7F7EEULL, 0x61B46FC9D6E6C907ULL,
+ 0xBC34F4DEF99C0238ULL, 0xDE355B3B6519035BULL, 0x886B4238611FCFDCULL,
+ 0xC6F34A26C1B2EFFAULL, 0xC58EF1837D1683B2ULL, 0x3BB5FCBC2EC22005ULL,
+ 0xC3FE3B1B4C6FAD73ULL, 0x8E4F1232EEF28183ULL, 0x9172FE9CE98583FFULL,
+ 0xC03404CD28342F61ULL, 0x9E02FCE1CDF7E2ECULL, 0x0B07A7C8EE0A6D70ULL,
+ 0xAE56EDE76372BB19ULL, 0x1D4F42A3DE394DF4ULL, 0xB96ADAB760D7F468ULL,
+ 0xD108A94BB2C8E3FBULL, 0xBC0AB182B324FB61ULL, 0x30ACCA4F483A797AULL,
+ 0x1DF158A136ADE735ULL, 0xE2A689DAF3EFE872ULL, 0x984F0C70E0E68B77ULL,
+ 0xB557135E7F57C935ULL, 0x856365553DED1AF3ULL, 0x2433F51F5F066ED0ULL,
+ 0xD3DF1ED5D5FD6561ULL, 0xF681B202AEC4617AULL, 0x7D2FE363630C75D8ULL,
+ 0xCC939DCE249B3EF9ULL, 0xA9E13641146433FBULL, 0xD8B9C583CE2D3695ULL,
+ 0xAFDC5620273D3CF1ULL, 0xADF85458A2BB4A9AULL, 0xFFFFFFFFFFFFFFFFULL
+};
+
+static const BN_ULONG ffdhe8192_p[] = {
+ 0xFFFFFFFFFFFFFFFFULL, 0xD68C8BB7C5C6424CULL, 0x011E2A94838FF88CULL,
+ 0x0822E506A9F4614EULL, 0x97D11D49F7A8443DULL, 0xA6BBFDE530677F0DULL,
+ 0x2F741EF8C1FE86FEULL, 0xFAFABE1C5D71A87EULL, 0xDED2FBABFBE58A30ULL,
+ 0xB6855DFE72B0A66EULL, 0x1EFC8CE0BA8A4FE8ULL, 0x83F81D4A3F2FA457ULL,
+ 0xA1FE3075A577E231ULL, 0xD5B8019488D9C0A0ULL, 0x624816CDAD9A95F9ULL,
+ 0x99E9E31650C1217BULL, 0x51AA691E0E423CFCULL, 0x1C217E6C3826E52CULL,
+ 0x51A8A93109703FEEULL, 0xBB7099876A460E74ULL, 0x541FC68C9C86B022ULL,
+ 0x59160CC046FD8251ULL, 0x2846C0BA35C35F5CULL, 0x54504AC78B758282ULL,
+ 0x29388839D2AF05E4ULL, 0xCB2C0F1CC01BD702ULL, 0x555B2F747C932665ULL,
+ 0x86B63142A3AB8829ULL, 0x0B8CC3BDF64B10EFULL, 0x687FEB69EDD1CC5EULL,
+ 0xFDB23FCEC9509D43ULL, 0x1E425A31D951AE64ULL, 0x36AD004CF600C838ULL,
+ 0xA40E329CCFF46AAAULL, 0xA41D570D7938DAD4ULL, 0x62A69526D43161C1ULL,
+ 0x3FDD4A8E9ADB1E69ULL, 0x5B3B71F9DC6B80D6ULL, 0xEC9D1810C6272B04ULL,
+ 0x8CCF2DD5CACEF403ULL, 0xE49F5235C95B9117ULL, 0x505DC82DB854338AULL,
+ 0x62292C311562A846ULL, 0xD72B03746AE77F5EULL, 0xF9C9091B462D538CULL,
+ 0x0AE8DB5847A67CBEULL, 0xB3A739C122611682ULL, 0xEEAAC0232A281BF6ULL,
+ 0x94C6651E77CAF992ULL, 0x763E4E4B94B2BBC1ULL, 0x587E38DA0077D9B4ULL,
+ 0x7FB29F8C183023C3ULL, 0x0ABEC1FFF9E3A26EULL, 0xA00EF092350511E3ULL,
+ 0xB855322EDB6340D8ULL, 0xA52471F7A9A96910ULL, 0x388147FB4CFDB477ULL,
+ 0x9B1F5C3E4E46041FULL, 0xCDAD0657FCCFEC71ULL, 0xB38E8C334C701C3AULL,
+ 0x917BDD64B1C0FD4CULL, 0x3BB454329B7624C8ULL, 0x23BA4442CAF53EA6ULL,
+ 0x4E677D2C38532A3AULL, 0x0BFD64B645036C7AULL, 0xC68A007E5E0DD902ULL,
+ 0x4DB5A851F44182E1ULL, 0x8EC9B55A7F88A46BULL, 0x0A8291CDCEC97DCFULL,
+ 0x2A4ECEA9F98D0ACCULL, 0x1A1DB93D7140003CULL, 0x092999A333CB8B7AULL,
+ 0x6DC778F971AD0038ULL, 0xA907600A918130C4ULL, 0xED6A1E012D9E6832ULL,
+ 0x7135C886EFB4318AULL, 0x87F55BA57E31CC7AULL, 0x7763CF1D55034004ULL,
+ 0xAC7D5F42D69F6D18ULL, 0x7930E9E4E58857B6ULL, 0x6E6F52C3164DF4FBULL,
+ 0x25E41D2B669E1EF1ULL, 0x3C1B20EE3FD59D7CULL, 0x0ABCD06BFA53DDEFULL,
+ 0x1DBF9A42D5C4484EULL, 0xABC521979B0DEADAULL, 0xE86D2BC522363A0DULL,
+ 0x5CAE82AB9C9DF69EULL, 0x64F2E21E71F54BFFULL, 0xF4FD4452E2D74DD3ULL,
+ 0xB4130C93BC437944ULL, 0xAEFE130985139270ULL, 0x598CB0FAC186D91CULL,
+ 0x7AD91D2691F7F7EEULL, 0x61B46FC9D6E6C907ULL, 0xBC34F4DEF99C0238ULL,
+ 0xDE355B3B6519035BULL, 0x886B4238611FCFDCULL, 0xC6F34A26C1B2EFFAULL,
+ 0xC58EF1837D1683B2ULL, 0x3BB5FCBC2EC22005ULL, 0xC3FE3B1B4C6FAD73ULL,
+ 0x8E4F1232EEF28183ULL, 0x9172FE9CE98583FFULL, 0xC03404CD28342F61ULL,
+ 0x9E02FCE1CDF7E2ECULL, 0x0B07A7C8EE0A6D70ULL, 0xAE56EDE76372BB19ULL,
+ 0x1D4F42A3DE394DF4ULL, 0xB96ADAB760D7F468ULL, 0xD108A94BB2C8E3FBULL,
+ 0xBC0AB182B324FB61ULL, 0x30ACCA4F483A797AULL, 0x1DF158A136ADE735ULL,
+ 0xE2A689DAF3EFE872ULL, 0x984F0C70E0E68B77ULL, 0xB557135E7F57C935ULL,
+ 0x856365553DED1AF3ULL, 0x2433F51F5F066ED0ULL, 0xD3DF1ED5D5FD6561ULL,
+ 0xF681B202AEC4617AULL, 0x7D2FE363630C75D8ULL, 0xCC939DCE249B3EF9ULL,
+ 0xA9E13641146433FBULL, 0xD8B9C583CE2D3695ULL, 0xAFDC5620273D3CF1ULL,
+ 0xADF85458A2BB4A9AULL, 0xFFFFFFFFFFFFFFFFULL
+};
+
+# elif BN_BITS2 == 32
+
+static const BN_ULONG dh1024_160_p[] = {
+ 0x2E4A4371, 0xDF1FB2BC, 0x6D4DA708, 0xE68CFDA7, 0x365C1A65, 0x45BF37DF,
+ 0x0DC8B4BD, 0xA151AF5F, 0xF55BCCC0, 0xFAA31A4F, 0xE5644738, 0x4EFFD6FA,
+ 0x219A7372, 0x98488E9C, 0x90C4BD70, 0xACCBDD7D, 0xD49B83BF, 0x24975C3C,
+ 0xA9061123, 0x13ECB4AE, 0x2EE652C0, 0x9838EF1E, 0x75A23D18, 0x6073E286,
+ 0x52D23B61, 0x9A6A9DCA, 0xFB06A3C6, 0x52C99FBC, 0xAE5D54EC, 0xDE92DE5E,
+ 0xA080E01D, 0xB10B8F96
+};
+
+static const BN_ULONG dh1024_160_g[] = {
+ 0x22B3B2E5, 0x855E6EEB, 0xF97C2A24, 0x858F4DCE, 0x18D08BC8, 0x2D779D59,
+ 0x8E73AFA3, 0xD662A4D1, 0x69B6A28A, 0x1DBF0A01, 0x7A091F53, 0xA6A24C08,
+ 0x63F80A76, 0x909D0D22, 0xB9A92EE1, 0xD7FBD7D3, 0x9E2749F4, 0x5E91547F,
+ 0xB01B886A, 0x160217B4, 0x5504F213, 0x777E690F, 0x5C41564B, 0x266FEA1E,
+ 0x14266D31, 0xD6406CFF, 0x58AC507F, 0xF8104DD2, 0xEFB99905, 0x6765A442,
+ 0xC3FD3412, 0xA4D1CBD5
+};
+
+static const BN_ULONG dh1024_160_q[] = {
+ 0x49462353, 0x64B7CB9D, 0x8ABA4E7D, 0x81A8DF27, 0xF518AA87
+};
+
+static const BN_ULONG dh2048_224_p[] = {
+ 0x0C10E64F, 0x0AC4DFFE, 0x4E71B81C, 0xCF9DE538, 0xFFA31F71, 0x7EF363E2,
+ 0x6B8E75B9, 0xE3FB73C1, 0x4BA80A29, 0xC9B53DCF, 0x16E79763, 0x23F10B0E,
+ 0x13042E9B, 0xC52172E4, 0xC928B2B9, 0xBE60E69C, 0xB9E587E8, 0x80CD86A1,
+ 0x98C641A4, 0x315D75E1, 0x44328387, 0xCDF93ACC, 0xDC0A486D, 0x15987D9A,
+ 0x1FD5A074, 0x7310F712, 0xDE31EFDC, 0x278273C7, 0x415D9330, 0x1602E714,
+ 0xBC8985DB, 0x81286130, 0x70918836, 0xB3BF8A31, 0xB9C49708, 0x6A00E0A0,
+ 0x8BBC27BE, 0xC6BA0B2C, 0xED34DBF6, 0xC9F98D11, 0xB6C12207, 0x7AD5B7D0,
+ 0x55B7394B, 0xD91E8FEF, 0xEFDA4DF8, 0x9037C9ED, 0xAD6AC212, 0x6D3F8152,
+ 0x1274A0A6, 0x1DE6B85A, 0x309C180E, 0xEB3D688A, 0x7BA1DF15, 0xAF9A3C40,
+ 0xF95A56DB, 0xE6FA141D, 0xB61D0A75, 0xB54B1597, 0x683B9FD1, 0xA20D64E5,
+ 0x9559C51F, 0xD660FAA7, 0x9123A9D0, 0xAD107E1E
+};
+
+static const BN_ULONG dh2048_224_g[] = {
+ 0x191F2BFA, 0x84B890D3, 0x2A7065B3, 0x81BC087F, 0xF6EC0179, 0x19C418E1,
+ 0x71CFFF4C, 0x7B5A0F1C, 0x9B6AA4BD, 0xEDFE72FE, 0x94B30269, 0x81E1BCFE,
+ 0x8D6C0191, 0x566AFBB4, 0x409D13CD, 0xB539CCE3, 0x5F2FF381, 0x6AA21E7F,
+ 0x770589EF, 0xD9E263E4, 0xD19963DD, 0x10E183ED, 0x150B8EEB, 0xB70A8137,
+ 0x28C8F8AC, 0x051AE3D4, 0x0C1AB15B, 0xBB77A86F, 0x16A330EF, 0x6E3025E3,
+ 0xD6F83456, 0x19529A45, 0x118E98D1, 0xF180EB34, 0x50717CBE, 0xB5F6C6B2,
+ 0xDA7460CD, 0x09939D54, 0x22EA1ED4, 0xE2471504, 0x521BC98A, 0xB8A762D0,
+ 0x5AC1348B, 0xF4D02727, 0x1999024A, 0xC1766910, 0xA8D66AD7, 0xBE5E9001,
+ 0x620A8652, 0xC57DB17C, 0x00C29F52, 0xAB739D77, 0xA70C4AFA, 0xDD921F01,
+ 0x10B9A6F0, 0xA6824A4E, 0xCFE4FFE3, 0x74866A08, 0x89998CAF, 0x6CDEBE7B,
+ 0x8FFDAC50, 0x9DF30B5C, 0x4F2D9AE3, 0xAC4032EF
+};
+
+static const BN_ULONG dh2048_224_q[] = {
+ 0xB36371EB, 0xBF389A99, 0x4738CEBC, 0x1F80535A, 0x99717710, 0xC58D93FE,
+ 0x801C0D34
+};
+
+static const BN_ULONG dh2048_256_p[] = {
+ 0x1E1A1597, 0xDB094AE9, 0xD7EF09CA, 0x693877FA, 0x6E11715F, 0x6116D227,
+ 0xC198AF12, 0xA4B54330, 0xD7014103, 0x75F26375, 0x54E710C3, 0xC3A3960A,
+ 0xBD0BE621, 0xDED4010A, 0x89962856, 0xC0B857F6, 0x71506026, 0xB3CA3F79,
+ 0xE6B486F6, 0x1CCACB83, 0x14056425, 0x67E144E5, 0xA41825D9, 0xF6A167B5,
+ 0x96524D8E, 0x3AD83477, 0x51BFA4AB, 0xF13C6D9A, 0x35488A0E, 0x2D525267,
+ 0xCAA6B790, 0xB63ACAE1, 0x81B23F76, 0x4FDB70C5, 0x12307F5C, 0xBC39A0BF,
+ 0xB1E59BB8, 0xB941F54E, 0xD45F9088, 0x6C5BFC11, 0x4275BF7B, 0x22E0B1EF,
+ 0x5B4758C0, 0x91F9E672, 0x6BCF67ED, 0x5A8A9D30, 0x97517ABD, 0x209E0C64,
+ 0x830E9A7C, 0x3BF4296D, 0x34096FAA, 0x16C3D911, 0x61B2AA30, 0xFAF7DF45,
+ 0xD61957D4, 0xE00DF8F1, 0x435E3B00, 0x5D2CEED4, 0x660DD0F2, 0x8CEEF608,
+ 0x65195999, 0xFFBBD19C, 0xB4B6663C, 0x87A8E61D
+};
+
+static const BN_ULONG dh2048_256_g[] = {
+ 0x6CC41659, 0x664B4C0F, 0xEF98C582, 0x5E2327CF, 0xD4795451, 0xD647D148,
+ 0x90F00EF8, 0x2F630784, 0x1DB246C3, 0x184B523D, 0xCDC67EB6, 0xC7891428,
+ 0x0DF92B52, 0x7FD02837, 0x64E0EC37, 0xB3353BBB, 0x57CD0915, 0xECD06E15,
+ 0xDF016199, 0xB7D2BBD2, 0x052588B9, 0xC8484B1E, 0x13D3FE14, 0xDB2A3B73,
+ 0xD182EA0A, 0xD052B985, 0xE83B9C80, 0xA4BD1BFF, 0xFB3F2E55, 0xDFC967C1,
+ 0x767164E1, 0xB5045AF2, 0x6F2F9193, 0x1D14348F, 0x428EBC83, 0x64E67982,
+ 0x82D6ED38, 0x8AC376D2, 0xAAB8A862, 0x777DE62A, 0xE9EC144B, 0xDDF463E5,
+ 0xC77A57F2, 0x0196F931, 0x41000A65, 0xA55AE313, 0xC28CBB18, 0x901228F8,
+ 0x7E8C6F62, 0xBC3773BF, 0x0C6B47B1, 0xBE3A6C1B, 0xAC0BB555, 0xFF4FED4A,
+ 0x77BE463F, 0x10DBC150, 0x1A0BA125, 0x07F4793A, 0x21EF2054, 0x4CA7B18F,
+ 0x60EDBD48, 0x2E775066, 0x73134D0B, 0x3FB32C9B
+};
+
+static const BN_ULONG dh2048_256_q[] = {
+ 0x64F5FBD3, 0xA308B0FE, 0x1EB3750B, 0x99B1A47D, 0x40129DA2, 0xB4479976,
+ 0xA709A097, 0x8CF83642
+};
+
+/* Primes from RFC 7919 */
+
+static const BN_ULONG ffdhe2048_p[] = {
+ 0xFFFFFFFF, 0xFFFFFFFF, 0x61285C97, 0x886B4238, 0xC1B2EFFA, 0xC6F34A26,
+ 0x7D1683B2, 0xC58EF183, 0x2EC22005, 0x3BB5FCBC, 0x4C6FAD73, 0xC3FE3B1B,
+ 0xEEF28183, 0x8E4F1232, 0xE98583FF, 0x9172FE9C, 0x28342F61, 0xC03404CD,
+ 0xCDF7E2EC, 0x9E02FCE1, 0xEE0A6D70, 0x0B07A7C8, 0x6372BB19, 0xAE56EDE7,
+ 0xDE394DF4, 0x1D4F42A3, 0x60D7F468, 0xB96ADAB7, 0xB2C8E3FB, 0xD108A94B,
+ 0xB324FB61, 0xBC0AB182, 0x483A797A, 0x30ACCA4F, 0x36ADE735, 0x1DF158A1,
+ 0xF3EFE872, 0xE2A689DA, 0xE0E68B77, 0x984F0C70, 0x7F57C935, 0xB557135E,
+ 0x3DED1AF3, 0x85636555, 0x5F066ED0, 0x2433F51F, 0xD5FD6561, 0xD3DF1ED5,
+ 0xAEC4617A, 0xF681B202, 0x630C75D8, 0x7D2FE363, 0x249B3EF9, 0xCC939DCE,
+ 0x146433FB, 0xA9E13641, 0xCE2D3695, 0xD8B9C583, 0x273D3CF1, 0xAFDC5620,
+ 0xA2BB4A9A, 0xADF85458, 0xFFFFFFFF, 0xFFFFFFFF
+};
+
+static const BN_ULONG ffdhe3072_p[] = {
+ 0xFFFFFFFF, 0xFFFFFFFF, 0x66C62E37, 0x25E41D2B, 0x3FD59D7C, 0x3C1B20EE,
+ 0xFA53DDEF, 0x0ABCD06B, 0xD5C4484E, 0x1DBF9A42, 0x9B0DEADA, 0xABC52197,
+ 0x22363A0D, 0xE86D2BC5, 0x9C9DF69E, 0x5CAE82AB, 0x71F54BFF, 0x64F2E21E,
+ 0xE2D74DD3, 0xF4FD4452, 0xBC437944, 0xB4130C93, 0x85139270, 0xAEFE1309,
+ 0xC186D91C, 0x598CB0FA, 0x91F7F7EE, 0x7AD91D26, 0xD6E6C907, 0x61B46FC9,
+ 0xF99C0238, 0xBC34F4DE, 0x6519035B, 0xDE355B3B, 0x611FCFDC, 0x886B4238,
+ 0xC1B2EFFA, 0xC6F34A26, 0x7D1683B2, 0xC58EF183, 0x2EC22005, 0x3BB5FCBC,
+ 0x4C6FAD73, 0xC3FE3B1B, 0xEEF28183, 0x8E4F1232, 0xE98583FF, 0x9172FE9C,
+ 0x28342F61, 0xC03404CD, 0xCDF7E2EC, 0x9E02FCE1, 0xEE0A6D70, 0x0B07A7C8,
+ 0x6372BB19, 0xAE56EDE7, 0xDE394DF4, 0x1D4F42A3, 0x60D7F468, 0xB96ADAB7,
+ 0xB2C8E3FB, 0xD108A94B, 0xB324FB61, 0xBC0AB182, 0x483A797A, 0x30ACCA4F,
+ 0x36ADE735, 0x1DF158A1, 0xF3EFE872, 0xE2A689DA, 0xE0E68B77, 0x984F0C70,
+ 0x7F57C935, 0xB557135E, 0x3DED1AF3, 0x85636555, 0x5F066ED0, 0x2433F51F,
+ 0xD5FD6561, 0xD3DF1ED5, 0xAEC4617A, 0xF681B202, 0x630C75D8, 0x7D2FE363,
+ 0x249B3EF9, 0xCC939DCE, 0x146433FB, 0xA9E13641, 0xCE2D3695, 0xD8B9C583,
+ 0x273D3CF1, 0xAFDC5620, 0xA2BB4A9A, 0xADF85458, 0xFFFFFFFF, 0xFFFFFFFF
+};
+
+static const BN_ULONG ffdhe4096_p[] = {
+ 0xFFFFFFFF, 0xFFFFFFFF, 0x5E655F6A, 0xC68A007E, 0xF44182E1, 0x4DB5A851,
+ 0x7F88A46B, 0x8EC9B55A, 0xCEC97DCF, 0x0A8291CD, 0xF98D0ACC, 0x2A4ECEA9,
+ 0x7140003C, 0x1A1DB93D, 0x33CB8B7A, 0x092999A3, 0x71AD0038, 0x6DC778F9,
+ 0x918130C4, 0xA907600A, 0x2D9E6832, 0xED6A1E01, 0xEFB4318A, 0x7135C886,
+ 0x7E31CC7A, 0x87F55BA5, 0x55034004, 0x7763CF1D, 0xD69F6D18, 0xAC7D5F42,
+ 0xE58857B6, 0x7930E9E4, 0x164DF4FB, 0x6E6F52C3, 0x669E1EF1, 0x25E41D2B,
+ 0x3FD59D7C, 0x3C1B20EE, 0xFA53DDEF, 0x0ABCD06B, 0xD5C4484E, 0x1DBF9A42,
+ 0x9B0DEADA, 0xABC52197, 0x22363A0D, 0xE86D2BC5, 0x9C9DF69E, 0x5CAE82AB,
+ 0x71F54BFF, 0x64F2E21E, 0xE2D74DD3, 0xF4FD4452, 0xBC437944, 0xB4130C93,
+ 0x85139270, 0xAEFE1309, 0xC186D91C, 0x598CB0FA, 0x91F7F7EE, 0x7AD91D26,
+ 0xD6E6C907, 0x61B46FC9, 0xF99C0238, 0xBC34F4DE, 0x6519035B, 0xDE355B3B,
+ 0x611FCFDC, 0x886B4238, 0xC1B2EFFA, 0xC6F34A26, 0x7D1683B2, 0xC58EF183,
+ 0x2EC22005, 0x3BB5FCBC, 0x4C6FAD73, 0xC3FE3B1B, 0xEEF28183, 0x8E4F1232,
+ 0xE98583FF, 0x9172FE9C, 0x28342F61, 0xC03404CD, 0xCDF7E2EC, 0x9E02FCE1,
+ 0xEE0A6D70, 0x0B07A7C8, 0x6372BB19, 0xAE56EDE7, 0xDE394DF4, 0x1D4F42A3,
+ 0x60D7F468, 0xB96ADAB7, 0xB2C8E3FB, 0xD108A94B, 0xB324FB61, 0xBC0AB182,
+ 0x483A797A, 0x30ACCA4F, 0x36ADE735, 0x1DF158A1, 0xF3EFE872, 0xE2A689DA,
+ 0xE0E68B77, 0x984F0C70, 0x7F57C935, 0xB557135E, 0x3DED1AF3, 0x85636555,
+ 0x5F066ED0, 0x2433F51F, 0xD5FD6561, 0xD3DF1ED5, 0xAEC4617A, 0xF681B202,
+ 0x630C75D8, 0x7D2FE363, 0x249B3EF9, 0xCC939DCE, 0x146433FB, 0xA9E13641,
+ 0xCE2D3695, 0xD8B9C583, 0x273D3CF1, 0xAFDC5620, 0xA2BB4A9A, 0xADF85458,
+ 0xFFFFFFFF, 0xFFFFFFFF
+};
+
+static const BN_ULONG ffdhe6144_p[] = {
+ 0xFFFFFFFF, 0xFFFFFFFF, 0xD0E40E65, 0xA40E329C, 0x7938DAD4, 0xA41D570D,
+ 0xD43161C1, 0x62A69526, 0x9ADB1E69, 0x3FDD4A8E, 0xDC6B80D6, 0x5B3B71F9,
+ 0xC6272B04, 0xEC9D1810, 0xCACEF403, 0x8CCF2DD5, 0xC95B9117, 0xE49F5235,
+ 0xB854338A, 0x505DC82D, 0x1562A846, 0x62292C31, 0x6AE77F5E, 0xD72B0374,
+ 0x462D538C, 0xF9C9091B, 0x47A67CBE, 0x0AE8DB58, 0x22611682, 0xB3A739C1,
+ 0x2A281BF6, 0xEEAAC023, 0x77CAF992, 0x94C6651E, 0x94B2BBC1, 0x763E4E4B,
+ 0x0077D9B4, 0x587E38DA, 0x183023C3, 0x7FB29F8C, 0xF9E3A26E, 0x0ABEC1FF,
+ 0x350511E3, 0xA00EF092, 0xDB6340D8, 0xB855322E, 0xA9A96910, 0xA52471F7,
+ 0x4CFDB477, 0x388147FB, 0x4E46041F, 0x9B1F5C3E, 0xFCCFEC71, 0xCDAD0657,
+ 0x4C701C3A, 0xB38E8C33, 0xB1C0FD4C, 0x917BDD64, 0x9B7624C8, 0x3BB45432,
+ 0xCAF53EA6, 0x23BA4442, 0x38532A3A, 0x4E677D2C, 0x45036C7A, 0x0BFD64B6,
+ 0x5E0DD902, 0xC68A007E, 0xF44182E1, 0x4DB5A851, 0x7F88A46B, 0x8EC9B55A,
+ 0xCEC97DCF, 0x0A8291CD, 0xF98D0ACC, 0x2A4ECEA9, 0x7140003C, 0x1A1DB93D,
+ 0x33CB8B7A, 0x092999A3, 0x71AD0038, 0x6DC778F9, 0x918130C4, 0xA907600A,
+ 0x2D9E6832, 0xED6A1E01, 0xEFB4318A, 0x7135C886, 0x7E31CC7A, 0x87F55BA5,
+ 0x55034004, 0x7763CF1D, 0xD69F6D18, 0xAC7D5F42, 0xE58857B6, 0x7930E9E4,
+ 0x164DF4FB, 0x6E6F52C3, 0x669E1EF1, 0x25E41D2B, 0x3FD59D7C, 0x3C1B20EE,
+ 0xFA53DDEF, 0x0ABCD06B, 0xD5C4484E, 0x1DBF9A42, 0x9B0DEADA, 0xABC52197,
+ 0x22363A0D, 0xE86D2BC5, 0x9C9DF69E, 0x5CAE82AB, 0x71F54BFF, 0x64F2E21E,
+ 0xE2D74DD3, 0xF4FD4452, 0xBC437944, 0xB4130C93, 0x85139270, 0xAEFE1309,
+ 0xC186D91C, 0x598CB0FA, 0x91F7F7EE, 0x7AD91D26, 0xD6E6C907, 0x61B46FC9,
+ 0xF99C0238, 0xBC34F4DE, 0x6519035B, 0xDE355B3B, 0x611FCFDC, 0x886B4238,
+ 0xC1B2EFFA, 0xC6F34A26, 0x7D1683B2, 0xC58EF183, 0x2EC22005, 0x3BB5FCBC,
+ 0x4C6FAD73, 0xC3FE3B1B, 0xEEF28183, 0x8E4F1232, 0xE98583FF, 0x9172FE9C,
+ 0x28342F61, 0xC03404CD, 0xCDF7E2EC, 0x9E02FCE1, 0xEE0A6D70, 0x0B07A7C8,
+ 0x6372BB19, 0xAE56EDE7, 0xDE394DF4, 0x1D4F42A3, 0x60D7F468, 0xB96ADAB7,
+ 0xB2C8E3FB, 0xD108A94B, 0xB324FB61, 0xBC0AB182, 0x483A797A, 0x30ACCA4F,
+ 0x36ADE735, 0x1DF158A1, 0xF3EFE872, 0xE2A689DA, 0xE0E68B77, 0x984F0C70,
+ 0x7F57C935, 0xB557135E, 0x3DED1AF3, 0x85636555, 0x5F066ED0, 0x2433F51F,
+ 0xD5FD6561, 0xD3DF1ED5, 0xAEC4617A, 0xF681B202, 0x630C75D8, 0x7D2FE363,
+ 0x249B3EF9, 0xCC939DCE, 0x146433FB, 0xA9E13641, 0xCE2D3695, 0xD8B9C583,
+ 0x273D3CF1, 0xAFDC5620, 0xA2BB4A9A, 0xADF85458, 0xFFFFFFFF, 0xFFFFFFFF
+};
+
+static const BN_ULONG ffdhe8192_p[] = {
+ 0xFFFFFFFF, 0xFFFFFFFF, 0xC5C6424C, 0xD68C8BB7, 0x838FF88C, 0x011E2A94,
+ 0xA9F4614E, 0x0822E506, 0xF7A8443D, 0x97D11D49, 0x30677F0D, 0xA6BBFDE5,
+ 0xC1FE86FE, 0x2F741EF8, 0x5D71A87E, 0xFAFABE1C, 0xFBE58A30, 0xDED2FBAB,
+ 0x72B0A66E, 0xB6855DFE, 0xBA8A4FE8, 0x1EFC8CE0, 0x3F2FA457, 0x83F81D4A,
+ 0xA577E231, 0xA1FE3075, 0x88D9C0A0, 0xD5B80194, 0xAD9A95F9, 0x624816CD,
+ 0x50C1217B, 0x99E9E316, 0x0E423CFC, 0x51AA691E, 0x3826E52C, 0x1C217E6C,
+ 0x09703FEE, 0x51A8A931, 0x6A460E74, 0xBB709987, 0x9C86B022, 0x541FC68C,
+ 0x46FD8251, 0x59160CC0, 0x35C35F5C, 0x2846C0BA, 0x8B758282, 0x54504AC7,
+ 0xD2AF05E4, 0x29388839, 0xC01BD702, 0xCB2C0F1C, 0x7C932665, 0x555B2F74,
+ 0xA3AB8829, 0x86B63142, 0xF64B10EF, 0x0B8CC3BD, 0xEDD1CC5E, 0x687FEB69,
+ 0xC9509D43, 0xFDB23FCE, 0xD951AE64, 0x1E425A31, 0xF600C838, 0x36AD004C,
+ 0xCFF46AAA, 0xA40E329C, 0x7938DAD4, 0xA41D570D, 0xD43161C1, 0x62A69526,
+ 0x9ADB1E69, 0x3FDD4A8E, 0xDC6B80D6, 0x5B3B71F9, 0xC6272B04, 0xEC9D1810,
+ 0xCACEF403, 0x8CCF2DD5, 0xC95B9117, 0xE49F5235, 0xB854338A, 0x505DC82D,
+ 0x1562A846, 0x62292C31, 0x6AE77F5E, 0xD72B0374, 0x462D538C, 0xF9C9091B,
+ 0x47A67CBE, 0x0AE8DB58, 0x22611682, 0xB3A739C1, 0x2A281BF6, 0xEEAAC023,
+ 0x77CAF992, 0x94C6651E, 0x94B2BBC1, 0x763E4E4B, 0x0077D9B4, 0x587E38DA,
+ 0x183023C3, 0x7FB29F8C, 0xF9E3A26E, 0x0ABEC1FF, 0x350511E3, 0xA00EF092,
+ 0xDB6340D8, 0xB855322E, 0xA9A96910, 0xA52471F7, 0x4CFDB477, 0x388147FB,
+ 0x4E46041F, 0x9B1F5C3E, 0xFCCFEC71, 0xCDAD0657, 0x4C701C3A, 0xB38E8C33,
+ 0xB1C0FD4C, 0x917BDD64, 0x9B7624C8, 0x3BB45432, 0xCAF53EA6, 0x23BA4442,
+ 0x38532A3A, 0x4E677D2C, 0x45036C7A, 0x0BFD64B6, 0x5E0DD902, 0xC68A007E,
+ 0xF44182E1, 0x4DB5A851, 0x7F88A46B, 0x8EC9B55A, 0xCEC97DCF, 0x0A8291CD,
+ 0xF98D0ACC, 0x2A4ECEA9, 0x7140003C, 0x1A1DB93D, 0x33CB8B7A, 0x092999A3,
+ 0x71AD0038, 0x6DC778F9, 0x918130C4, 0xA907600A, 0x2D9E6832, 0xED6A1E01,
+ 0xEFB4318A, 0x7135C886, 0x7E31CC7A, 0x87F55BA5, 0x55034004, 0x7763CF1D,
+ 0xD69F6D18, 0xAC7D5F42, 0xE58857B6, 0x7930E9E4, 0x164DF4FB, 0x6E6F52C3,
+ 0x669E1EF1, 0x25E41D2B, 0x3FD59D7C, 0x3C1B20EE, 0xFA53DDEF, 0x0ABCD06B,
+ 0xD5C4484E, 0x1DBF9A42, 0x9B0DEADA, 0xABC52197, 0x22363A0D, 0xE86D2BC5,
+ 0x9C9DF69E, 0x5CAE82AB, 0x71F54BFF, 0x64F2E21E, 0xE2D74DD3, 0xF4FD4452,
+ 0xBC437944, 0xB4130C93, 0x85139270, 0xAEFE1309, 0xC186D91C, 0x598CB0FA,
+ 0x91F7F7EE, 0x7AD91D26, 0xD6E6C907, 0x61B46FC9, 0xF99C0238, 0xBC34F4DE,
+ 0x6519035B, 0xDE355B3B, 0x611FCFDC, 0x886B4238, 0xC1B2EFFA, 0xC6F34A26,
+ 0x7D1683B2, 0xC58EF183, 0x2EC22005, 0x3BB5FCBC, 0x4C6FAD73, 0xC3FE3B1B,
+ 0xEEF28183, 0x8E4F1232, 0xE98583FF, 0x9172FE9C, 0x28342F61, 0xC03404CD,
+ 0xCDF7E2EC, 0x9E02FCE1, 0xEE0A6D70, 0x0B07A7C8, 0x6372BB19, 0xAE56EDE7,
+ 0xDE394DF4, 0x1D4F42A3, 0x60D7F468, 0xB96ADAB7, 0xB2C8E3FB, 0xD108A94B,
+ 0xB324FB61, 0xBC0AB182, 0x483A797A, 0x30ACCA4F, 0x36ADE735, 0x1DF158A1,
+ 0xF3EFE872, 0xE2A689DA, 0xE0E68B77, 0x984F0C70, 0x7F57C935, 0xB557135E,
+ 0x3DED1AF3, 0x85636555, 0x5F066ED0, 0x2433F51F, 0xD5FD6561, 0xD3DF1ED5,
+ 0xAEC4617A, 0xF681B202, 0x630C75D8, 0x7D2FE363, 0x249B3EF9, 0xCC939DCE,
+ 0x146433FB, 0xA9E13641, 0xCE2D3695, 0xD8B9C583, 0x273D3CF1, 0xAFDC5620,
+ 0xA2BB4A9A, 0xADF85458, 0xFFFFFFFF, 0xFFFFFFFF
+};
+
+# else
+# error "unsupported BN_BITS2"
+# endif
+
+/* Macro to make a BIGNUM from static data */
+
+# define make_dh_bn(x) extern const BIGNUM _bignum_##x; \
+ const BIGNUM _bignum_##x = { (BN_ULONG *) x, \
+ OSSL_NELEM(x),\
+ OSSL_NELEM(x),\
+ 0, BN_FLG_STATIC_DATA };
+
+static const BN_ULONG value_2 = 2;
+
+const BIGNUM _bignum_const_2 =
+ { (BN_ULONG *)&value_2, 1, 1, 0, BN_FLG_STATIC_DATA };
+
+make_dh_bn(dh1024_160_p)
+make_dh_bn(dh1024_160_g)
+make_dh_bn(dh1024_160_q)
+make_dh_bn(dh2048_224_p)
+make_dh_bn(dh2048_224_g)
+make_dh_bn(dh2048_224_q)
+make_dh_bn(dh2048_256_p)
+make_dh_bn(dh2048_256_g)
+make_dh_bn(dh2048_256_q)
+
+make_dh_bn(ffdhe2048_p)
+make_dh_bn(ffdhe3072_p)
+make_dh_bn(ffdhe4096_p)
+make_dh_bn(ffdhe6144_p)
+make_dh_bn(ffdhe8192_p)
+
+
+#endif
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_div.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_div.c
new file mode 100644
index 0000000..4273618
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_div.c
@@ -0,0 +1,460 @@
+/*
+ * Copyright 1995-2022 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include <assert.h>
+#include <openssl/bn.h>
+#include "internal/cryptlib.h"
+#include "bn_local.h"
+
+/* The old slow way */
+#if 0
+int BN_div(BIGNUM *dv, BIGNUM *rem, const BIGNUM *m, const BIGNUM *d,
+ BN_CTX *ctx)
+{
+ int i, nm, nd;
+ int ret = 0;
+ BIGNUM *D;
+
+ bn_check_top(m);
+ bn_check_top(d);
+ if (BN_is_zero(d)) {
+ BNerr(BN_F_BN_DIV, BN_R_DIV_BY_ZERO);
+ return 0;
+ }
+
+ if (BN_ucmp(m, d) < 0) {
+ if (rem != NULL) {
+ if (BN_copy(rem, m) == NULL)
+ return 0;
+ }
+ if (dv != NULL)
+ BN_zero(dv);
+ return 1;
+ }
+
+ BN_CTX_start(ctx);
+ D = BN_CTX_get(ctx);
+ if (dv == NULL)
+ dv = BN_CTX_get(ctx);
+ if (rem == NULL)
+ rem = BN_CTX_get(ctx);
+ if (D == NULL || dv == NULL || rem == NULL)
+ goto end;
+
+ nd = BN_num_bits(d);
+ nm = BN_num_bits(m);
+ if (BN_copy(D, d) == NULL)
+ goto end;
+ if (BN_copy(rem, m) == NULL)
+ goto end;
+
+ /*
+ * The next 2 are needed so we can do a dv->d[0]|=1 later since
+ * BN_lshift1 will only work once there is a value :-)
+ */
+ BN_zero(dv);
+ if (bn_wexpand(dv, 1) == NULL)
+ goto end;
+ dv->top = 1;
+
+ if (!BN_lshift(D, D, nm - nd))
+ goto end;
+ for (i = nm - nd; i >= 0; i--) {
+ if (!BN_lshift1(dv, dv))
+ goto end;
+ if (BN_ucmp(rem, D) >= 0) {
+ dv->d[0] |= 1;
+ if (!BN_usub(rem, rem, D))
+ goto end;
+ }
+/* CAN IMPROVE (and have now :=) */
+ if (!BN_rshift1(D, D))
+ goto end;
+ }
+ rem->neg = BN_is_zero(rem) ? 0 : m->neg;
+ dv->neg = m->neg ^ d->neg;
+ ret = 1;
+ end:
+ BN_CTX_end(ctx);
+ return ret;
+}
+
+#else
+
+# if defined(BN_DIV3W)
+BN_ULONG bn_div_3_words(const BN_ULONG *m, BN_ULONG d1, BN_ULONG d0);
+# elif 0
+/*
+ * This is #if-ed away, because it's a reference for assembly implementations,
+ * where it can and should be made constant-time. But if you want to test it,
+ * just replace 0 with 1.
+ */
+# if BN_BITS2 == 64 && defined(__SIZEOF_INT128__) && __SIZEOF_INT128__==16
+# undef BN_ULLONG
+# define BN_ULLONG __uint128_t
+# define BN_LLONG
+# endif
+
+# ifdef BN_LLONG
+# define BN_DIV3W
+/*
+ * Interface is somewhat quirky, |m| is pointer to most significant limb,
+ * and less significant limb is referred at |m[-1]|. This means that caller
+ * is responsible for ensuring that |m[-1]| is valid. Second condition that
+ * has to be met is that |d0|'s most significant bit has to be set. Or in
+ * other words divisor has to be "bit-aligned to the left." bn_div_fixed_top
+ * does all this. The subroutine considers four limbs, two of which are
+ * "overlapping," hence the name...
+ */
+static BN_ULONG bn_div_3_words(const BN_ULONG *m, BN_ULONG d1, BN_ULONG d0)
+{
+ BN_ULLONG R = ((BN_ULLONG)m[0] << BN_BITS2) | m[-1];
+ BN_ULLONG D = ((BN_ULLONG)d0 << BN_BITS2) | d1;
+ BN_ULONG Q = 0, mask;
+ int i;
+
+ for (i = 0; i < BN_BITS2; i++) {
+ Q <<= 1;
+ if (R >= D) {
+ Q |= 1;
+ R -= D;
+ }
+ D >>= 1;
+ }
+
+ mask = 0 - (Q >> (BN_BITS2 - 1)); /* does it overflow? */
+
+ Q <<= 1;
+ Q |= (R >= D);
+
+ return (Q | mask) & BN_MASK2;
+}
+# endif
+# endif
+
+static int bn_left_align(BIGNUM *num)
+{
+ BN_ULONG *d = num->d, n, m, rmask;
+ int top = num->top;
+ int rshift = BN_num_bits_word(d[top - 1]), lshift, i;
+
+ lshift = BN_BITS2 - rshift;
+ rshift %= BN_BITS2; /* say no to undefined behaviour */
+ rmask = (BN_ULONG)0 - rshift; /* rmask = 0 - (rshift != 0) */
+ rmask |= rmask >> 8;
+
+ for (i = 0, m = 0; i < top; i++) {
+ n = d[i];
+ d[i] = ((n << lshift) | m) & BN_MASK2;
+ m = (n >> rshift) & rmask;
+ }
+
+ return lshift;
+}
+
+# if !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM) \
+ && !defined(PEDANTIC) && !defined(BN_DIV3W)
+# if defined(__GNUC__) && __GNUC__>=2
+# if defined(__i386) || defined (__i386__)
+ /*-
+ * There were two reasons for implementing this template:
+ * - GNU C generates a call to a function (__udivdi3 to be exact)
+ * in reply to ((((BN_ULLONG)n0)<<BN_BITS2)|n1)/d0 (I fail to
+ * understand why...);
+ * - divl doesn't only calculate quotient, but also leaves
+ * remainder in %edx which we can definitely use here:-)
+ */
+# undef bn_div_words
+# define bn_div_words(n0,n1,d0) \
+ ({ asm volatile ( \
+ "divl %4" \
+ : "=a"(q), "=d"(rem) \
+ : "a"(n1), "d"(n0), "r"(d0) \
+ : "cc"); \
+ q; \
+ })
+# define REMAINDER_IS_ALREADY_CALCULATED
+# elif defined(__x86_64) && defined(SIXTY_FOUR_BIT_LONG)
+ /*
+ * Same story here, but it's 128-bit by 64-bit division. Wow!
+ */
+# undef bn_div_words
+# define bn_div_words(n0,n1,d0) \
+ ({ asm volatile ( \
+ "divq %4" \
+ : "=a"(q), "=d"(rem) \
+ : "a"(n1), "d"(n0), "r"(d0) \
+ : "cc"); \
+ q; \
+ })
+# define REMAINDER_IS_ALREADY_CALCULATED
+# endif /* __<cpu> */
+# endif /* __GNUC__ */
+# endif /* OPENSSL_NO_ASM */
+
+/*-
+ * BN_div computes dv := num / divisor, rounding towards
+ * zero, and sets up rm such that dv*divisor + rm = num holds.
+ * Thus:
+ * dv->neg == num->neg ^ divisor->neg (unless the result is zero)
+ * rm->neg == num->neg (unless the remainder is zero)
+ * If 'dv' or 'rm' is NULL, the respective value is not returned.
+ */
+int BN_div(BIGNUM *dv, BIGNUM *rm, const BIGNUM *num, const BIGNUM *divisor,
+ BN_CTX *ctx)
+{
+ int ret;
+
+ if (BN_is_zero(divisor)) {
+ BNerr(BN_F_BN_DIV, BN_R_DIV_BY_ZERO);
+ return 0;
+ }
+
+ /*
+ * Invalid zero-padding would have particularly bad consequences so don't
+ * just rely on bn_check_top() here (bn_check_top() works only for
+ * BN_DEBUG builds)
+ */
+ if (divisor->d[divisor->top - 1] == 0) {
+ BNerr(BN_F_BN_DIV, BN_R_NOT_INITIALIZED);
+ return 0;
+ }
+
+ ret = bn_div_fixed_top(dv, rm, num, divisor, ctx);
+
+ if (ret) {
+ if (dv != NULL)
+ bn_correct_top(dv);
+ if (rm != NULL)
+ bn_correct_top(rm);
+ }
+
+ return ret;
+}
+
+/*
+ * It's argued that *length* of *significant* part of divisor is public.
+ * Even if it's private modulus that is. Again, *length* is assumed
+ * public, but not *value*. Former is likely to be pre-defined by
+ * algorithm with bit granularity, though below subroutine is invariant
+ * of limb length. Thanks to this assumption we can require that |divisor|
+ * may not be zero-padded, yet claim this subroutine "constant-time"(*).
+ * This is because zero-padded dividend, |num|, is tolerated, so that
+ * caller can pass dividend of public length(*), but with smaller amount
+ * of significant limbs. This naturally means that quotient, |dv|, would
+ * contain correspongly less significant limbs as well, and will be zero-
+ * padded accordingly. Returned remainder, |rm|, will have same bit length
+ * as divisor, also zero-padded if needed. These actually leave sign bits
+ * in ambiguous state. In sense that we try to avoid negative zeros, while
+ * zero-padded zeros would retain sign.
+ *
+ * (*) "Constant-time-ness" has two pre-conditions:
+ *
+ * - availability of constant-time bn_div_3_words;
+ * - dividend is at least as "wide" as divisor, limb-wise, zero-padded
+ * if so required, which shouldn't be a privacy problem, because
+ * divisor's length is considered public;
+ */
+int bn_div_fixed_top(BIGNUM *dv, BIGNUM *rm, const BIGNUM *num,
+ const BIGNUM *divisor, BN_CTX *ctx)
+{
+ int norm_shift, i, j, loop;
+ BIGNUM *tmp, *snum, *sdiv, *res;
+ BN_ULONG *resp, *wnum, *wnumtop;
+ BN_ULONG d0, d1;
+ int num_n, div_n, num_neg;
+
+ assert(divisor->top > 0 && divisor->d[divisor->top - 1] != 0);
+
+ bn_check_top(num);
+ bn_check_top(divisor);
+ bn_check_top(dv);
+ bn_check_top(rm);
+
+ BN_CTX_start(ctx);
+ res = (dv == NULL) ? BN_CTX_get(ctx) : dv;
+ tmp = BN_CTX_get(ctx);
+ snum = BN_CTX_get(ctx);
+ sdiv = BN_CTX_get(ctx);
+ if (sdiv == NULL)
+ goto err;
+
+ /* First we normalise the numbers */
+ if (!BN_copy(sdiv, divisor))
+ goto err;
+ norm_shift = bn_left_align(sdiv);
+ sdiv->neg = 0;
+ /*
+ * Note that bn_lshift_fixed_top's output is always one limb longer
+ * than input, even when norm_shift is zero. This means that amount of
+ * inner loop iterations is invariant of dividend value, and that one
+ * doesn't need to compare dividend and divisor if they were originally
+ * of the same bit length.
+ */
+ if (!(bn_lshift_fixed_top(snum, num, norm_shift)))
+ goto err;
+
+ div_n = sdiv->top;
+ num_n = snum->top;
+
+ if (num_n <= div_n) {
+ /* caller didn't pad dividend -> no constant-time guarantee... */
+ if (bn_wexpand(snum, div_n + 1) == NULL)
+ goto err;
+ memset(&(snum->d[num_n]), 0, (div_n - num_n + 1) * sizeof(BN_ULONG));
+ snum->top = num_n = div_n + 1;
+ }
+
+ loop = num_n - div_n;
+ /*
+ * Lets setup a 'window' into snum This is the part that corresponds to
+ * the current 'area' being divided
+ */
+ wnum = &(snum->d[loop]);
+ wnumtop = &(snum->d[num_n - 1]);
+
+ /* Get the top 2 words of sdiv */
+ d0 = sdiv->d[div_n - 1];
+ d1 = (div_n == 1) ? 0 : sdiv->d[div_n - 2];
+
+ /* Setup quotient */
+ if (!bn_wexpand(res, loop))
+ goto err;
+ num_neg = num->neg;
+ res->neg = (num_neg ^ divisor->neg);
+ res->top = loop;
+ res->flags |= BN_FLG_FIXED_TOP;
+ resp = &(res->d[loop]);
+
+ /* space for temp */
+ if (!bn_wexpand(tmp, (div_n + 1)))
+ goto err;
+
+ for (i = 0; i < loop; i++, wnumtop--) {
+ BN_ULONG q, l0;
+ /*
+ * the first part of the loop uses the top two words of snum and sdiv
+ * to calculate a BN_ULONG q such that | wnum - sdiv * q | < sdiv
+ */
+# if defined(BN_DIV3W)
+ q = bn_div_3_words(wnumtop, d1, d0);
+# else
+ BN_ULONG n0, n1, rem = 0;
+
+ n0 = wnumtop[0];
+ n1 = wnumtop[-1];
+ if (n0 == d0)
+ q = BN_MASK2;
+ else { /* n0 < d0 */
+ BN_ULONG n2 = (wnumtop == wnum) ? 0 : wnumtop[-2];
+# ifdef BN_LLONG
+ BN_ULLONG t2;
+
+# if defined(BN_LLONG) && defined(BN_DIV2W) && !defined(bn_div_words)
+ q = (BN_ULONG)(((((BN_ULLONG) n0) << BN_BITS2) | n1) / d0);
+# else
+ q = bn_div_words(n0, n1, d0);
+# endif
+
+# ifndef REMAINDER_IS_ALREADY_CALCULATED
+ /*
+ * rem doesn't have to be BN_ULLONG. The least we
+ * know it's less that d0, isn't it?
+ */
+ rem = (n1 - q * d0) & BN_MASK2;
+# endif
+ t2 = (BN_ULLONG) d1 *q;
+
+ for (;;) {
+ if (t2 <= ((((BN_ULLONG) rem) << BN_BITS2) | n2))
+ break;
+ q--;
+ rem += d0;
+ if (rem < d0)
+ break; /* don't let rem overflow */
+ t2 -= d1;
+ }
+# else /* !BN_LLONG */
+ BN_ULONG t2l, t2h;
+
+ q = bn_div_words(n0, n1, d0);
+# ifndef REMAINDER_IS_ALREADY_CALCULATED
+ rem = (n1 - q * d0) & BN_MASK2;
+# endif
+
+# if defined(BN_UMULT_LOHI)
+ BN_UMULT_LOHI(t2l, t2h, d1, q);
+# elif defined(BN_UMULT_HIGH)
+ t2l = d1 * q;
+ t2h = BN_UMULT_HIGH(d1, q);
+# else
+ {
+ BN_ULONG ql, qh;
+ t2l = LBITS(d1);
+ t2h = HBITS(d1);
+ ql = LBITS(q);
+ qh = HBITS(q);
+ mul64(t2l, t2h, ql, qh); /* t2=(BN_ULLONG)d1*q; */
+ }
+# endif
+
+ for (;;) {
+ if ((t2h < rem) || ((t2h == rem) && (t2l <= n2)))
+ break;
+ q--;
+ rem += d0;
+ if (rem < d0)
+ break; /* don't let rem overflow */
+ if (t2l < d1)
+ t2h--;
+ t2l -= d1;
+ }
+# endif /* !BN_LLONG */
+ }
+# endif /* !BN_DIV3W */
+
+ l0 = bn_mul_words(tmp->d, sdiv->d, div_n, q);
+ tmp->d[div_n] = l0;
+ wnum--;
+ /*
+ * ignore top values of the bignums just sub the two BN_ULONG arrays
+ * with bn_sub_words
+ */
+ l0 = bn_sub_words(wnum, wnum, tmp->d, div_n + 1);
+ q -= l0;
+ /*
+ * Note: As we have considered only the leading two BN_ULONGs in
+ * the calculation of q, sdiv * q might be greater than wnum (but
+ * then (q-1) * sdiv is less or equal than wnum)
+ */
+ for (l0 = 0 - l0, j = 0; j < div_n; j++)
+ tmp->d[j] = sdiv->d[j] & l0;
+ l0 = bn_add_words(wnum, wnum, tmp->d, div_n);
+ (*wnumtop) += l0;
+ assert((*wnumtop) == 0);
+
+ /* store part of the result */
+ *--resp = q;
+ }
+ /* snum holds remainder, it's as wide as divisor */
+ snum->neg = num_neg;
+ snum->top = div_n;
+ snum->flags |= BN_FLG_FIXED_TOP;
+
+ if (rm != NULL && bn_rshift_fixed_top(rm, snum, norm_shift) == 0)
+ goto err;
+
+ BN_CTX_end(ctx);
+ return 1;
+ err:
+ bn_check_top(rm);
+ BN_CTX_end(ctx);
+ return 0;
+}
+#endif
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_err.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_err.c
new file mode 100644
index 0000000..dd87c15
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_err.c
@@ -0,0 +1,118 @@
+/*
+ * Generated by util/mkerr.pl DO NOT EDIT
+ * Copyright 1995-2018 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include <openssl/err.h>
+#include <openssl/bnerr.h>
+
+#ifndef OPENSSL_NO_ERR
+
+static const ERR_STRING_DATA BN_str_functs[] = {
+ {ERR_PACK(ERR_LIB_BN, BN_F_BNRAND, 0), "bnrand"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BNRAND_RANGE, 0), "bnrand_range"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_BLINDING_CONVERT_EX, 0),
+ "BN_BLINDING_convert_ex"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_BLINDING_CREATE_PARAM, 0),
+ "BN_BLINDING_create_param"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_BLINDING_INVERT_EX, 0),
+ "BN_BLINDING_invert_ex"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_BLINDING_NEW, 0), "BN_BLINDING_new"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_BLINDING_UPDATE, 0), "BN_BLINDING_update"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_BN2DEC, 0), "BN_bn2dec"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_BN2HEX, 0), "BN_bn2hex"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_COMPUTE_WNAF, 0), "bn_compute_wNAF"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_CTX_GET, 0), "BN_CTX_get"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_CTX_NEW, 0), "BN_CTX_new"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_CTX_START, 0), "BN_CTX_start"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_DIV, 0), "BN_div"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_DIV_RECP, 0), "BN_div_recp"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_EXP, 0), "BN_exp"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_EXPAND_INTERNAL, 0), "bn_expand_internal"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_GENCB_NEW, 0), "BN_GENCB_new"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_GENERATE_DSA_NONCE, 0),
+ "BN_generate_dsa_nonce"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_GENERATE_PRIME_EX, 0),
+ "BN_generate_prime_ex"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_GF2M_MOD, 0), "BN_GF2m_mod"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_GF2M_MOD_EXP, 0), "BN_GF2m_mod_exp"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_GF2M_MOD_MUL, 0), "BN_GF2m_mod_mul"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_GF2M_MOD_SOLVE_QUAD, 0),
+ "BN_GF2m_mod_solve_quad"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_GF2M_MOD_SOLVE_QUAD_ARR, 0),
+ "BN_GF2m_mod_solve_quad_arr"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_GF2M_MOD_SQR, 0), "BN_GF2m_mod_sqr"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_GF2M_MOD_SQRT, 0), "BN_GF2m_mod_sqrt"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_LSHIFT, 0), "BN_lshift"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_MOD_EXP2_MONT, 0), "BN_mod_exp2_mont"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_MOD_EXP_MONT, 0), "BN_mod_exp_mont"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_MOD_EXP_MONT_CONSTTIME, 0),
+ "BN_mod_exp_mont_consttime"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_MOD_EXP_MONT_WORD, 0),
+ "BN_mod_exp_mont_word"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_MOD_EXP_RECP, 0), "BN_mod_exp_recp"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_MOD_EXP_SIMPLE, 0), "BN_mod_exp_simple"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_MOD_INVERSE, 0), "BN_mod_inverse"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_MOD_INVERSE_NO_BRANCH, 0),
+ "BN_mod_inverse_no_branch"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_MOD_LSHIFT_QUICK, 0), "BN_mod_lshift_quick"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_MOD_SQRT, 0), "BN_mod_sqrt"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_MONT_CTX_NEW, 0), "BN_MONT_CTX_new"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_MPI2BN, 0), "BN_mpi2bn"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_NEW, 0), "BN_new"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_POOL_GET, 0), "BN_POOL_get"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_RAND, 0), "BN_rand"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_RAND_RANGE, 0), "BN_rand_range"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_RECP_CTX_NEW, 0), "BN_RECP_CTX_new"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_RSHIFT, 0), "BN_rshift"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_SET_WORDS, 0), "bn_set_words"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_STACK_PUSH, 0), "BN_STACK_push"},
+ {ERR_PACK(ERR_LIB_BN, BN_F_BN_USUB, 0), "BN_usub"},
+ {0, NULL}
+};
+
+static const ERR_STRING_DATA BN_str_reasons[] = {
+ {ERR_PACK(ERR_LIB_BN, 0, BN_R_ARG2_LT_ARG3), "arg2 lt arg3"},
+ {ERR_PACK(ERR_LIB_BN, 0, BN_R_BAD_RECIPROCAL), "bad reciprocal"},
+ {ERR_PACK(ERR_LIB_BN, 0, BN_R_BIGNUM_TOO_LONG), "bignum too long"},
+ {ERR_PACK(ERR_LIB_BN, 0, BN_R_BITS_TOO_SMALL), "bits too small"},
+ {ERR_PACK(ERR_LIB_BN, 0, BN_R_CALLED_WITH_EVEN_MODULUS),
+ "called with even modulus"},
+ {ERR_PACK(ERR_LIB_BN, 0, BN_R_DIV_BY_ZERO), "div by zero"},
+ {ERR_PACK(ERR_LIB_BN, 0, BN_R_ENCODING_ERROR), "encoding error"},
+ {ERR_PACK(ERR_LIB_BN, 0, BN_R_EXPAND_ON_STATIC_BIGNUM_DATA),
+ "expand on static bignum data"},
+ {ERR_PACK(ERR_LIB_BN, 0, BN_R_INPUT_NOT_REDUCED), "input not reduced"},
+ {ERR_PACK(ERR_LIB_BN, 0, BN_R_INVALID_LENGTH), "invalid length"},
+ {ERR_PACK(ERR_LIB_BN, 0, BN_R_INVALID_RANGE), "invalid range"},
+ {ERR_PACK(ERR_LIB_BN, 0, BN_R_INVALID_SHIFT), "invalid shift"},
+ {ERR_PACK(ERR_LIB_BN, 0, BN_R_NOT_A_SQUARE), "not a square"},
+ {ERR_PACK(ERR_LIB_BN, 0, BN_R_NOT_INITIALIZED), "not initialized"},
+ {ERR_PACK(ERR_LIB_BN, 0, BN_R_NO_INVERSE), "no inverse"},
+ {ERR_PACK(ERR_LIB_BN, 0, BN_R_NO_SOLUTION), "no solution"},
+ {ERR_PACK(ERR_LIB_BN, 0, BN_R_PRIVATE_KEY_TOO_LARGE),
+ "private key too large"},
+ {ERR_PACK(ERR_LIB_BN, 0, BN_R_P_IS_NOT_PRIME), "p is not prime"},
+ {ERR_PACK(ERR_LIB_BN, 0, BN_R_TOO_MANY_ITERATIONS), "too many iterations"},
+ {ERR_PACK(ERR_LIB_BN, 0, BN_R_TOO_MANY_TEMPORARY_VARIABLES),
+ "too many temporary variables"},
+ {0, NULL}
+};
+
+#endif
+
+int ERR_load_BN_strings(void)
+{
+#ifndef OPENSSL_NO_ERR
+ if (ERR_func_error_string(BN_str_functs[0].error) == NULL) {
+ ERR_load_strings_const(BN_str_functs);
+ ERR_load_strings_const(BN_str_reasons);
+ }
+#endif
+ return 1;
+}
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_exp.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_exp.c
new file mode 100644
index 0000000..e21dcff
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_exp.c
@@ -0,0 +1,1404 @@
+/*
+ * Copyright 1995-2022 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include "internal/cryptlib.h"
+#include "internal/constant_time.h"
+#include "bn_local.h"
+
+#include <stdlib.h>
+#ifdef _WIN32
+# include <malloc.h>
+# ifndef alloca
+# define alloca _alloca
+# endif
+#elif defined(__GNUC__)
+# ifndef alloca
+# define alloca(s) __builtin_alloca((s))
+# endif
+#elif defined(__sun)
+# include <alloca.h>
+#endif
+
+#include "rsaz_exp.h"
+
+#undef SPARC_T4_MONT
+#if defined(OPENSSL_BN_ASM_MONT) && (defined(__sparc__) || defined(__sparc))
+# include "sparc_arch.h"
+extern unsigned int OPENSSL_sparcv9cap_P[];
+# define SPARC_T4_MONT
+#endif
+
+/* maximum precomputation table size for *variable* sliding windows */
+#define TABLE_SIZE 32
+
+/* this one works - simple but works */
+int BN_exp(BIGNUM *r, const BIGNUM *a, const BIGNUM *p, BN_CTX *ctx)
+{
+ int i, bits, ret = 0;
+ BIGNUM *v, *rr;
+
+ if (BN_get_flags(p, BN_FLG_CONSTTIME) != 0
+ || BN_get_flags(a, BN_FLG_CONSTTIME) != 0) {
+ /* BN_FLG_CONSTTIME only supported by BN_mod_exp_mont() */
+ BNerr(BN_F_BN_EXP, ERR_R_SHOULD_NOT_HAVE_BEEN_CALLED);
+ return 0;
+ }
+
+ BN_CTX_start(ctx);
+ rr = ((r == a) || (r == p)) ? BN_CTX_get(ctx) : r;
+ v = BN_CTX_get(ctx);
+ if (rr == NULL || v == NULL)
+ goto err;
+
+ if (BN_copy(v, a) == NULL)
+ goto err;
+ bits = BN_num_bits(p);
+
+ if (BN_is_odd(p)) {
+ if (BN_copy(rr, a) == NULL)
+ goto err;
+ } else {
+ if (!BN_one(rr))
+ goto err;
+ }
+
+ for (i = 1; i < bits; i++) {
+ if (!BN_sqr(v, v, ctx))
+ goto err;
+ if (BN_is_bit_set(p, i)) {
+ if (!BN_mul(rr, rr, v, ctx))
+ goto err;
+ }
+ }
+ if (r != rr && BN_copy(r, rr) == NULL)
+ goto err;
+
+ ret = 1;
+ err:
+ BN_CTX_end(ctx);
+ bn_check_top(r);
+ return ret;
+}
+
+int BN_mod_exp(BIGNUM *r, const BIGNUM *a, const BIGNUM *p, const BIGNUM *m,
+ BN_CTX *ctx)
+{
+ int ret;
+
+ bn_check_top(a);
+ bn_check_top(p);
+ bn_check_top(m);
+
+ /*-
+ * For even modulus m = 2^k*m_odd, it might make sense to compute
+ * a^p mod m_odd and a^p mod 2^k separately (with Montgomery
+ * exponentiation for the odd part), using appropriate exponent
+ * reductions, and combine the results using the CRT.
+ *
+ * For now, we use Montgomery only if the modulus is odd; otherwise,
+ * exponentiation using the reciprocal-based quick remaindering
+ * algorithm is used.
+ *
+ * (Timing obtained with expspeed.c [computations a^p mod m
+ * where a, p, m are of the same length: 256, 512, 1024, 2048,
+ * 4096, 8192 bits], compared to the running time of the
+ * standard algorithm:
+ *
+ * BN_mod_exp_mont 33 .. 40 % [AMD K6-2, Linux, debug configuration]
+ * 55 .. 77 % [UltraSparc processor, but
+ * debug-solaris-sparcv8-gcc conf.]
+ *
+ * BN_mod_exp_recp 50 .. 70 % [AMD K6-2, Linux, debug configuration]
+ * 62 .. 118 % [UltraSparc, debug-solaris-sparcv8-gcc]
+ *
+ * On the Sparc, BN_mod_exp_recp was faster than BN_mod_exp_mont
+ * at 2048 and more bits, but at 512 and 1024 bits, it was
+ * slower even than the standard algorithm!
+ *
+ * "Real" timings [linux-elf, solaris-sparcv9-gcc configurations]
+ * should be obtained when the new Montgomery reduction code
+ * has been integrated into OpenSSL.)
+ */
+
+#define MONT_MUL_MOD
+#define MONT_EXP_WORD
+#define RECP_MUL_MOD
+
+#ifdef MONT_MUL_MOD
+ if (BN_is_odd(m)) {
+# ifdef MONT_EXP_WORD
+ if (a->top == 1 && !a->neg
+ && (BN_get_flags(p, BN_FLG_CONSTTIME) == 0)
+ && (BN_get_flags(a, BN_FLG_CONSTTIME) == 0)
+ && (BN_get_flags(m, BN_FLG_CONSTTIME) == 0)) {
+ BN_ULONG A = a->d[0];
+ ret = BN_mod_exp_mont_word(r, A, p, m, ctx, NULL);
+ } else
+# endif
+ ret = BN_mod_exp_mont(r, a, p, m, ctx, NULL);
+ } else
+#endif
+#ifdef RECP_MUL_MOD
+ {
+ ret = BN_mod_exp_recp(r, a, p, m, ctx);
+ }
+#else
+ {
+ ret = BN_mod_exp_simple(r, a, p, m, ctx);
+ }
+#endif
+
+ bn_check_top(r);
+ return ret;
+}
+
+int BN_mod_exp_recp(BIGNUM *r, const BIGNUM *a, const BIGNUM *p,
+ const BIGNUM *m, BN_CTX *ctx)
+{
+ int i, j, bits, ret = 0, wstart, wend, window, wvalue;
+ int start = 1;
+ BIGNUM *aa;
+ /* Table of variables obtained from 'ctx' */
+ BIGNUM *val[TABLE_SIZE];
+ BN_RECP_CTX recp;
+
+ if (BN_get_flags(p, BN_FLG_CONSTTIME) != 0
+ || BN_get_flags(a, BN_FLG_CONSTTIME) != 0
+ || BN_get_flags(m, BN_FLG_CONSTTIME) != 0) {
+ /* BN_FLG_CONSTTIME only supported by BN_mod_exp_mont() */
+ BNerr(BN_F_BN_MOD_EXP_RECP, ERR_R_SHOULD_NOT_HAVE_BEEN_CALLED);
+ return 0;
+ }
+
+ bits = BN_num_bits(p);
+ if (bits == 0) {
+ /* x**0 mod 1, or x**0 mod -1 is still zero. */
+ if (BN_abs_is_word(m, 1)) {
+ ret = 1;
+ BN_zero(r);
+ } else {
+ ret = BN_one(r);
+ }
+ return ret;
+ }
+
+ BN_RECP_CTX_init(&recp);
+
+ BN_CTX_start(ctx);
+ aa = BN_CTX_get(ctx);
+ val[0] = BN_CTX_get(ctx);
+ if (val[0] == NULL)
+ goto err;
+
+ if (m->neg) {
+ /* ignore sign of 'm' */
+ if (!BN_copy(aa, m))
+ goto err;
+ aa->neg = 0;
+ if (BN_RECP_CTX_set(&recp, aa, ctx) <= 0)
+ goto err;
+ } else {
+ if (BN_RECP_CTX_set(&recp, m, ctx) <= 0)
+ goto err;
+ }
+
+ if (!BN_nnmod(val[0], a, m, ctx))
+ goto err; /* 1 */
+ if (BN_is_zero(val[0])) {
+ BN_zero(r);
+ ret = 1;
+ goto err;
+ }
+
+ window = BN_window_bits_for_exponent_size(bits);
+ if (window > 1) {
+ if (!BN_mod_mul_reciprocal(aa, val[0], val[0], &recp, ctx))
+ goto err; /* 2 */
+ j = 1 << (window - 1);
+ for (i = 1; i < j; i++) {
+ if (((val[i] = BN_CTX_get(ctx)) == NULL) ||
+ !BN_mod_mul_reciprocal(val[i], val[i - 1], aa, &recp, ctx))
+ goto err;
+ }
+ }
+
+ start = 1; /* This is used to avoid multiplication etc
+ * when there is only the value '1' in the
+ * buffer. */
+ wvalue = 0; /* The 'value' of the window */
+ wstart = bits - 1; /* The top bit of the window */
+ wend = 0; /* The bottom bit of the window */
+
+ if (!BN_one(r))
+ goto err;
+
+ for (;;) {
+ if (BN_is_bit_set(p, wstart) == 0) {
+ if (!start)
+ if (!BN_mod_mul_reciprocal(r, r, r, &recp, ctx))
+ goto err;
+ if (wstart == 0)
+ break;
+ wstart--;
+ continue;
+ }
+ /*
+ * We now have wstart on a 'set' bit, we now need to work out how bit
+ * a window to do. To do this we need to scan forward until the last
+ * set bit before the end of the window
+ */
+ j = wstart;
+ wvalue = 1;
+ wend = 0;
+ for (i = 1; i < window; i++) {
+ if (wstart - i < 0)
+ break;
+ if (BN_is_bit_set(p, wstart - i)) {
+ wvalue <<= (i - wend);
+ wvalue |= 1;
+ wend = i;
+ }
+ }
+
+ /* wend is the size of the current window */
+ j = wend + 1;
+ /* add the 'bytes above' */
+ if (!start)
+ for (i = 0; i < j; i++) {
+ if (!BN_mod_mul_reciprocal(r, r, r, &recp, ctx))
+ goto err;
+ }
+
+ /* wvalue will be an odd number < 2^window */
+ if (!BN_mod_mul_reciprocal(r, r, val[wvalue >> 1], &recp, ctx))
+ goto err;
+
+ /* move the 'window' down further */
+ wstart -= wend + 1;
+ wvalue = 0;
+ start = 0;
+ if (wstart < 0)
+ break;
+ }
+ ret = 1;
+ err:
+ BN_CTX_end(ctx);
+ BN_RECP_CTX_free(&recp);
+ bn_check_top(r);
+ return ret;
+}
+
+int BN_mod_exp_mont(BIGNUM *rr, const BIGNUM *a, const BIGNUM *p,
+ const BIGNUM *m, BN_CTX *ctx, BN_MONT_CTX *in_mont)
+{
+ int i, j, bits, ret = 0, wstart, wend, window, wvalue;
+ int start = 1;
+ BIGNUM *d, *r;
+ const BIGNUM *aa;
+ /* Table of variables obtained from 'ctx' */
+ BIGNUM *val[TABLE_SIZE];
+ BN_MONT_CTX *mont = NULL;
+
+ if (BN_get_flags(p, BN_FLG_CONSTTIME) != 0
+ || BN_get_flags(a, BN_FLG_CONSTTIME) != 0
+ || BN_get_flags(m, BN_FLG_CONSTTIME) != 0) {
+ return BN_mod_exp_mont_consttime(rr, a, p, m, ctx, in_mont);
+ }
+
+ bn_check_top(a);
+ bn_check_top(p);
+ bn_check_top(m);
+
+ if (!BN_is_odd(m)) {
+ BNerr(BN_F_BN_MOD_EXP_MONT, BN_R_CALLED_WITH_EVEN_MODULUS);
+ return 0;
+ }
+ bits = BN_num_bits(p);
+ if (bits == 0) {
+ /* x**0 mod 1, or x**0 mod -1 is still zero. */
+ if (BN_abs_is_word(m, 1)) {
+ ret = 1;
+ BN_zero(rr);
+ } else {
+ ret = BN_one(rr);
+ }
+ return ret;
+ }
+
+ BN_CTX_start(ctx);
+ d = BN_CTX_get(ctx);
+ r = BN_CTX_get(ctx);
+ val[0] = BN_CTX_get(ctx);
+ if (val[0] == NULL)
+ goto err;
+
+ /*
+ * If this is not done, things will break in the montgomery part
+ */
+
+ if (in_mont != NULL)
+ mont = in_mont;
+ else {
+ if ((mont = BN_MONT_CTX_new()) == NULL)
+ goto err;
+ if (!BN_MONT_CTX_set(mont, m, ctx))
+ goto err;
+ }
+
+ if (a->neg || BN_ucmp(a, m) >= 0) {
+ if (!BN_nnmod(val[0], a, m, ctx))
+ goto err;
+ aa = val[0];
+ } else
+ aa = a;
+ if (!bn_to_mont_fixed_top(val[0], aa, mont, ctx))
+ goto err; /* 1 */
+
+ window = BN_window_bits_for_exponent_size(bits);
+ if (window > 1) {
+ if (!bn_mul_mont_fixed_top(d, val[0], val[0], mont, ctx))
+ goto err; /* 2 */
+ j = 1 << (window - 1);
+ for (i = 1; i < j; i++) {
+ if (((val[i] = BN_CTX_get(ctx)) == NULL) ||
+ !bn_mul_mont_fixed_top(val[i], val[i - 1], d, mont, ctx))
+ goto err;
+ }
+ }
+
+ start = 1; /* This is used to avoid multiplication etc
+ * when there is only the value '1' in the
+ * buffer. */
+ wvalue = 0; /* The 'value' of the window */
+ wstart = bits - 1; /* The top bit of the window */
+ wend = 0; /* The bottom bit of the window */
+
+#if 1 /* by Shay Gueron's suggestion */
+ j = m->top; /* borrow j */
+ if (m->d[j - 1] & (((BN_ULONG)1) << (BN_BITS2 - 1))) {
+ if (bn_wexpand(r, j) == NULL)
+ goto err;
+ /* 2^(top*BN_BITS2) - m */
+ r->d[0] = (0 - m->d[0]) & BN_MASK2;
+ for (i = 1; i < j; i++)
+ r->d[i] = (~m->d[i]) & BN_MASK2;
+ r->top = j;
+ r->flags |= BN_FLG_FIXED_TOP;
+ } else
+#endif
+ if (!bn_to_mont_fixed_top(r, BN_value_one(), mont, ctx))
+ goto err;
+ for (;;) {
+ if (BN_is_bit_set(p, wstart) == 0) {
+ if (!start) {
+ if (!bn_mul_mont_fixed_top(r, r, r, mont, ctx))
+ goto err;
+ }
+ if (wstart == 0)
+ break;
+ wstart--;
+ continue;
+ }
+ /*
+ * We now have wstart on a 'set' bit, we now need to work out how bit
+ * a window to do. To do this we need to scan forward until the last
+ * set bit before the end of the window
+ */
+ j = wstart;
+ wvalue = 1;
+ wend = 0;
+ for (i = 1; i < window; i++) {
+ if (wstart - i < 0)
+ break;
+ if (BN_is_bit_set(p, wstart - i)) {
+ wvalue <<= (i - wend);
+ wvalue |= 1;
+ wend = i;
+ }
+ }
+
+ /* wend is the size of the current window */
+ j = wend + 1;
+ /* add the 'bytes above' */
+ if (!start)
+ for (i = 0; i < j; i++) {
+ if (!bn_mul_mont_fixed_top(r, r, r, mont, ctx))
+ goto err;
+ }
+
+ /* wvalue will be an odd number < 2^window */
+ if (!bn_mul_mont_fixed_top(r, r, val[wvalue >> 1], mont, ctx))
+ goto err;
+
+ /* move the 'window' down further */
+ wstart -= wend + 1;
+ wvalue = 0;
+ start = 0;
+ if (wstart < 0)
+ break;
+ }
+ /*
+ * Done with zero-padded intermediate BIGNUMs. Final BN_from_montgomery
+ * removes padding [if any] and makes return value suitable for public
+ * API consumer.
+ */
+#if defined(SPARC_T4_MONT)
+ if (OPENSSL_sparcv9cap_P[0] & (SPARCV9_VIS3 | SPARCV9_PREFER_FPU)) {
+ j = mont->N.top; /* borrow j */
+ val[0]->d[0] = 1; /* borrow val[0] */
+ for (i = 1; i < j; i++)
+ val[0]->d[i] = 0;
+ val[0]->top = j;
+ if (!BN_mod_mul_montgomery(rr, r, val[0], mont, ctx))
+ goto err;
+ } else
+#endif
+ if (!BN_from_montgomery(rr, r, mont, ctx))
+ goto err;
+ ret = 1;
+ err:
+ if (in_mont == NULL)
+ BN_MONT_CTX_free(mont);
+ BN_CTX_end(ctx);
+ bn_check_top(rr);
+ return ret;
+}
+
+static BN_ULONG bn_get_bits(const BIGNUM *a, int bitpos)
+{
+ BN_ULONG ret = 0;
+ int wordpos;
+
+ wordpos = bitpos / BN_BITS2;
+ bitpos %= BN_BITS2;
+ if (wordpos >= 0 && wordpos < a->top) {
+ ret = a->d[wordpos] & BN_MASK2;
+ if (bitpos) {
+ ret >>= bitpos;
+ if (++wordpos < a->top)
+ ret |= a->d[wordpos] << (BN_BITS2 - bitpos);
+ }
+ }
+
+ return ret & BN_MASK2;
+}
+
+/*
+ * BN_mod_exp_mont_consttime() stores the precomputed powers in a specific
+ * layout so that accessing any of these table values shows the same access
+ * pattern as far as cache lines are concerned. The following functions are
+ * used to transfer a BIGNUM from/to that table.
+ */
+
+static int MOD_EXP_CTIME_COPY_TO_PREBUF(const BIGNUM *b, int top,
+ unsigned char *buf, int idx,
+ int window)
+{
+ int i, j;
+ int width = 1 << window;
+ BN_ULONG *table = (BN_ULONG *)buf;
+
+ if (top > b->top)
+ top = b->top; /* this works because 'buf' is explicitly
+ * zeroed */
+ for (i = 0, j = idx; i < top; i++, j += width) {
+ table[j] = b->d[i];
+ }
+
+ return 1;
+}
+
+static int MOD_EXP_CTIME_COPY_FROM_PREBUF(BIGNUM *b, int top,
+ unsigned char *buf, int idx,
+ int window)
+{
+ int i, j;
+ int width = 1 << window;
+ /*
+ * We declare table 'volatile' in order to discourage compiler
+ * from reordering loads from the table. Concern is that if
+ * reordered in specific manner loads might give away the
+ * information we are trying to conceal. Some would argue that
+ * compiler can reorder them anyway, but it can as well be
+ * argued that doing so would be violation of standard...
+ */
+ volatile BN_ULONG *table = (volatile BN_ULONG *)buf;
+
+ if (bn_wexpand(b, top) == NULL)
+ return 0;
+
+ if (window <= 3) {
+ for (i = 0; i < top; i++, table += width) {
+ BN_ULONG acc = 0;
+
+ for (j = 0; j < width; j++) {
+ acc |= table[j] &
+ ((BN_ULONG)0 - (constant_time_eq_int(j,idx)&1));
+ }
+
+ b->d[i] = acc;
+ }
+ } else {
+ int xstride = 1 << (window - 2);
+ BN_ULONG y0, y1, y2, y3;
+
+ i = idx >> (window - 2); /* equivalent of idx / xstride */
+ idx &= xstride - 1; /* equivalent of idx % xstride */
+
+ y0 = (BN_ULONG)0 - (constant_time_eq_int(i,0)&1);
+ y1 = (BN_ULONG)0 - (constant_time_eq_int(i,1)&1);
+ y2 = (BN_ULONG)0 - (constant_time_eq_int(i,2)&1);
+ y3 = (BN_ULONG)0 - (constant_time_eq_int(i,3)&1);
+
+ for (i = 0; i < top; i++, table += width) {
+ BN_ULONG acc = 0;
+
+ for (j = 0; j < xstride; j++) {
+ acc |= ( (table[j + 0 * xstride] & y0) |
+ (table[j + 1 * xstride] & y1) |
+ (table[j + 2 * xstride] & y2) |
+ (table[j + 3 * xstride] & y3) )
+ & ((BN_ULONG)0 - (constant_time_eq_int(j,idx)&1));
+ }
+
+ b->d[i] = acc;
+ }
+ }
+
+ b->top = top;
+ b->flags |= BN_FLG_FIXED_TOP;
+ return 1;
+}
+
+/*
+ * Given a pointer value, compute the next address that is a cache line
+ * multiple.
+ */
+#define MOD_EXP_CTIME_ALIGN(x_) \
+ ((unsigned char*)(x_) + (MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH - (((size_t)(x_)) & (MOD_EXP_CTIME_MIN_CACHE_LINE_MASK))))
+
+/*
+ * This variant of BN_mod_exp_mont() uses fixed windows and the special
+ * precomputation memory layout to limit data-dependency to a minimum to
+ * protect secret exponents (cf. the hyper-threading timing attacks pointed
+ * out by Colin Percival,
+ * http://www.daemonology.net/hyperthreading-considered-harmful/)
+ */
+int BN_mod_exp_mont_consttime(BIGNUM *rr, const BIGNUM *a, const BIGNUM *p,
+ const BIGNUM *m, BN_CTX *ctx,
+ BN_MONT_CTX *in_mont)
+{
+ int i, bits, ret = 0, window, wvalue, wmask, window0;
+ int top;
+ BN_MONT_CTX *mont = NULL;
+
+ int numPowers;
+ unsigned char *powerbufFree = NULL;
+ int powerbufLen = 0;
+ unsigned char *powerbuf = NULL;
+ BIGNUM tmp, am;
+#if defined(SPARC_T4_MONT)
+ unsigned int t4 = 0;
+#endif
+
+ bn_check_top(a);
+ bn_check_top(p);
+ bn_check_top(m);
+
+ if (!BN_is_odd(m)) {
+ BNerr(BN_F_BN_MOD_EXP_MONT_CONSTTIME, BN_R_CALLED_WITH_EVEN_MODULUS);
+ return 0;
+ }
+
+ top = m->top;
+
+ /*
+ * Use all bits stored in |p|, rather than |BN_num_bits|, so we do not leak
+ * whether the top bits are zero.
+ */
+ bits = p->top * BN_BITS2;
+ if (bits == 0) {
+ /* x**0 mod 1, or x**0 mod -1 is still zero. */
+ if (BN_abs_is_word(m, 1)) {
+ ret = 1;
+ BN_zero(rr);
+ } else {
+ ret = BN_one(rr);
+ }
+ return ret;
+ }
+
+ BN_CTX_start(ctx);
+
+ /*
+ * Allocate a montgomery context if it was not supplied by the caller. If
+ * this is not done, things will break in the montgomery part.
+ */
+ if (in_mont != NULL)
+ mont = in_mont;
+ else {
+ if ((mont = BN_MONT_CTX_new()) == NULL)
+ goto err;
+ if (!BN_MONT_CTX_set(mont, m, ctx))
+ goto err;
+ }
+
+ if (a->neg || BN_ucmp(a, m) >= 0) {
+ BIGNUM *reduced = BN_CTX_get(ctx);
+ if (reduced == NULL
+ || !BN_nnmod(reduced, a, m, ctx)) {
+ goto err;
+ }
+ a = reduced;
+ }
+
+#ifdef RSAZ_ENABLED
+ /*
+ * If the size of the operands allow it, perform the optimized
+ * RSAZ exponentiation. For further information see
+ * crypto/bn/rsaz_exp.c and accompanying assembly modules.
+ */
+ if ((16 == a->top) && (16 == p->top) && (BN_num_bits(m) == 1024)
+ && rsaz_avx2_eligible()) {
+ if (NULL == bn_wexpand(rr, 16))
+ goto err;
+ RSAZ_1024_mod_exp_avx2(rr->d, a->d, p->d, m->d, mont->RR.d,
+ mont->n0[0]);
+ rr->top = 16;
+ rr->neg = 0;
+ bn_correct_top(rr);
+ ret = 1;
+ goto err;
+ } else if ((8 == a->top) && (8 == p->top) && (BN_num_bits(m) == 512)) {
+ if (NULL == bn_wexpand(rr, 8))
+ goto err;
+ RSAZ_512_mod_exp(rr->d, a->d, p->d, m->d, mont->n0[0], mont->RR.d);
+ rr->top = 8;
+ rr->neg = 0;
+ bn_correct_top(rr);
+ ret = 1;
+ goto err;
+ }
+#endif
+
+ /* Get the window size to use with size of p. */
+ window = BN_window_bits_for_ctime_exponent_size(bits);
+#if defined(SPARC_T4_MONT)
+ if (window >= 5 && (top & 15) == 0 && top <= 64 &&
+ (OPENSSL_sparcv9cap_P[1] & (CFR_MONTMUL | CFR_MONTSQR)) ==
+ (CFR_MONTMUL | CFR_MONTSQR) && (t4 = OPENSSL_sparcv9cap_P[0]))
+ window = 5;
+ else
+#endif
+#if defined(OPENSSL_BN_ASM_MONT5)
+ if (window >= 5) {
+ window = 5; /* ~5% improvement for RSA2048 sign, and even
+ * for RSA4096 */
+ /* reserve space for mont->N.d[] copy */
+ powerbufLen += top * sizeof(mont->N.d[0]);
+ }
+#endif
+ (void)0;
+
+ /*
+ * Allocate a buffer large enough to hold all of the pre-computed powers
+ * of am, am itself and tmp.
+ */
+ numPowers = 1 << window;
+ powerbufLen += sizeof(m->d[0]) * (top * numPowers +
+ ((2 * top) >
+ numPowers ? (2 * top) : numPowers));
+#ifdef alloca
+ if (powerbufLen < 3072)
+ powerbufFree =
+ alloca(powerbufLen + MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH);
+ else
+#endif
+ if ((powerbufFree =
+ OPENSSL_malloc(powerbufLen + MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH))
+ == NULL)
+ goto err;
+
+ powerbuf = MOD_EXP_CTIME_ALIGN(powerbufFree);
+ memset(powerbuf, 0, powerbufLen);
+
+#ifdef alloca
+ if (powerbufLen < 3072)
+ powerbufFree = NULL;
+#endif
+
+ /* lay down tmp and am right after powers table */
+ tmp.d = (BN_ULONG *)(powerbuf + sizeof(m->d[0]) * top * numPowers);
+ am.d = tmp.d + top;
+ tmp.top = am.top = 0;
+ tmp.dmax = am.dmax = top;
+ tmp.neg = am.neg = 0;
+ tmp.flags = am.flags = BN_FLG_STATIC_DATA;
+
+ /* prepare a^0 in Montgomery domain */
+#if 1 /* by Shay Gueron's suggestion */
+ if (m->d[top - 1] & (((BN_ULONG)1) << (BN_BITS2 - 1))) {
+ /* 2^(top*BN_BITS2) - m */
+ tmp.d[0] = (0 - m->d[0]) & BN_MASK2;
+ for (i = 1; i < top; i++)
+ tmp.d[i] = (~m->d[i]) & BN_MASK2;
+ tmp.top = top;
+ } else
+#endif
+ if (!bn_to_mont_fixed_top(&tmp, BN_value_one(), mont, ctx))
+ goto err;
+
+ /* prepare a^1 in Montgomery domain */
+ if (!bn_to_mont_fixed_top(&am, a, mont, ctx))
+ goto err;
+
+#if defined(SPARC_T4_MONT)
+ if (t4) {
+ typedef int (*bn_pwr5_mont_f) (BN_ULONG *tp, const BN_ULONG *np,
+ const BN_ULONG *n0, const void *table,
+ int power, int bits);
+ int bn_pwr5_mont_t4_8(BN_ULONG *tp, const BN_ULONG *np,
+ const BN_ULONG *n0, const void *table,
+ int power, int bits);
+ int bn_pwr5_mont_t4_16(BN_ULONG *tp, const BN_ULONG *np,
+ const BN_ULONG *n0, const void *table,
+ int power, int bits);
+ int bn_pwr5_mont_t4_24(BN_ULONG *tp, const BN_ULONG *np,
+ const BN_ULONG *n0, const void *table,
+ int power, int bits);
+ int bn_pwr5_mont_t4_32(BN_ULONG *tp, const BN_ULONG *np,
+ const BN_ULONG *n0, const void *table,
+ int power, int bits);
+ static const bn_pwr5_mont_f pwr5_funcs[4] = {
+ bn_pwr5_mont_t4_8, bn_pwr5_mont_t4_16,
+ bn_pwr5_mont_t4_24, bn_pwr5_mont_t4_32
+ };
+ bn_pwr5_mont_f pwr5_worker = pwr5_funcs[top / 16 - 1];
+
+ typedef int (*bn_mul_mont_f) (BN_ULONG *rp, const BN_ULONG *ap,
+ const void *bp, const BN_ULONG *np,
+ const BN_ULONG *n0);
+ int bn_mul_mont_t4_8(BN_ULONG *rp, const BN_ULONG *ap, const void *bp,
+ const BN_ULONG *np, const BN_ULONG *n0);
+ int bn_mul_mont_t4_16(BN_ULONG *rp, const BN_ULONG *ap,
+ const void *bp, const BN_ULONG *np,
+ const BN_ULONG *n0);
+ int bn_mul_mont_t4_24(BN_ULONG *rp, const BN_ULONG *ap,
+ const void *bp, const BN_ULONG *np,
+ const BN_ULONG *n0);
+ int bn_mul_mont_t4_32(BN_ULONG *rp, const BN_ULONG *ap,
+ const void *bp, const BN_ULONG *np,
+ const BN_ULONG *n0);
+ static const bn_mul_mont_f mul_funcs[4] = {
+ bn_mul_mont_t4_8, bn_mul_mont_t4_16,
+ bn_mul_mont_t4_24, bn_mul_mont_t4_32
+ };
+ bn_mul_mont_f mul_worker = mul_funcs[top / 16 - 1];
+
+ void bn_mul_mont_vis3(BN_ULONG *rp, const BN_ULONG *ap,
+ const void *bp, const BN_ULONG *np,
+ const BN_ULONG *n0, int num);
+ void bn_mul_mont_t4(BN_ULONG *rp, const BN_ULONG *ap,
+ const void *bp, const BN_ULONG *np,
+ const BN_ULONG *n0, int num);
+ void bn_mul_mont_gather5_t4(BN_ULONG *rp, const BN_ULONG *ap,
+ const void *table, const BN_ULONG *np,
+ const BN_ULONG *n0, int num, int power);
+ void bn_flip_n_scatter5_t4(const BN_ULONG *inp, size_t num,
+ void *table, size_t power);
+ void bn_gather5_t4(BN_ULONG *out, size_t num,
+ void *table, size_t power);
+ void bn_flip_t4(BN_ULONG *dst, BN_ULONG *src, size_t num);
+
+ BN_ULONG *np = mont->N.d, *n0 = mont->n0;
+ int stride = 5 * (6 - (top / 16 - 1)); /* multiple of 5, but less
+ * than 32 */
+
+ /*
+ * BN_to_montgomery can contaminate words above .top [in
+ * BN_DEBUG[_DEBUG] build]...
+ */
+ for (i = am.top; i < top; i++)
+ am.d[i] = 0;
+ for (i = tmp.top; i < top; i++)
+ tmp.d[i] = 0;
+
+ bn_flip_n_scatter5_t4(tmp.d, top, powerbuf, 0);
+ bn_flip_n_scatter5_t4(am.d, top, powerbuf, 1);
+ if (!(*mul_worker) (tmp.d, am.d, am.d, np, n0) &&
+ !(*mul_worker) (tmp.d, am.d, am.d, np, n0))
+ bn_mul_mont_vis3(tmp.d, am.d, am.d, np, n0, top);
+ bn_flip_n_scatter5_t4(tmp.d, top, powerbuf, 2);
+
+ for (i = 3; i < 32; i++) {
+ /* Calculate a^i = a^(i-1) * a */
+ if (!(*mul_worker) (tmp.d, tmp.d, am.d, np, n0) &&
+ !(*mul_worker) (tmp.d, tmp.d, am.d, np, n0))
+ bn_mul_mont_vis3(tmp.d, tmp.d, am.d, np, n0, top);
+ bn_flip_n_scatter5_t4(tmp.d, top, powerbuf, i);
+ }
+
+ /* switch to 64-bit domain */
+ np = alloca(top * sizeof(BN_ULONG));
+ top /= 2;
+ bn_flip_t4(np, mont->N.d, top);
+
+ /*
+ * The exponent may not have a whole number of fixed-size windows.
+ * To simplify the main loop, the initial window has between 1 and
+ * full-window-size bits such that what remains is always a whole
+ * number of windows
+ */
+ window0 = (bits - 1) % 5 + 1;
+ wmask = (1 << window0) - 1;
+ bits -= window0;
+ wvalue = bn_get_bits(p, bits) & wmask;
+ bn_gather5_t4(tmp.d, top, powerbuf, wvalue);
+
+ /*
+ * Scan the exponent one window at a time starting from the most
+ * significant bits.
+ */
+ while (bits > 0) {
+ if (bits < stride)
+ stride = bits;
+ bits -= stride;
+ wvalue = bn_get_bits(p, bits);
+
+ if ((*pwr5_worker) (tmp.d, np, n0, powerbuf, wvalue, stride))
+ continue;
+ /* retry once and fall back */
+ if ((*pwr5_worker) (tmp.d, np, n0, powerbuf, wvalue, stride))
+ continue;
+
+ bits += stride - 5;
+ wvalue >>= stride - 5;
+ wvalue &= 31;
+ bn_mul_mont_t4(tmp.d, tmp.d, tmp.d, np, n0, top);
+ bn_mul_mont_t4(tmp.d, tmp.d, tmp.d, np, n0, top);
+ bn_mul_mont_t4(tmp.d, tmp.d, tmp.d, np, n0, top);
+ bn_mul_mont_t4(tmp.d, tmp.d, tmp.d, np, n0, top);
+ bn_mul_mont_t4(tmp.d, tmp.d, tmp.d, np, n0, top);
+ bn_mul_mont_gather5_t4(tmp.d, tmp.d, powerbuf, np, n0, top,
+ wvalue);
+ }
+
+ bn_flip_t4(tmp.d, tmp.d, top);
+ top *= 2;
+ /* back to 32-bit domain */
+ tmp.top = top;
+ bn_correct_top(&tmp);
+ OPENSSL_cleanse(np, top * sizeof(BN_ULONG));
+ } else
+#endif
+#if defined(OPENSSL_BN_ASM_MONT5)
+ if (window == 5 && top > 1) {
+ /*
+ * This optimization uses ideas from https://eprint.iacr.org/2011/239,
+ * specifically optimization of cache-timing attack countermeasures,
+ * pre-computation optimization, and Almost Montgomery Multiplication.
+ *
+ * The paper discusses a 4-bit window to optimize 512-bit modular
+ * exponentiation, used in RSA-1024 with CRT, but RSA-1024 is no longer
+ * important.
+ *
+ * |bn_mul_mont_gather5| and |bn_power5| implement the "almost"
+ * reduction variant, so the values here may not be fully reduced.
+ * They are bounded by R (i.e. they fit in |top| words), not |m|.
+ * Additionally, we pass these "almost" reduced inputs into
+ * |bn_mul_mont|, which implements the normal reduction variant.
+ * Given those inputs, |bn_mul_mont| may not give reduced
+ * output, but it will still produce "almost" reduced output.
+ */
+ void bn_mul_mont_gather5(BN_ULONG *rp, const BN_ULONG *ap,
+ const void *table, const BN_ULONG *np,
+ const BN_ULONG *n0, int num, int power);
+ void bn_scatter5(const BN_ULONG *inp, size_t num,
+ void *table, size_t power);
+ void bn_gather5(BN_ULONG *out, size_t num, void *table, size_t power);
+ void bn_power5(BN_ULONG *rp, const BN_ULONG *ap,
+ const void *table, const BN_ULONG *np,
+ const BN_ULONG *n0, int num, int power);
+ int bn_get_bits5(const BN_ULONG *ap, int off);
+
+ BN_ULONG *n0 = mont->n0, *np;
+
+ /*
+ * BN_to_montgomery can contaminate words above .top [in
+ * BN_DEBUG[_DEBUG] build]...
+ */
+ for (i = am.top; i < top; i++)
+ am.d[i] = 0;
+ for (i = tmp.top; i < top; i++)
+ tmp.d[i] = 0;
+
+ /*
+ * copy mont->N.d[] to improve cache locality
+ */
+ for (np = am.d + top, i = 0; i < top; i++)
+ np[i] = mont->N.d[i];
+
+ bn_scatter5(tmp.d, top, powerbuf, 0);
+ bn_scatter5(am.d, am.top, powerbuf, 1);
+ bn_mul_mont(tmp.d, am.d, am.d, np, n0, top);
+ bn_scatter5(tmp.d, top, powerbuf, 2);
+
+# if 0
+ for (i = 3; i < 32; i++) {
+ /* Calculate a^i = a^(i-1) * a */
+ bn_mul_mont_gather5(tmp.d, am.d, powerbuf, np, n0, top, i - 1);
+ bn_scatter5(tmp.d, top, powerbuf, i);
+ }
+# else
+ /* same as above, but uses squaring for 1/2 of operations */
+ for (i = 4; i < 32; i *= 2) {
+ bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
+ bn_scatter5(tmp.d, top, powerbuf, i);
+ }
+ for (i = 3; i < 8; i += 2) {
+ int j;
+ bn_mul_mont_gather5(tmp.d, am.d, powerbuf, np, n0, top, i - 1);
+ bn_scatter5(tmp.d, top, powerbuf, i);
+ for (j = 2 * i; j < 32; j *= 2) {
+ bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
+ bn_scatter5(tmp.d, top, powerbuf, j);
+ }
+ }
+ for (; i < 16; i += 2) {
+ bn_mul_mont_gather5(tmp.d, am.d, powerbuf, np, n0, top, i - 1);
+ bn_scatter5(tmp.d, top, powerbuf, i);
+ bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
+ bn_scatter5(tmp.d, top, powerbuf, 2 * i);
+ }
+ for (; i < 32; i += 2) {
+ bn_mul_mont_gather5(tmp.d, am.d, powerbuf, np, n0, top, i - 1);
+ bn_scatter5(tmp.d, top, powerbuf, i);
+ }
+# endif
+ /*
+ * The exponent may not have a whole number of fixed-size windows.
+ * To simplify the main loop, the initial window has between 1 and
+ * full-window-size bits such that what remains is always a whole
+ * number of windows
+ */
+ window0 = (bits - 1) % 5 + 1;
+ wmask = (1 << window0) - 1;
+ bits -= window0;
+ wvalue = bn_get_bits(p, bits) & wmask;
+ bn_gather5(tmp.d, top, powerbuf, wvalue);
+
+ /*
+ * Scan the exponent one window at a time starting from the most
+ * significant bits.
+ */
+ if (top & 7) {
+ while (bits > 0) {
+ bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
+ bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
+ bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
+ bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
+ bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
+ bn_mul_mont_gather5(tmp.d, tmp.d, powerbuf, np, n0, top,
+ bn_get_bits5(p->d, bits -= 5));
+ }
+ } else {
+ while (bits > 0) {
+ bn_power5(tmp.d, tmp.d, powerbuf, np, n0, top,
+ bn_get_bits5(p->d, bits -= 5));
+ }
+ }
+
+ tmp.top = top;
+ /*
+ * The result is now in |tmp| in Montgomery form, but it may not be
+ * fully reduced. This is within bounds for |BN_from_montgomery|
+ * (tmp < R <= m*R) so it will, when converting from Montgomery form,
+ * produce a fully reduced result.
+ *
+ * This differs from Figure 2 of the paper, which uses AMM(h, 1) to
+ * convert from Montgomery form with unreduced output, followed by an
+ * extra reduction step. In the paper's terminology, we replace
+ * steps 9 and 10 with MM(h, 1).
+ */
+ } else
+#endif
+ {
+ if (!MOD_EXP_CTIME_COPY_TO_PREBUF(&tmp, top, powerbuf, 0, window))
+ goto err;
+ if (!MOD_EXP_CTIME_COPY_TO_PREBUF(&am, top, powerbuf, 1, window))
+ goto err;
+
+ /*
+ * If the window size is greater than 1, then calculate
+ * val[i=2..2^winsize-1]. Powers are computed as a*a^(i-1) (even
+ * powers could instead be computed as (a^(i/2))^2 to use the slight
+ * performance advantage of sqr over mul).
+ */
+ if (window > 1) {
+ if (!bn_mul_mont_fixed_top(&tmp, &am, &am, mont, ctx))
+ goto err;
+ if (!MOD_EXP_CTIME_COPY_TO_PREBUF(&tmp, top, powerbuf, 2,
+ window))
+ goto err;
+ for (i = 3; i < numPowers; i++) {
+ /* Calculate a^i = a^(i-1) * a */
+ if (!bn_mul_mont_fixed_top(&tmp, &am, &tmp, mont, ctx))
+ goto err;
+ if (!MOD_EXP_CTIME_COPY_TO_PREBUF(&tmp, top, powerbuf, i,
+ window))
+ goto err;
+ }
+ }
+
+ /*
+ * The exponent may not have a whole number of fixed-size windows.
+ * To simplify the main loop, the initial window has between 1 and
+ * full-window-size bits such that what remains is always a whole
+ * number of windows
+ */
+ window0 = (bits - 1) % window + 1;
+ wmask = (1 << window0) - 1;
+ bits -= window0;
+ wvalue = bn_get_bits(p, bits) & wmask;
+ if (!MOD_EXP_CTIME_COPY_FROM_PREBUF(&tmp, top, powerbuf, wvalue,
+ window))
+ goto err;
+
+ wmask = (1 << window) - 1;
+ /*
+ * Scan the exponent one window at a time starting from the most
+ * significant bits.
+ */
+ while (bits > 0) {
+
+ /* Square the result window-size times */
+ for (i = 0; i < window; i++)
+ if (!bn_mul_mont_fixed_top(&tmp, &tmp, &tmp, mont, ctx))
+ goto err;
+
+ /*
+ * Get a window's worth of bits from the exponent
+ * This avoids calling BN_is_bit_set for each bit, which
+ * is not only slower but also makes each bit vulnerable to
+ * EM (and likely other) side-channel attacks like One&Done
+ * (for details see "One&Done: A Single-Decryption EM-Based
+ * Attack on OpenSSL's Constant-Time Blinded RSA" by M. Alam,
+ * H. Khan, M. Dey, N. Sinha, R. Callan, A. Zajic, and
+ * M. Prvulovic, in USENIX Security'18)
+ */
+ bits -= window;
+ wvalue = bn_get_bits(p, bits) & wmask;
+ /*
+ * Fetch the appropriate pre-computed value from the pre-buf
+ */
+ if (!MOD_EXP_CTIME_COPY_FROM_PREBUF(&am, top, powerbuf, wvalue,
+ window))
+ goto err;
+
+ /* Multiply the result into the intermediate result */
+ if (!bn_mul_mont_fixed_top(&tmp, &tmp, &am, mont, ctx))
+ goto err;
+ }
+ }
+
+ /*
+ * Done with zero-padded intermediate BIGNUMs. Final BN_from_montgomery
+ * removes padding [if any] and makes return value suitable for public
+ * API consumer.
+ */
+#if defined(SPARC_T4_MONT)
+ if (OPENSSL_sparcv9cap_P[0] & (SPARCV9_VIS3 | SPARCV9_PREFER_FPU)) {
+ am.d[0] = 1; /* borrow am */
+ for (i = 1; i < top; i++)
+ am.d[i] = 0;
+ if (!BN_mod_mul_montgomery(rr, &tmp, &am, mont, ctx))
+ goto err;
+ } else
+#endif
+ if (!BN_from_montgomery(rr, &tmp, mont, ctx))
+ goto err;
+ ret = 1;
+ err:
+ if (in_mont == NULL)
+ BN_MONT_CTX_free(mont);
+ if (powerbuf != NULL) {
+ OPENSSL_cleanse(powerbuf, powerbufLen);
+ OPENSSL_free(powerbufFree);
+ }
+ BN_CTX_end(ctx);
+ return ret;
+}
+
+int BN_mod_exp_mont_word(BIGNUM *rr, BN_ULONG a, const BIGNUM *p,
+ const BIGNUM *m, BN_CTX *ctx, BN_MONT_CTX *in_mont)
+{
+ BN_MONT_CTX *mont = NULL;
+ int b, bits, ret = 0;
+ int r_is_one;
+ BN_ULONG w, next_w;
+ BIGNUM *r, *t;
+ BIGNUM *swap_tmp;
+#define BN_MOD_MUL_WORD(r, w, m) \
+ (BN_mul_word(r, (w)) && \
+ (/* BN_ucmp(r, (m)) < 0 ? 1 :*/ \
+ (BN_mod(t, r, m, ctx) && (swap_tmp = r, r = t, t = swap_tmp, 1))))
+ /*
+ * BN_MOD_MUL_WORD is only used with 'w' large, so the BN_ucmp test is
+ * probably more overhead than always using BN_mod (which uses BN_copy if
+ * a similar test returns true).
+ */
+ /*
+ * We can use BN_mod and do not need BN_nnmod because our accumulator is
+ * never negative (the result of BN_mod does not depend on the sign of
+ * the modulus).
+ */
+#define BN_TO_MONTGOMERY_WORD(r, w, mont) \
+ (BN_set_word(r, (w)) && BN_to_montgomery(r, r, (mont), ctx))
+
+ if (BN_get_flags(p, BN_FLG_CONSTTIME) != 0
+ || BN_get_flags(m, BN_FLG_CONSTTIME) != 0) {
+ /* BN_FLG_CONSTTIME only supported by BN_mod_exp_mont() */
+ BNerr(BN_F_BN_MOD_EXP_MONT_WORD, ERR_R_SHOULD_NOT_HAVE_BEEN_CALLED);
+ return 0;
+ }
+
+ bn_check_top(p);
+ bn_check_top(m);
+
+ if (!BN_is_odd(m)) {
+ BNerr(BN_F_BN_MOD_EXP_MONT_WORD, BN_R_CALLED_WITH_EVEN_MODULUS);
+ return 0;
+ }
+ if (m->top == 1)
+ a %= m->d[0]; /* make sure that 'a' is reduced */
+
+ bits = BN_num_bits(p);
+ if (bits == 0) {
+ /* x**0 mod 1, or x**0 mod -1 is still zero. */
+ if (BN_abs_is_word(m, 1)) {
+ ret = 1;
+ BN_zero(rr);
+ } else {
+ ret = BN_one(rr);
+ }
+ return ret;
+ }
+ if (a == 0) {
+ BN_zero(rr);
+ ret = 1;
+ return ret;
+ }
+
+ BN_CTX_start(ctx);
+ r = BN_CTX_get(ctx);
+ t = BN_CTX_get(ctx);
+ if (t == NULL)
+ goto err;
+
+ if (in_mont != NULL)
+ mont = in_mont;
+ else {
+ if ((mont = BN_MONT_CTX_new()) == NULL)
+ goto err;
+ if (!BN_MONT_CTX_set(mont, m, ctx))
+ goto err;
+ }
+
+ r_is_one = 1; /* except for Montgomery factor */
+
+ /* bits-1 >= 0 */
+
+ /* The result is accumulated in the product r*w. */
+ w = a; /* bit 'bits-1' of 'p' is always set */
+ for (b = bits - 2; b >= 0; b--) {
+ /* First, square r*w. */
+ next_w = w * w;
+ if ((next_w / w) != w) { /* overflow */
+ if (r_is_one) {
+ if (!BN_TO_MONTGOMERY_WORD(r, w, mont))
+ goto err;
+ r_is_one = 0;
+ } else {
+ if (!BN_MOD_MUL_WORD(r, w, m))
+ goto err;
+ }
+ next_w = 1;
+ }
+ w = next_w;
+ if (!r_is_one) {
+ if (!BN_mod_mul_montgomery(r, r, r, mont, ctx))
+ goto err;
+ }
+
+ /* Second, multiply r*w by 'a' if exponent bit is set. */
+ if (BN_is_bit_set(p, b)) {
+ next_w = w * a;
+ if ((next_w / a) != w) { /* overflow */
+ if (r_is_one) {
+ if (!BN_TO_MONTGOMERY_WORD(r, w, mont))
+ goto err;
+ r_is_one = 0;
+ } else {
+ if (!BN_MOD_MUL_WORD(r, w, m))
+ goto err;
+ }
+ next_w = a;
+ }
+ w = next_w;
+ }
+ }
+
+ /* Finally, set r:=r*w. */
+ if (w != 1) {
+ if (r_is_one) {
+ if (!BN_TO_MONTGOMERY_WORD(r, w, mont))
+ goto err;
+ r_is_one = 0;
+ } else {
+ if (!BN_MOD_MUL_WORD(r, w, m))
+ goto err;
+ }
+ }
+
+ if (r_is_one) { /* can happen only if a == 1 */
+ if (!BN_one(rr))
+ goto err;
+ } else {
+ if (!BN_from_montgomery(rr, r, mont, ctx))
+ goto err;
+ }
+ ret = 1;
+ err:
+ if (in_mont == NULL)
+ BN_MONT_CTX_free(mont);
+ BN_CTX_end(ctx);
+ bn_check_top(rr);
+ return ret;
+}
+
+/* The old fallback, simple version :-) */
+int BN_mod_exp_simple(BIGNUM *r, const BIGNUM *a, const BIGNUM *p,
+ const BIGNUM *m, BN_CTX *ctx)
+{
+ int i, j, bits, ret = 0, wstart, wend, window, wvalue;
+ int start = 1;
+ BIGNUM *d;
+ /* Table of variables obtained from 'ctx' */
+ BIGNUM *val[TABLE_SIZE];
+
+ if (BN_get_flags(p, BN_FLG_CONSTTIME) != 0
+ || BN_get_flags(a, BN_FLG_CONSTTIME) != 0
+ || BN_get_flags(m, BN_FLG_CONSTTIME) != 0) {
+ /* BN_FLG_CONSTTIME only supported by BN_mod_exp_mont() */
+ BNerr(BN_F_BN_MOD_EXP_SIMPLE, ERR_R_SHOULD_NOT_HAVE_BEEN_CALLED);
+ return 0;
+ }
+
+ bits = BN_num_bits(p);
+ if (bits == 0) {
+ /* x**0 mod 1, or x**0 mod -1 is still zero. */
+ if (BN_abs_is_word(m, 1)) {
+ ret = 1;
+ BN_zero(r);
+ } else {
+ ret = BN_one(r);
+ }
+ return ret;
+ }
+
+ BN_CTX_start(ctx);
+ d = BN_CTX_get(ctx);
+ val[0] = BN_CTX_get(ctx);
+ if (val[0] == NULL)
+ goto err;
+
+ if (!BN_nnmod(val[0], a, m, ctx))
+ goto err; /* 1 */
+ if (BN_is_zero(val[0])) {
+ BN_zero(r);
+ ret = 1;
+ goto err;
+ }
+
+ window = BN_window_bits_for_exponent_size(bits);
+ if (window > 1) {
+ if (!BN_mod_mul(d, val[0], val[0], m, ctx))
+ goto err; /* 2 */
+ j = 1 << (window - 1);
+ for (i = 1; i < j; i++) {
+ if (((val[i] = BN_CTX_get(ctx)) == NULL) ||
+ !BN_mod_mul(val[i], val[i - 1], d, m, ctx))
+ goto err;
+ }
+ }
+
+ start = 1; /* This is used to avoid multiplication etc
+ * when there is only the value '1' in the
+ * buffer. */
+ wvalue = 0; /* The 'value' of the window */
+ wstart = bits - 1; /* The top bit of the window */
+ wend = 0; /* The bottom bit of the window */
+
+ if (!BN_one(r))
+ goto err;
+
+ for (;;) {
+ if (BN_is_bit_set(p, wstart) == 0) {
+ if (!start)
+ if (!BN_mod_mul(r, r, r, m, ctx))
+ goto err;
+ if (wstart == 0)
+ break;
+ wstart--;
+ continue;
+ }
+ /*
+ * We now have wstart on a 'set' bit, we now need to work out how bit
+ * a window to do. To do this we need to scan forward until the last
+ * set bit before the end of the window
+ */
+ j = wstart;
+ wvalue = 1;
+ wend = 0;
+ for (i = 1; i < window; i++) {
+ if (wstart - i < 0)
+ break;
+ if (BN_is_bit_set(p, wstart - i)) {
+ wvalue <<= (i - wend);
+ wvalue |= 1;
+ wend = i;
+ }
+ }
+
+ /* wend is the size of the current window */
+ j = wend + 1;
+ /* add the 'bytes above' */
+ if (!start)
+ for (i = 0; i < j; i++) {
+ if (!BN_mod_mul(r, r, r, m, ctx))
+ goto err;
+ }
+
+ /* wvalue will be an odd number < 2^window */
+ if (!BN_mod_mul(r, r, val[wvalue >> 1], m, ctx))
+ goto err;
+
+ /* move the 'window' down further */
+ wstart -= wend + 1;
+ wvalue = 0;
+ start = 0;
+ if (wstart < 0)
+ break;
+ }
+ ret = 1;
+ err:
+ BN_CTX_end(ctx);
+ bn_check_top(r);
+ return ret;
+}
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_exp2.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_exp2.c
new file mode 100644
index 0000000..eac0896
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_exp2.c
@@ -0,0 +1,201 @@
+/*
+ * Copyright 1995-2022 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include <stdio.h>
+#include "internal/cryptlib.h"
+#include "bn_local.h"
+
+#define TABLE_SIZE 32
+
+int BN_mod_exp2_mont(BIGNUM *rr, const BIGNUM *a1, const BIGNUM *p1,
+ const BIGNUM *a2, const BIGNUM *p2, const BIGNUM *m,
+ BN_CTX *ctx, BN_MONT_CTX *in_mont)
+{
+ int i, j, bits, b, bits1, bits2, ret =
+ 0, wpos1, wpos2, window1, window2, wvalue1, wvalue2;
+ int r_is_one = 1;
+ BIGNUM *d, *r;
+ const BIGNUM *a_mod_m;
+ /* Tables of variables obtained from 'ctx' */
+ BIGNUM *val1[TABLE_SIZE], *val2[TABLE_SIZE];
+ BN_MONT_CTX *mont = NULL;
+
+ bn_check_top(a1);
+ bn_check_top(p1);
+ bn_check_top(a2);
+ bn_check_top(p2);
+ bn_check_top(m);
+
+ if (!BN_is_odd(m)) {
+ BNerr(BN_F_BN_MOD_EXP2_MONT, BN_R_CALLED_WITH_EVEN_MODULUS);
+ return 0;
+ }
+ bits1 = BN_num_bits(p1);
+ bits2 = BN_num_bits(p2);
+ if ((bits1 == 0) && (bits2 == 0)) {
+ ret = BN_one(rr);
+ return ret;
+ }
+
+ bits = (bits1 > bits2) ? bits1 : bits2;
+
+ BN_CTX_start(ctx);
+ d = BN_CTX_get(ctx);
+ r = BN_CTX_get(ctx);
+ val1[0] = BN_CTX_get(ctx);
+ val2[0] = BN_CTX_get(ctx);
+ if (val2[0] == NULL)
+ goto err;
+
+ if (in_mont != NULL)
+ mont = in_mont;
+ else {
+ if ((mont = BN_MONT_CTX_new()) == NULL)
+ goto err;
+ if (!BN_MONT_CTX_set(mont, m, ctx))
+ goto err;
+ }
+
+ window1 = BN_window_bits_for_exponent_size(bits1);
+ window2 = BN_window_bits_for_exponent_size(bits2);
+
+ /*
+ * Build table for a1: val1[i] := a1^(2*i + 1) mod m for i = 0 .. 2^(window1-1)
+ */
+ if (a1->neg || BN_ucmp(a1, m) >= 0) {
+ if (!BN_mod(val1[0], a1, m, ctx))
+ goto err;
+ a_mod_m = val1[0];
+ } else
+ a_mod_m = a1;
+ if (BN_is_zero(a_mod_m)) {
+ BN_zero(rr);
+ ret = 1;
+ goto err;
+ }
+
+ if (!BN_to_montgomery(val1[0], a_mod_m, mont, ctx))
+ goto err;
+ if (window1 > 1) {
+ if (!BN_mod_mul_montgomery(d, val1[0], val1[0], mont, ctx))
+ goto err;
+
+ j = 1 << (window1 - 1);
+ for (i = 1; i < j; i++) {
+ if (((val1[i] = BN_CTX_get(ctx)) == NULL) ||
+ !BN_mod_mul_montgomery(val1[i], val1[i - 1], d, mont, ctx))
+ goto err;
+ }
+ }
+
+ /*
+ * Build table for a2: val2[i] := a2^(2*i + 1) mod m for i = 0 .. 2^(window2-1)
+ */
+ if (a2->neg || BN_ucmp(a2, m) >= 0) {
+ if (!BN_mod(val2[0], a2, m, ctx))
+ goto err;
+ a_mod_m = val2[0];
+ } else
+ a_mod_m = a2;
+ if (BN_is_zero(a_mod_m)) {
+ BN_zero(rr);
+ ret = 1;
+ goto err;
+ }
+ if (!BN_to_montgomery(val2[0], a_mod_m, mont, ctx))
+ goto err;
+ if (window2 > 1) {
+ if (!BN_mod_mul_montgomery(d, val2[0], val2[0], mont, ctx))
+ goto err;
+
+ j = 1 << (window2 - 1);
+ for (i = 1; i < j; i++) {
+ if (((val2[i] = BN_CTX_get(ctx)) == NULL) ||
+ !BN_mod_mul_montgomery(val2[i], val2[i - 1], d, mont, ctx))
+ goto err;
+ }
+ }
+
+ /* Now compute the power product, using independent windows. */
+ r_is_one = 1;
+ wvalue1 = 0; /* The 'value' of the first window */
+ wvalue2 = 0; /* The 'value' of the second window */
+ wpos1 = 0; /* If wvalue1 > 0, the bottom bit of the
+ * first window */
+ wpos2 = 0; /* If wvalue2 > 0, the bottom bit of the
+ * second window */
+
+ if (!BN_to_montgomery(r, BN_value_one(), mont, ctx))
+ goto err;
+ for (b = bits - 1; b >= 0; b--) {
+ if (!r_is_one) {
+ if (!BN_mod_mul_montgomery(r, r, r, mont, ctx))
+ goto err;
+ }
+
+ if (!wvalue1)
+ if (BN_is_bit_set(p1, b)) {
+ /*
+ * consider bits b-window1+1 .. b for this window
+ */
+ i = b - window1 + 1;
+ while (!BN_is_bit_set(p1, i)) /* works for i<0 */
+ i++;
+ wpos1 = i;
+ wvalue1 = 1;
+ for (i = b - 1; i >= wpos1; i--) {
+ wvalue1 <<= 1;
+ if (BN_is_bit_set(p1, i))
+ wvalue1++;
+ }
+ }
+
+ if (!wvalue2)
+ if (BN_is_bit_set(p2, b)) {
+ /*
+ * consider bits b-window2+1 .. b for this window
+ */
+ i = b - window2 + 1;
+ while (!BN_is_bit_set(p2, i))
+ i++;
+ wpos2 = i;
+ wvalue2 = 1;
+ for (i = b - 1; i >= wpos2; i--) {
+ wvalue2 <<= 1;
+ if (BN_is_bit_set(p2, i))
+ wvalue2++;
+ }
+ }
+
+ if (wvalue1 && b == wpos1) {
+ /* wvalue1 is odd and < 2^window1 */
+ if (!BN_mod_mul_montgomery(r, r, val1[wvalue1 >> 1], mont, ctx))
+ goto err;
+ wvalue1 = 0;
+ r_is_one = 0;
+ }
+
+ if (wvalue2 && b == wpos2) {
+ /* wvalue2 is odd and < 2^window2 */
+ if (!BN_mod_mul_montgomery(r, r, val2[wvalue2 >> 1], mont, ctx))
+ goto err;
+ wvalue2 = 0;
+ r_is_one = 0;
+ }
+ }
+ if (!BN_from_montgomery(rr, r, mont, ctx))
+ goto err;
+ ret = 1;
+ err:
+ if (in_mont == NULL)
+ BN_MONT_CTX_free(mont);
+ BN_CTX_end(ctx);
+ bn_check_top(rr);
+ return ret;
+}
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_gcd.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_gcd.c
new file mode 100644
index 0000000..0941f7b
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_gcd.c
@@ -0,0 +1,645 @@
+/*
+ * Copyright 1995-2020 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include "internal/cryptlib.h"
+#include "bn_local.h"
+
+/*
+ * bn_mod_inverse_no_branch is a special version of BN_mod_inverse. It does
+ * not contain branches that may leak sensitive information.
+ *
+ * This is a static function, we ensure all callers in this file pass valid
+ * arguments: all passed pointers here are non-NULL.
+ */
+static ossl_inline
+BIGNUM *bn_mod_inverse_no_branch(BIGNUM *in,
+ const BIGNUM *a, const BIGNUM *n,
+ BN_CTX *ctx, int *pnoinv)
+{
+ BIGNUM *A, *B, *X, *Y, *M, *D, *T, *R = NULL;
+ BIGNUM *ret = NULL;
+ int sign;
+
+ bn_check_top(a);
+ bn_check_top(n);
+
+ BN_CTX_start(ctx);
+ A = BN_CTX_get(ctx);
+ B = BN_CTX_get(ctx);
+ X = BN_CTX_get(ctx);
+ D = BN_CTX_get(ctx);
+ M = BN_CTX_get(ctx);
+ Y = BN_CTX_get(ctx);
+ T = BN_CTX_get(ctx);
+ if (T == NULL)
+ goto err;
+
+ if (in == NULL)
+ R = BN_new();
+ else
+ R = in;
+ if (R == NULL)
+ goto err;
+
+ BN_one(X);
+ BN_zero(Y);
+ if (BN_copy(B, a) == NULL)
+ goto err;
+ if (BN_copy(A, n) == NULL)
+ goto err;
+ A->neg = 0;
+
+ if (B->neg || (BN_ucmp(B, A) >= 0)) {
+ /*
+ * Turn BN_FLG_CONSTTIME flag on, so that when BN_div is invoked,
+ * BN_div_no_branch will be called eventually.
+ */
+ {
+ BIGNUM local_B;
+ bn_init(&local_B);
+ BN_with_flags(&local_B, B, BN_FLG_CONSTTIME);
+ if (!BN_nnmod(B, &local_B, A, ctx))
+ goto err;
+ /* Ensure local_B goes out of scope before any further use of B */
+ }
+ }
+ sign = -1;
+ /*-
+ * From B = a mod |n|, A = |n| it follows that
+ *
+ * 0 <= B < A,
+ * -sign*X*a == B (mod |n|),
+ * sign*Y*a == A (mod |n|).
+ */
+
+ while (!BN_is_zero(B)) {
+ BIGNUM *tmp;
+
+ /*-
+ * 0 < B < A,
+ * (*) -sign*X*a == B (mod |n|),
+ * sign*Y*a == A (mod |n|)
+ */
+
+ /*
+ * Turn BN_FLG_CONSTTIME flag on, so that when BN_div is invoked,
+ * BN_div_no_branch will be called eventually.
+ */
+ {
+ BIGNUM local_A;
+ bn_init(&local_A);
+ BN_with_flags(&local_A, A, BN_FLG_CONSTTIME);
+
+ /* (D, M) := (A/B, A%B) ... */
+ if (!BN_div(D, M, &local_A, B, ctx))
+ goto err;
+ /* Ensure local_A goes out of scope before any further use of A */
+ }
+
+ /*-
+ * Now
+ * A = D*B + M;
+ * thus we have
+ * (**) sign*Y*a == D*B + M (mod |n|).
+ */
+
+ tmp = A; /* keep the BIGNUM object, the value does not
+ * matter */
+
+ /* (A, B) := (B, A mod B) ... */
+ A = B;
+ B = M;
+ /* ... so we have 0 <= B < A again */
+
+ /*-
+ * Since the former M is now B and the former B is now A,
+ * (**) translates into
+ * sign*Y*a == D*A + B (mod |n|),
+ * i.e.
+ * sign*Y*a - D*A == B (mod |n|).
+ * Similarly, (*) translates into
+ * -sign*X*a == A (mod |n|).
+ *
+ * Thus,
+ * sign*Y*a + D*sign*X*a == B (mod |n|),
+ * i.e.
+ * sign*(Y + D*X)*a == B (mod |n|).
+ *
+ * So if we set (X, Y, sign) := (Y + D*X, X, -sign), we arrive back at
+ * -sign*X*a == B (mod |n|),
+ * sign*Y*a == A (mod |n|).
+ * Note that X and Y stay non-negative all the time.
+ */
+
+ if (!BN_mul(tmp, D, X, ctx))
+ goto err;
+ if (!BN_add(tmp, tmp, Y))
+ goto err;
+
+ M = Y; /* keep the BIGNUM object, the value does not
+ * matter */
+ Y = X;
+ X = tmp;
+ sign = -sign;
+ }
+
+ /*-
+ * The while loop (Euclid's algorithm) ends when
+ * A == gcd(a,n);
+ * we have
+ * sign*Y*a == A (mod |n|),
+ * where Y is non-negative.
+ */
+
+ if (sign < 0) {
+ if (!BN_sub(Y, n, Y))
+ goto err;
+ }
+ /* Now Y*a == A (mod |n|). */
+
+ if (BN_is_one(A)) {
+ /* Y*a == 1 (mod |n|) */
+ if (!Y->neg && BN_ucmp(Y, n) < 0) {
+ if (!BN_copy(R, Y))
+ goto err;
+ } else {
+ if (!BN_nnmod(R, Y, n, ctx))
+ goto err;
+ }
+ } else {
+ *pnoinv = 1;
+ /* caller sets the BN_R_NO_INVERSE error */
+ goto err;
+ }
+
+ ret = R;
+ *pnoinv = 0;
+
+ err:
+ if ((ret == NULL) && (in == NULL))
+ BN_free(R);
+ BN_CTX_end(ctx);
+ bn_check_top(ret);
+ return ret;
+}
+
+/*
+ * This is an internal function, we assume all callers pass valid arguments:
+ * all pointers passed here are assumed non-NULL.
+ */
+BIGNUM *int_bn_mod_inverse(BIGNUM *in,
+ const BIGNUM *a, const BIGNUM *n, BN_CTX *ctx,
+ int *pnoinv)
+{
+ BIGNUM *A, *B, *X, *Y, *M, *D, *T, *R = NULL;
+ BIGNUM *ret = NULL;
+ int sign;
+
+ /* This is invalid input so we don't worry about constant time here */
+ if (BN_abs_is_word(n, 1) || BN_is_zero(n)) {
+ *pnoinv = 1;
+ return NULL;
+ }
+
+ *pnoinv = 0;
+
+ if ((BN_get_flags(a, BN_FLG_CONSTTIME) != 0)
+ || (BN_get_flags(n, BN_FLG_CONSTTIME) != 0)) {
+ return bn_mod_inverse_no_branch(in, a, n, ctx, pnoinv);
+ }
+
+ bn_check_top(a);
+ bn_check_top(n);
+
+ BN_CTX_start(ctx);
+ A = BN_CTX_get(ctx);
+ B = BN_CTX_get(ctx);
+ X = BN_CTX_get(ctx);
+ D = BN_CTX_get(ctx);
+ M = BN_CTX_get(ctx);
+ Y = BN_CTX_get(ctx);
+ T = BN_CTX_get(ctx);
+ if (T == NULL)
+ goto err;
+
+ if (in == NULL)
+ R = BN_new();
+ else
+ R = in;
+ if (R == NULL)
+ goto err;
+
+ BN_one(X);
+ BN_zero(Y);
+ if (BN_copy(B, a) == NULL)
+ goto err;
+ if (BN_copy(A, n) == NULL)
+ goto err;
+ A->neg = 0;
+ if (B->neg || (BN_ucmp(B, A) >= 0)) {
+ if (!BN_nnmod(B, B, A, ctx))
+ goto err;
+ }
+ sign = -1;
+ /*-
+ * From B = a mod |n|, A = |n| it follows that
+ *
+ * 0 <= B < A,
+ * -sign*X*a == B (mod |n|),
+ * sign*Y*a == A (mod |n|).
+ */
+
+ if (BN_is_odd(n) && (BN_num_bits(n) <= 2048)) {
+ /*
+ * Binary inversion algorithm; requires odd modulus. This is faster
+ * than the general algorithm if the modulus is sufficiently small
+ * (about 400 .. 500 bits on 32-bit systems, but much more on 64-bit
+ * systems)
+ */
+ int shift;
+
+ while (!BN_is_zero(B)) {
+ /*-
+ * 0 < B < |n|,
+ * 0 < A <= |n|,
+ * (1) -sign*X*a == B (mod |n|),
+ * (2) sign*Y*a == A (mod |n|)
+ */
+
+ /*
+ * Now divide B by the maximum possible power of two in the
+ * integers, and divide X by the same value mod |n|. When we're
+ * done, (1) still holds.
+ */
+ shift = 0;
+ while (!BN_is_bit_set(B, shift)) { /* note that 0 < B */
+ shift++;
+
+ if (BN_is_odd(X)) {
+ if (!BN_uadd(X, X, n))
+ goto err;
+ }
+ /*
+ * now X is even, so we can easily divide it by two
+ */
+ if (!BN_rshift1(X, X))
+ goto err;
+ }
+ if (shift > 0) {
+ if (!BN_rshift(B, B, shift))
+ goto err;
+ }
+
+ /*
+ * Same for A and Y. Afterwards, (2) still holds.
+ */
+ shift = 0;
+ while (!BN_is_bit_set(A, shift)) { /* note that 0 < A */
+ shift++;
+
+ if (BN_is_odd(Y)) {
+ if (!BN_uadd(Y, Y, n))
+ goto err;
+ }
+ /* now Y is even */
+ if (!BN_rshift1(Y, Y))
+ goto err;
+ }
+ if (shift > 0) {
+ if (!BN_rshift(A, A, shift))
+ goto err;
+ }
+
+ /*-
+ * We still have (1) and (2).
+ * Both A and B are odd.
+ * The following computations ensure that
+ *
+ * 0 <= B < |n|,
+ * 0 < A < |n|,
+ * (1) -sign*X*a == B (mod |n|),
+ * (2) sign*Y*a == A (mod |n|),
+ *
+ * and that either A or B is even in the next iteration.
+ */
+ if (BN_ucmp(B, A) >= 0) {
+ /* -sign*(X + Y)*a == B - A (mod |n|) */
+ if (!BN_uadd(X, X, Y))
+ goto err;
+ /*
+ * NB: we could use BN_mod_add_quick(X, X, Y, n), but that
+ * actually makes the algorithm slower
+ */
+ if (!BN_usub(B, B, A))
+ goto err;
+ } else {
+ /* sign*(X + Y)*a == A - B (mod |n|) */
+ if (!BN_uadd(Y, Y, X))
+ goto err;
+ /*
+ * as above, BN_mod_add_quick(Y, Y, X, n) would slow things down
+ */
+ if (!BN_usub(A, A, B))
+ goto err;
+ }
+ }
+ } else {
+ /* general inversion algorithm */
+
+ while (!BN_is_zero(B)) {
+ BIGNUM *tmp;
+
+ /*-
+ * 0 < B < A,
+ * (*) -sign*X*a == B (mod |n|),
+ * sign*Y*a == A (mod |n|)
+ */
+
+ /* (D, M) := (A/B, A%B) ... */
+ if (BN_num_bits(A) == BN_num_bits(B)) {
+ if (!BN_one(D))
+ goto err;
+ if (!BN_sub(M, A, B))
+ goto err;
+ } else if (BN_num_bits(A) == BN_num_bits(B) + 1) {
+ /* A/B is 1, 2, or 3 */
+ if (!BN_lshift1(T, B))
+ goto err;
+ if (BN_ucmp(A, T) < 0) {
+ /* A < 2*B, so D=1 */
+ if (!BN_one(D))
+ goto err;
+ if (!BN_sub(M, A, B))
+ goto err;
+ } else {
+ /* A >= 2*B, so D=2 or D=3 */
+ if (!BN_sub(M, A, T))
+ goto err;
+ if (!BN_add(D, T, B))
+ goto err; /* use D (:= 3*B) as temp */
+ if (BN_ucmp(A, D) < 0) {
+ /* A < 3*B, so D=2 */
+ if (!BN_set_word(D, 2))
+ goto err;
+ /*
+ * M (= A - 2*B) already has the correct value
+ */
+ } else {
+ /* only D=3 remains */
+ if (!BN_set_word(D, 3))
+ goto err;
+ /*
+ * currently M = A - 2*B, but we need M = A - 3*B
+ */
+ if (!BN_sub(M, M, B))
+ goto err;
+ }
+ }
+ } else {
+ if (!BN_div(D, M, A, B, ctx))
+ goto err;
+ }
+
+ /*-
+ * Now
+ * A = D*B + M;
+ * thus we have
+ * (**) sign*Y*a == D*B + M (mod |n|).
+ */
+
+ tmp = A; /* keep the BIGNUM object, the value does not matter */
+
+ /* (A, B) := (B, A mod B) ... */
+ A = B;
+ B = M;
+ /* ... so we have 0 <= B < A again */
+
+ /*-
+ * Since the former M is now B and the former B is now A,
+ * (**) translates into
+ * sign*Y*a == D*A + B (mod |n|),
+ * i.e.
+ * sign*Y*a - D*A == B (mod |n|).
+ * Similarly, (*) translates into
+ * -sign*X*a == A (mod |n|).
+ *
+ * Thus,
+ * sign*Y*a + D*sign*X*a == B (mod |n|),
+ * i.e.
+ * sign*(Y + D*X)*a == B (mod |n|).
+ *
+ * So if we set (X, Y, sign) := (Y + D*X, X, -sign), we arrive back at
+ * -sign*X*a == B (mod |n|),
+ * sign*Y*a == A (mod |n|).
+ * Note that X and Y stay non-negative all the time.
+ */
+
+ /*
+ * most of the time D is very small, so we can optimize tmp := D*X+Y
+ */
+ if (BN_is_one(D)) {
+ if (!BN_add(tmp, X, Y))
+ goto err;
+ } else {
+ if (BN_is_word(D, 2)) {
+ if (!BN_lshift1(tmp, X))
+ goto err;
+ } else if (BN_is_word(D, 4)) {
+ if (!BN_lshift(tmp, X, 2))
+ goto err;
+ } else if (D->top == 1) {
+ if (!BN_copy(tmp, X))
+ goto err;
+ if (!BN_mul_word(tmp, D->d[0]))
+ goto err;
+ } else {
+ if (!BN_mul(tmp, D, X, ctx))
+ goto err;
+ }
+ if (!BN_add(tmp, tmp, Y))
+ goto err;
+ }
+
+ M = Y; /* keep the BIGNUM object, the value does not matter */
+ Y = X;
+ X = tmp;
+ sign = -sign;
+ }
+ }
+
+ /*-
+ * The while loop (Euclid's algorithm) ends when
+ * A == gcd(a,n);
+ * we have
+ * sign*Y*a == A (mod |n|),
+ * where Y is non-negative.
+ */
+
+ if (sign < 0) {
+ if (!BN_sub(Y, n, Y))
+ goto err;
+ }
+ /* Now Y*a == A (mod |n|). */
+
+ if (BN_is_one(A)) {
+ /* Y*a == 1 (mod |n|) */
+ if (!Y->neg && BN_ucmp(Y, n) < 0) {
+ if (!BN_copy(R, Y))
+ goto err;
+ } else {
+ if (!BN_nnmod(R, Y, n, ctx))
+ goto err;
+ }
+ } else {
+ *pnoinv = 1;
+ goto err;
+ }
+ ret = R;
+ err:
+ if ((ret == NULL) && (in == NULL))
+ BN_free(R);
+ BN_CTX_end(ctx);
+ bn_check_top(ret);
+ return ret;
+}
+
+/* solves ax == 1 (mod n) */
+BIGNUM *BN_mod_inverse(BIGNUM *in,
+ const BIGNUM *a, const BIGNUM *n, BN_CTX *ctx)
+{
+ BN_CTX *new_ctx = NULL;
+ BIGNUM *rv;
+ int noinv = 0;
+
+ if (ctx == NULL) {
+ ctx = new_ctx = BN_CTX_new();
+ if (ctx == NULL) {
+ BNerr(BN_F_BN_MOD_INVERSE, ERR_R_MALLOC_FAILURE);
+ return NULL;
+ }
+ }
+
+ rv = int_bn_mod_inverse(in, a, n, ctx, &noinv);
+ if (noinv)
+ BNerr(BN_F_BN_MOD_INVERSE, BN_R_NO_INVERSE);
+ BN_CTX_free(new_ctx);
+ return rv;
+}
+
+/*-
+ * This function is based on the constant-time GCD work by Bernstein and Yang:
+ * https://eprint.iacr.org/2019/266
+ * Generalized fast GCD function to allow even inputs.
+ * The algorithm first finds the shared powers of 2 between
+ * the inputs, and removes them, reducing at least one of the
+ * inputs to an odd value. Then it proceeds to calculate the GCD.
+ * Before returning the resulting GCD, we take care of adding
+ * back the powers of two removed at the beginning.
+ * Note 1: we assume the bit length of both inputs is public information,
+ * since access to top potentially leaks this information.
+ */
+int BN_gcd(BIGNUM *r, const BIGNUM *in_a, const BIGNUM *in_b, BN_CTX *ctx)
+{
+ BIGNUM *g, *temp = NULL;
+ BN_ULONG mask = 0;
+ int i, j, top, rlen, glen, m, bit = 1, delta = 1, cond = 0, shifts = 0, ret = 0;
+
+ /* Note 2: zero input corner cases are not constant-time since they are
+ * handled immediately. An attacker can run an attack under this
+ * assumption without the need of side-channel information. */
+ if (BN_is_zero(in_b)) {
+ ret = BN_copy(r, in_a) != NULL;
+ r->neg = 0;
+ return ret;
+ }
+ if (BN_is_zero(in_a)) {
+ ret = BN_copy(r, in_b) != NULL;
+ r->neg = 0;
+ return ret;
+ }
+
+ bn_check_top(in_a);
+ bn_check_top(in_b);
+
+ BN_CTX_start(ctx);
+ temp = BN_CTX_get(ctx);
+ g = BN_CTX_get(ctx);
+
+ /* make r != 0, g != 0 even, so BN_rshift is not a potential nop */
+ if (g == NULL
+ || !BN_lshift1(g, in_b)
+ || !BN_lshift1(r, in_a))
+ goto err;
+
+ /* find shared powers of two, i.e. "shifts" >= 1 */
+ for (i = 0; i < r->dmax && i < g->dmax; i++) {
+ mask = ~(r->d[i] | g->d[i]);
+ for (j = 0; j < BN_BITS2; j++) {
+ bit &= mask;
+ shifts += bit;
+ mask >>= 1;
+ }
+ }
+
+ /* subtract shared powers of two; shifts >= 1 */
+ if (!BN_rshift(r, r, shifts)
+ || !BN_rshift(g, g, shifts))
+ goto err;
+
+ /* expand to biggest nword, with room for a possible extra word */
+ top = 1 + ((r->top >= g->top) ? r->top : g->top);
+ if (bn_wexpand(r, top) == NULL
+ || bn_wexpand(g, top) == NULL
+ || bn_wexpand(temp, top) == NULL)
+ goto err;
+
+ /* re arrange inputs s.t. r is odd */
+ BN_consttime_swap((~r->d[0]) & 1, r, g, top);
+
+ /* compute the number of iterations */
+ rlen = BN_num_bits(r);
+ glen = BN_num_bits(g);
+ m = 4 + 3 * ((rlen >= glen) ? rlen : glen);
+
+ for (i = 0; i < m; i++) {
+ /* conditionally flip signs if delta is positive and g is odd */
+ cond = (-delta >> (8 * sizeof(delta) - 1)) & g->d[0] & 1
+ /* make sure g->top > 0 (i.e. if top == 0 then g == 0 always) */
+ & (~((g->top - 1) >> (sizeof(g->top) * 8 - 1)));
+ delta = (-cond & -delta) | ((cond - 1) & delta);
+ r->neg ^= cond;
+ /* swap */
+ BN_consttime_swap(cond, r, g, top);
+
+ /* elimination step */
+ delta++;
+ if (!BN_add(temp, g, r))
+ goto err;
+ BN_consttime_swap(g->d[0] & 1 /* g is odd */
+ /* make sure g->top > 0 (i.e. if top == 0 then g == 0 always) */
+ & (~((g->top - 1) >> (sizeof(g->top) * 8 - 1))),
+ g, temp, top);
+ if (!BN_rshift1(g, g))
+ goto err;
+ }
+
+ /* remove possible negative sign */
+ r->neg = 0;
+ /* add powers of 2 removed, then correct the artificial shift */
+ if (!BN_lshift(r, r, shifts)
+ || !BN_rshift1(r, r))
+ goto err;
+
+ ret = 1;
+
+ err:
+ BN_CTX_end(ctx);
+ bn_check_top(r);
+ return ret;
+}
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_gf2m.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_gf2m.c
new file mode 100644
index 0000000..a2ea867
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_gf2m.c
@@ -0,0 +1,1166 @@
+/*
+ * Copyright 2002-2018 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright (c) 2002, Oracle and/or its affiliates. All rights reserved
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include <assert.h>
+#include <limits.h>
+#include <stdio.h>
+#include "internal/cryptlib.h"
+#include "bn_local.h"
+
+#ifndef OPENSSL_NO_EC2M
+
+/*
+ * Maximum number of iterations before BN_GF2m_mod_solve_quad_arr should
+ * fail.
+ */
+# define MAX_ITERATIONS 50
+
+# define SQR_nibble(w) ((((w) & 8) << 3) \
+ | (((w) & 4) << 2) \
+ | (((w) & 2) << 1) \
+ | ((w) & 1))
+
+
+/* Platform-specific macros to accelerate squaring. */
+# if defined(SIXTY_FOUR_BIT) || defined(SIXTY_FOUR_BIT_LONG)
+# define SQR1(w) \
+ SQR_nibble((w) >> 60) << 56 | SQR_nibble((w) >> 56) << 48 | \
+ SQR_nibble((w) >> 52) << 40 | SQR_nibble((w) >> 48) << 32 | \
+ SQR_nibble((w) >> 44) << 24 | SQR_nibble((w) >> 40) << 16 | \
+ SQR_nibble((w) >> 36) << 8 | SQR_nibble((w) >> 32)
+# define SQR0(w) \
+ SQR_nibble((w) >> 28) << 56 | SQR_nibble((w) >> 24) << 48 | \
+ SQR_nibble((w) >> 20) << 40 | SQR_nibble((w) >> 16) << 32 | \
+ SQR_nibble((w) >> 12) << 24 | SQR_nibble((w) >> 8) << 16 | \
+ SQR_nibble((w) >> 4) << 8 | SQR_nibble((w) )
+# endif
+# ifdef THIRTY_TWO_BIT
+# define SQR1(w) \
+ SQR_nibble((w) >> 28) << 24 | SQR_nibble((w) >> 24) << 16 | \
+ SQR_nibble((w) >> 20) << 8 | SQR_nibble((w) >> 16)
+# define SQR0(w) \
+ SQR_nibble((w) >> 12) << 24 | SQR_nibble((w) >> 8) << 16 | \
+ SQR_nibble((w) >> 4) << 8 | SQR_nibble((w) )
+# endif
+
+# if !defined(OPENSSL_BN_ASM_GF2m)
+/*
+ * Product of two polynomials a, b each with degree < BN_BITS2 - 1, result is
+ * a polynomial r with degree < 2 * BN_BITS - 1 The caller MUST ensure that
+ * the variables have the right amount of space allocated.
+ */
+# ifdef THIRTY_TWO_BIT
+static void bn_GF2m_mul_1x1(BN_ULONG *r1, BN_ULONG *r0, const BN_ULONG a,
+ const BN_ULONG b)
+{
+ register BN_ULONG h, l, s;
+ BN_ULONG tab[8], top2b = a >> 30;
+ register BN_ULONG a1, a2, a4;
+
+ a1 = a & (0x3FFFFFFF);
+ a2 = a1 << 1;
+ a4 = a2 << 1;
+
+ tab[0] = 0;
+ tab[1] = a1;
+ tab[2] = a2;
+ tab[3] = a1 ^ a2;
+ tab[4] = a4;
+ tab[5] = a1 ^ a4;
+ tab[6] = a2 ^ a4;
+ tab[7] = a1 ^ a2 ^ a4;
+
+ s = tab[b & 0x7];
+ l = s;
+ s = tab[b >> 3 & 0x7];
+ l ^= s << 3;
+ h = s >> 29;
+ s = tab[b >> 6 & 0x7];
+ l ^= s << 6;
+ h ^= s >> 26;
+ s = tab[b >> 9 & 0x7];
+ l ^= s << 9;
+ h ^= s >> 23;
+ s = tab[b >> 12 & 0x7];
+ l ^= s << 12;
+ h ^= s >> 20;
+ s = tab[b >> 15 & 0x7];
+ l ^= s << 15;
+ h ^= s >> 17;
+ s = tab[b >> 18 & 0x7];
+ l ^= s << 18;
+ h ^= s >> 14;
+ s = tab[b >> 21 & 0x7];
+ l ^= s << 21;
+ h ^= s >> 11;
+ s = tab[b >> 24 & 0x7];
+ l ^= s << 24;
+ h ^= s >> 8;
+ s = tab[b >> 27 & 0x7];
+ l ^= s << 27;
+ h ^= s >> 5;
+ s = tab[b >> 30];
+ l ^= s << 30;
+ h ^= s >> 2;
+
+ /* compensate for the top two bits of a */
+
+ if (top2b & 01) {
+ l ^= b << 30;
+ h ^= b >> 2;
+ }
+ if (top2b & 02) {
+ l ^= b << 31;
+ h ^= b >> 1;
+ }
+
+ *r1 = h;
+ *r0 = l;
+}
+# endif
+# if defined(SIXTY_FOUR_BIT) || defined(SIXTY_FOUR_BIT_LONG)
+static void bn_GF2m_mul_1x1(BN_ULONG *r1, BN_ULONG *r0, const BN_ULONG a,
+ const BN_ULONG b)
+{
+ register BN_ULONG h, l, s;
+ BN_ULONG tab[16], top3b = a >> 61;
+ register BN_ULONG a1, a2, a4, a8;
+
+ a1 = a & (0x1FFFFFFFFFFFFFFFULL);
+ a2 = a1 << 1;
+ a4 = a2 << 1;
+ a8 = a4 << 1;
+
+ tab[0] = 0;
+ tab[1] = a1;
+ tab[2] = a2;
+ tab[3] = a1 ^ a2;
+ tab[4] = a4;
+ tab[5] = a1 ^ a4;
+ tab[6] = a2 ^ a4;
+ tab[7] = a1 ^ a2 ^ a4;
+ tab[8] = a8;
+ tab[9] = a1 ^ a8;
+ tab[10] = a2 ^ a8;
+ tab[11] = a1 ^ a2 ^ a8;
+ tab[12] = a4 ^ a8;
+ tab[13] = a1 ^ a4 ^ a8;
+ tab[14] = a2 ^ a4 ^ a8;
+ tab[15] = a1 ^ a2 ^ a4 ^ a8;
+
+ s = tab[b & 0xF];
+ l = s;
+ s = tab[b >> 4 & 0xF];
+ l ^= s << 4;
+ h = s >> 60;
+ s = tab[b >> 8 & 0xF];
+ l ^= s << 8;
+ h ^= s >> 56;
+ s = tab[b >> 12 & 0xF];
+ l ^= s << 12;
+ h ^= s >> 52;
+ s = tab[b >> 16 & 0xF];
+ l ^= s << 16;
+ h ^= s >> 48;
+ s = tab[b >> 20 & 0xF];
+ l ^= s << 20;
+ h ^= s >> 44;
+ s = tab[b >> 24 & 0xF];
+ l ^= s << 24;
+ h ^= s >> 40;
+ s = tab[b >> 28 & 0xF];
+ l ^= s << 28;
+ h ^= s >> 36;
+ s = tab[b >> 32 & 0xF];
+ l ^= s << 32;
+ h ^= s >> 32;
+ s = tab[b >> 36 & 0xF];
+ l ^= s << 36;
+ h ^= s >> 28;
+ s = tab[b >> 40 & 0xF];
+ l ^= s << 40;
+ h ^= s >> 24;
+ s = tab[b >> 44 & 0xF];
+ l ^= s << 44;
+ h ^= s >> 20;
+ s = tab[b >> 48 & 0xF];
+ l ^= s << 48;
+ h ^= s >> 16;
+ s = tab[b >> 52 & 0xF];
+ l ^= s << 52;
+ h ^= s >> 12;
+ s = tab[b >> 56 & 0xF];
+ l ^= s << 56;
+ h ^= s >> 8;
+ s = tab[b >> 60];
+ l ^= s << 60;
+ h ^= s >> 4;
+
+ /* compensate for the top three bits of a */
+
+ if (top3b & 01) {
+ l ^= b << 61;
+ h ^= b >> 3;
+ }
+ if (top3b & 02) {
+ l ^= b << 62;
+ h ^= b >> 2;
+ }
+ if (top3b & 04) {
+ l ^= b << 63;
+ h ^= b >> 1;
+ }
+
+ *r1 = h;
+ *r0 = l;
+}
+# endif
+
+/*
+ * Product of two polynomials a, b each with degree < 2 * BN_BITS2 - 1,
+ * result is a polynomial r with degree < 4 * BN_BITS2 - 1 The caller MUST
+ * ensure that the variables have the right amount of space allocated.
+ */
+static void bn_GF2m_mul_2x2(BN_ULONG *r, const BN_ULONG a1, const BN_ULONG a0,
+ const BN_ULONG b1, const BN_ULONG b0)
+{
+ BN_ULONG m1, m0;
+ /* r[3] = h1, r[2] = h0; r[1] = l1; r[0] = l0 */
+ bn_GF2m_mul_1x1(r + 3, r + 2, a1, b1);
+ bn_GF2m_mul_1x1(r + 1, r, a0, b0);
+ bn_GF2m_mul_1x1(&m1, &m0, a0 ^ a1, b0 ^ b1);
+ /* Correction on m1 ^= l1 ^ h1; m0 ^= l0 ^ h0; */
+ r[2] ^= m1 ^ r[1] ^ r[3]; /* h0 ^= m1 ^ l1 ^ h1; */
+ r[1] = r[3] ^ r[2] ^ r[0] ^ m1 ^ m0; /* l1 ^= l0 ^ h0 ^ m0; */
+}
+# else
+void bn_GF2m_mul_2x2(BN_ULONG *r, BN_ULONG a1, BN_ULONG a0, BN_ULONG b1,
+ BN_ULONG b0);
+# endif
+
+/*
+ * Add polynomials a and b and store result in r; r could be a or b, a and b
+ * could be equal; r is the bitwise XOR of a and b.
+ */
+int BN_GF2m_add(BIGNUM *r, const BIGNUM *a, const BIGNUM *b)
+{
+ int i;
+ const BIGNUM *at, *bt;
+
+ bn_check_top(a);
+ bn_check_top(b);
+
+ if (a->top < b->top) {
+ at = b;
+ bt = a;
+ } else {
+ at = a;
+ bt = b;
+ }
+
+ if (bn_wexpand(r, at->top) == NULL)
+ return 0;
+
+ for (i = 0; i < bt->top; i++) {
+ r->d[i] = at->d[i] ^ bt->d[i];
+ }
+ for (; i < at->top; i++) {
+ r->d[i] = at->d[i];
+ }
+
+ r->top = at->top;
+ bn_correct_top(r);
+
+ return 1;
+}
+
+/*-
+ * Some functions allow for representation of the irreducible polynomials
+ * as an int[], say p. The irreducible f(t) is then of the form:
+ * t^p[0] + t^p[1] + ... + t^p[k]
+ * where m = p[0] > p[1] > ... > p[k] = 0.
+ */
+
+/* Performs modular reduction of a and store result in r. r could be a. */
+int BN_GF2m_mod_arr(BIGNUM *r, const BIGNUM *a, const int p[])
+{
+ int j, k;
+ int n, dN, d0, d1;
+ BN_ULONG zz, *z;
+
+ bn_check_top(a);
+
+ if (!p[0]) {
+ /* reduction mod 1 => return 0 */
+ BN_zero(r);
+ return 1;
+ }
+
+ /*
+ * Since the algorithm does reduction in the r value, if a != r, copy the
+ * contents of a into r so we can do reduction in r.
+ */
+ if (a != r) {
+ if (!bn_wexpand(r, a->top))
+ return 0;
+ for (j = 0; j < a->top; j++) {
+ r->d[j] = a->d[j];
+ }
+ r->top = a->top;
+ }
+ z = r->d;
+
+ /* start reduction */
+ dN = p[0] / BN_BITS2;
+ for (j = r->top - 1; j > dN;) {
+ zz = z[j];
+ if (z[j] == 0) {
+ j--;
+ continue;
+ }
+ z[j] = 0;
+
+ for (k = 1; p[k] != 0; k++) {
+ /* reducing component t^p[k] */
+ n = p[0] - p[k];
+ d0 = n % BN_BITS2;
+ d1 = BN_BITS2 - d0;
+ n /= BN_BITS2;
+ z[j - n] ^= (zz >> d0);
+ if (d0)
+ z[j - n - 1] ^= (zz << d1);
+ }
+
+ /* reducing component t^0 */
+ n = dN;
+ d0 = p[0] % BN_BITS2;
+ d1 = BN_BITS2 - d0;
+ z[j - n] ^= (zz >> d0);
+ if (d0)
+ z[j - n - 1] ^= (zz << d1);
+ }
+
+ /* final round of reduction */
+ while (j == dN) {
+
+ d0 = p[0] % BN_BITS2;
+ zz = z[dN] >> d0;
+ if (zz == 0)
+ break;
+ d1 = BN_BITS2 - d0;
+
+ /* clear up the top d1 bits */
+ if (d0)
+ z[dN] = (z[dN] << d1) >> d1;
+ else
+ z[dN] = 0;
+ z[0] ^= zz; /* reduction t^0 component */
+
+ for (k = 1; p[k] != 0; k++) {
+ BN_ULONG tmp_ulong;
+
+ /* reducing component t^p[k] */
+ n = p[k] / BN_BITS2;
+ d0 = p[k] % BN_BITS2;
+ d1 = BN_BITS2 - d0;
+ z[n] ^= (zz << d0);
+ if (d0 && (tmp_ulong = zz >> d1))
+ z[n + 1] ^= tmp_ulong;
+ }
+
+ }
+
+ bn_correct_top(r);
+ return 1;
+}
+
+/*
+ * Performs modular reduction of a by p and store result in r. r could be a.
+ * This function calls down to the BN_GF2m_mod_arr implementation; this wrapper
+ * function is only provided for convenience; for best performance, use the
+ * BN_GF2m_mod_arr function.
+ */
+int BN_GF2m_mod(BIGNUM *r, const BIGNUM *a, const BIGNUM *p)
+{
+ int ret = 0;
+ int arr[6];
+ bn_check_top(a);
+ bn_check_top(p);
+ ret = BN_GF2m_poly2arr(p, arr, OSSL_NELEM(arr));
+ if (!ret || ret > (int)OSSL_NELEM(arr)) {
+ BNerr(BN_F_BN_GF2M_MOD, BN_R_INVALID_LENGTH);
+ return 0;
+ }
+ ret = BN_GF2m_mod_arr(r, a, arr);
+ bn_check_top(r);
+ return ret;
+}
+
+/*
+ * Compute the product of two polynomials a and b, reduce modulo p, and store
+ * the result in r. r could be a or b; a could be b.
+ */
+int BN_GF2m_mod_mul_arr(BIGNUM *r, const BIGNUM *a, const BIGNUM *b,
+ const int p[], BN_CTX *ctx)
+{
+ int zlen, i, j, k, ret = 0;
+ BIGNUM *s;
+ BN_ULONG x1, x0, y1, y0, zz[4];
+
+ bn_check_top(a);
+ bn_check_top(b);
+
+ if (a == b) {
+ return BN_GF2m_mod_sqr_arr(r, a, p, ctx);
+ }
+
+ BN_CTX_start(ctx);
+ if ((s = BN_CTX_get(ctx)) == NULL)
+ goto err;
+
+ zlen = a->top + b->top + 4;
+ if (!bn_wexpand(s, zlen))
+ goto err;
+ s->top = zlen;
+
+ for (i = 0; i < zlen; i++)
+ s->d[i] = 0;
+
+ for (j = 0; j < b->top; j += 2) {
+ y0 = b->d[j];
+ y1 = ((j + 1) == b->top) ? 0 : b->d[j + 1];
+ for (i = 0; i < a->top; i += 2) {
+ x0 = a->d[i];
+ x1 = ((i + 1) == a->top) ? 0 : a->d[i + 1];
+ bn_GF2m_mul_2x2(zz, x1, x0, y1, y0);
+ for (k = 0; k < 4; k++)
+ s->d[i + j + k] ^= zz[k];
+ }
+ }
+
+ bn_correct_top(s);
+ if (BN_GF2m_mod_arr(r, s, p))
+ ret = 1;
+ bn_check_top(r);
+
+ err:
+ BN_CTX_end(ctx);
+ return ret;
+}
+
+/*
+ * Compute the product of two polynomials a and b, reduce modulo p, and store
+ * the result in r. r could be a or b; a could equal b. This function calls
+ * down to the BN_GF2m_mod_mul_arr implementation; this wrapper function is
+ * only provided for convenience; for best performance, use the
+ * BN_GF2m_mod_mul_arr function.
+ */
+int BN_GF2m_mod_mul(BIGNUM *r, const BIGNUM *a, const BIGNUM *b,
+ const BIGNUM *p, BN_CTX *ctx)
+{
+ int ret = 0;
+ const int max = BN_num_bits(p) + 1;
+ int *arr = NULL;
+ bn_check_top(a);
+ bn_check_top(b);
+ bn_check_top(p);
+ if ((arr = OPENSSL_malloc(sizeof(*arr) * max)) == NULL)
+ goto err;
+ ret = BN_GF2m_poly2arr(p, arr, max);
+ if (!ret || ret > max) {
+ BNerr(BN_F_BN_GF2M_MOD_MUL, BN_R_INVALID_LENGTH);
+ goto err;
+ }
+ ret = BN_GF2m_mod_mul_arr(r, a, b, arr, ctx);
+ bn_check_top(r);
+ err:
+ OPENSSL_free(arr);
+ return ret;
+}
+
+/* Square a, reduce the result mod p, and store it in a. r could be a. */
+int BN_GF2m_mod_sqr_arr(BIGNUM *r, const BIGNUM *a, const int p[],
+ BN_CTX *ctx)
+{
+ int i, ret = 0;
+ BIGNUM *s;
+
+ bn_check_top(a);
+ BN_CTX_start(ctx);
+ if ((s = BN_CTX_get(ctx)) == NULL)
+ goto err;
+ if (!bn_wexpand(s, 2 * a->top))
+ goto err;
+
+ for (i = a->top - 1; i >= 0; i--) {
+ s->d[2 * i + 1] = SQR1(a->d[i]);
+ s->d[2 * i] = SQR0(a->d[i]);
+ }
+
+ s->top = 2 * a->top;
+ bn_correct_top(s);
+ if (!BN_GF2m_mod_arr(r, s, p))
+ goto err;
+ bn_check_top(r);
+ ret = 1;
+ err:
+ BN_CTX_end(ctx);
+ return ret;
+}
+
+/*
+ * Square a, reduce the result mod p, and store it in a. r could be a. This
+ * function calls down to the BN_GF2m_mod_sqr_arr implementation; this
+ * wrapper function is only provided for convenience; for best performance,
+ * use the BN_GF2m_mod_sqr_arr function.
+ */
+int BN_GF2m_mod_sqr(BIGNUM *r, const BIGNUM *a, const BIGNUM *p, BN_CTX *ctx)
+{
+ int ret = 0;
+ const int max = BN_num_bits(p) + 1;
+ int *arr = NULL;
+
+ bn_check_top(a);
+ bn_check_top(p);
+ if ((arr = OPENSSL_malloc(sizeof(*arr) * max)) == NULL)
+ goto err;
+ ret = BN_GF2m_poly2arr(p, arr, max);
+ if (!ret || ret > max) {
+ BNerr(BN_F_BN_GF2M_MOD_SQR, BN_R_INVALID_LENGTH);
+ goto err;
+ }
+ ret = BN_GF2m_mod_sqr_arr(r, a, arr, ctx);
+ bn_check_top(r);
+ err:
+ OPENSSL_free(arr);
+ return ret;
+}
+
+/*
+ * Invert a, reduce modulo p, and store the result in r. r could be a. Uses
+ * Modified Almost Inverse Algorithm (Algorithm 10) from Hankerson, D.,
+ * Hernandez, J.L., and Menezes, A. "Software Implementation of Elliptic
+ * Curve Cryptography Over Binary Fields".
+ */
+static int BN_GF2m_mod_inv_vartime(BIGNUM *r, const BIGNUM *a,
+ const BIGNUM *p, BN_CTX *ctx)
+{
+ BIGNUM *b, *c = NULL, *u = NULL, *v = NULL, *tmp;
+ int ret = 0;
+
+ bn_check_top(a);
+ bn_check_top(p);
+
+ BN_CTX_start(ctx);
+
+ b = BN_CTX_get(ctx);
+ c = BN_CTX_get(ctx);
+ u = BN_CTX_get(ctx);
+ v = BN_CTX_get(ctx);
+ if (v == NULL)
+ goto err;
+
+ if (!BN_GF2m_mod(u, a, p))
+ goto err;
+ if (BN_is_zero(u))
+ goto err;
+
+ if (!BN_copy(v, p))
+ goto err;
+# if 0
+ if (!BN_one(b))
+ goto err;
+
+ while (1) {
+ while (!BN_is_odd(u)) {
+ if (BN_is_zero(u))
+ goto err;
+ if (!BN_rshift1(u, u))
+ goto err;
+ if (BN_is_odd(b)) {
+ if (!BN_GF2m_add(b, b, p))
+ goto err;
+ }
+ if (!BN_rshift1(b, b))
+ goto err;
+ }
+
+ if (BN_abs_is_word(u, 1))
+ break;
+
+ if (BN_num_bits(u) < BN_num_bits(v)) {
+ tmp = u;
+ u = v;
+ v = tmp;
+ tmp = b;
+ b = c;
+ c = tmp;
+ }
+
+ if (!BN_GF2m_add(u, u, v))
+ goto err;
+ if (!BN_GF2m_add(b, b, c))
+ goto err;
+ }
+# else
+ {
+ int i;
+ int ubits = BN_num_bits(u);
+ int vbits = BN_num_bits(v); /* v is copy of p */
+ int top = p->top;
+ BN_ULONG *udp, *bdp, *vdp, *cdp;
+
+ if (!bn_wexpand(u, top))
+ goto err;
+ udp = u->d;
+ for (i = u->top; i < top; i++)
+ udp[i] = 0;
+ u->top = top;
+ if (!bn_wexpand(b, top))
+ goto err;
+ bdp = b->d;
+ bdp[0] = 1;
+ for (i = 1; i < top; i++)
+ bdp[i] = 0;
+ b->top = top;
+ if (!bn_wexpand(c, top))
+ goto err;
+ cdp = c->d;
+ for (i = 0; i < top; i++)
+ cdp[i] = 0;
+ c->top = top;
+ vdp = v->d; /* It pays off to "cache" *->d pointers,
+ * because it allows optimizer to be more
+ * aggressive. But we don't have to "cache"
+ * p->d, because *p is declared 'const'... */
+ while (1) {
+ while (ubits && !(udp[0] & 1)) {
+ BN_ULONG u0, u1, b0, b1, mask;
+
+ u0 = udp[0];
+ b0 = bdp[0];
+ mask = (BN_ULONG)0 - (b0 & 1);
+ b0 ^= p->d[0] & mask;
+ for (i = 0; i < top - 1; i++) {
+ u1 = udp[i + 1];
+ udp[i] = ((u0 >> 1) | (u1 << (BN_BITS2 - 1))) & BN_MASK2;
+ u0 = u1;
+ b1 = bdp[i + 1] ^ (p->d[i + 1] & mask);
+ bdp[i] = ((b0 >> 1) | (b1 << (BN_BITS2 - 1))) & BN_MASK2;
+ b0 = b1;
+ }
+ udp[i] = u0 >> 1;
+ bdp[i] = b0 >> 1;
+ ubits--;
+ }
+
+ if (ubits <= BN_BITS2) {
+ if (udp[0] == 0) /* poly was reducible */
+ goto err;
+ if (udp[0] == 1)
+ break;
+ }
+
+ if (ubits < vbits) {
+ i = ubits;
+ ubits = vbits;
+ vbits = i;
+ tmp = u;
+ u = v;
+ v = tmp;
+ tmp = b;
+ b = c;
+ c = tmp;
+ udp = vdp;
+ vdp = v->d;
+ bdp = cdp;
+ cdp = c->d;
+ }
+ for (i = 0; i < top; i++) {
+ udp[i] ^= vdp[i];
+ bdp[i] ^= cdp[i];
+ }
+ if (ubits == vbits) {
+ BN_ULONG ul;
+ int utop = (ubits - 1) / BN_BITS2;
+
+ while ((ul = udp[utop]) == 0 && utop)
+ utop--;
+ ubits = utop * BN_BITS2 + BN_num_bits_word(ul);
+ }
+ }
+ bn_correct_top(b);
+ }
+# endif
+
+ if (!BN_copy(r, b))
+ goto err;
+ bn_check_top(r);
+ ret = 1;
+
+ err:
+# ifdef BN_DEBUG /* BN_CTX_end would complain about the
+ * expanded form */
+ bn_correct_top(c);
+ bn_correct_top(u);
+ bn_correct_top(v);
+# endif
+ BN_CTX_end(ctx);
+ return ret;
+}
+
+/*-
+ * Wrapper for BN_GF2m_mod_inv_vartime that blinds the input before calling.
+ * This is not constant time.
+ * But it does eliminate first order deduction on the input.
+ */
+int BN_GF2m_mod_inv(BIGNUM *r, const BIGNUM *a, const BIGNUM *p, BN_CTX *ctx)
+{
+ BIGNUM *b = NULL;
+ int ret = 0;
+
+ BN_CTX_start(ctx);
+ if ((b = BN_CTX_get(ctx)) == NULL)
+ goto err;
+
+ /* generate blinding value */
+ do {
+ if (!BN_priv_rand(b, BN_num_bits(p) - 1,
+ BN_RAND_TOP_ANY, BN_RAND_BOTTOM_ANY))
+ goto err;
+ } while (BN_is_zero(b));
+
+ /* r := a * b */
+ if (!BN_GF2m_mod_mul(r, a, b, p, ctx))
+ goto err;
+
+ /* r := 1/(a * b) */
+ if (!BN_GF2m_mod_inv_vartime(r, r, p, ctx))
+ goto err;
+
+ /* r := b/(a * b) = 1/a */
+ if (!BN_GF2m_mod_mul(r, r, b, p, ctx))
+ goto err;
+
+ ret = 1;
+
+ err:
+ BN_CTX_end(ctx);
+ return ret;
+}
+
+/*
+ * Invert xx, reduce modulo p, and store the result in r. r could be xx.
+ * This function calls down to the BN_GF2m_mod_inv implementation; this
+ * wrapper function is only provided for convenience; for best performance,
+ * use the BN_GF2m_mod_inv function.
+ */
+int BN_GF2m_mod_inv_arr(BIGNUM *r, const BIGNUM *xx, const int p[],
+ BN_CTX *ctx)
+{
+ BIGNUM *field;
+ int ret = 0;
+
+ bn_check_top(xx);
+ BN_CTX_start(ctx);
+ if ((field = BN_CTX_get(ctx)) == NULL)
+ goto err;
+ if (!BN_GF2m_arr2poly(p, field))
+ goto err;
+
+ ret = BN_GF2m_mod_inv(r, xx, field, ctx);
+ bn_check_top(r);
+
+ err:
+ BN_CTX_end(ctx);
+ return ret;
+}
+
+/*
+ * Divide y by x, reduce modulo p, and store the result in r. r could be x
+ * or y, x could equal y.
+ */
+int BN_GF2m_mod_div(BIGNUM *r, const BIGNUM *y, const BIGNUM *x,
+ const BIGNUM *p, BN_CTX *ctx)
+{
+ BIGNUM *xinv = NULL;
+ int ret = 0;
+
+ bn_check_top(y);
+ bn_check_top(x);
+ bn_check_top(p);
+
+ BN_CTX_start(ctx);
+ xinv = BN_CTX_get(ctx);
+ if (xinv == NULL)
+ goto err;
+
+ if (!BN_GF2m_mod_inv(xinv, x, p, ctx))
+ goto err;
+ if (!BN_GF2m_mod_mul(r, y, xinv, p, ctx))
+ goto err;
+ bn_check_top(r);
+ ret = 1;
+
+ err:
+ BN_CTX_end(ctx);
+ return ret;
+}
+
+/*
+ * Divide yy by xx, reduce modulo p, and store the result in r. r could be xx
+ * * or yy, xx could equal yy. This function calls down to the
+ * BN_GF2m_mod_div implementation; this wrapper function is only provided for
+ * convenience; for best performance, use the BN_GF2m_mod_div function.
+ */
+int BN_GF2m_mod_div_arr(BIGNUM *r, const BIGNUM *yy, const BIGNUM *xx,
+ const int p[], BN_CTX *ctx)
+{
+ BIGNUM *field;
+ int ret = 0;
+
+ bn_check_top(yy);
+ bn_check_top(xx);
+
+ BN_CTX_start(ctx);
+ if ((field = BN_CTX_get(ctx)) == NULL)
+ goto err;
+ if (!BN_GF2m_arr2poly(p, field))
+ goto err;
+
+ ret = BN_GF2m_mod_div(r, yy, xx, field, ctx);
+ bn_check_top(r);
+
+ err:
+ BN_CTX_end(ctx);
+ return ret;
+}
+
+/*
+ * Compute the bth power of a, reduce modulo p, and store the result in r. r
+ * could be a. Uses simple square-and-multiply algorithm A.5.1 from IEEE
+ * P1363.
+ */
+int BN_GF2m_mod_exp_arr(BIGNUM *r, const BIGNUM *a, const BIGNUM *b,
+ const int p[], BN_CTX *ctx)
+{
+ int ret = 0, i, n;
+ BIGNUM *u;
+
+ bn_check_top(a);
+ bn_check_top(b);
+
+ if (BN_is_zero(b))
+ return BN_one(r);
+
+ if (BN_abs_is_word(b, 1))
+ return (BN_copy(r, a) != NULL);
+
+ BN_CTX_start(ctx);
+ if ((u = BN_CTX_get(ctx)) == NULL)
+ goto err;
+
+ if (!BN_GF2m_mod_arr(u, a, p))
+ goto err;
+
+ n = BN_num_bits(b) - 1;
+ for (i = n - 1; i >= 0; i--) {
+ if (!BN_GF2m_mod_sqr_arr(u, u, p, ctx))
+ goto err;
+ if (BN_is_bit_set(b, i)) {
+ if (!BN_GF2m_mod_mul_arr(u, u, a, p, ctx))
+ goto err;
+ }
+ }
+ if (!BN_copy(r, u))
+ goto err;
+ bn_check_top(r);
+ ret = 1;
+ err:
+ BN_CTX_end(ctx);
+ return ret;
+}
+
+/*
+ * Compute the bth power of a, reduce modulo p, and store the result in r. r
+ * could be a. This function calls down to the BN_GF2m_mod_exp_arr
+ * implementation; this wrapper function is only provided for convenience;
+ * for best performance, use the BN_GF2m_mod_exp_arr function.
+ */
+int BN_GF2m_mod_exp(BIGNUM *r, const BIGNUM *a, const BIGNUM *b,
+ const BIGNUM *p, BN_CTX *ctx)
+{
+ int ret = 0;
+ const int max = BN_num_bits(p) + 1;
+ int *arr = NULL;
+ bn_check_top(a);
+ bn_check_top(b);
+ bn_check_top(p);
+ if ((arr = OPENSSL_malloc(sizeof(*arr) * max)) == NULL)
+ goto err;
+ ret = BN_GF2m_poly2arr(p, arr, max);
+ if (!ret || ret > max) {
+ BNerr(BN_F_BN_GF2M_MOD_EXP, BN_R_INVALID_LENGTH);
+ goto err;
+ }
+ ret = BN_GF2m_mod_exp_arr(r, a, b, arr, ctx);
+ bn_check_top(r);
+ err:
+ OPENSSL_free(arr);
+ return ret;
+}
+
+/*
+ * Compute the square root of a, reduce modulo p, and store the result in r.
+ * r could be a. Uses exponentiation as in algorithm A.4.1 from IEEE P1363.
+ */
+int BN_GF2m_mod_sqrt_arr(BIGNUM *r, const BIGNUM *a, const int p[],
+ BN_CTX *ctx)
+{
+ int ret = 0;
+ BIGNUM *u;
+
+ bn_check_top(a);
+
+ if (!p[0]) {
+ /* reduction mod 1 => return 0 */
+ BN_zero(r);
+ return 1;
+ }
+
+ BN_CTX_start(ctx);
+ if ((u = BN_CTX_get(ctx)) == NULL)
+ goto err;
+
+ if (!BN_set_bit(u, p[0] - 1))
+ goto err;
+ ret = BN_GF2m_mod_exp_arr(r, a, u, p, ctx);
+ bn_check_top(r);
+
+ err:
+ BN_CTX_end(ctx);
+ return ret;
+}
+
+/*
+ * Compute the square root of a, reduce modulo p, and store the result in r.
+ * r could be a. This function calls down to the BN_GF2m_mod_sqrt_arr
+ * implementation; this wrapper function is only provided for convenience;
+ * for best performance, use the BN_GF2m_mod_sqrt_arr function.
+ */
+int BN_GF2m_mod_sqrt(BIGNUM *r, const BIGNUM *a, const BIGNUM *p, BN_CTX *ctx)
+{
+ int ret = 0;
+ const int max = BN_num_bits(p) + 1;
+ int *arr = NULL;
+ bn_check_top(a);
+ bn_check_top(p);
+ if ((arr = OPENSSL_malloc(sizeof(*arr) * max)) == NULL)
+ goto err;
+ ret = BN_GF2m_poly2arr(p, arr, max);
+ if (!ret || ret > max) {
+ BNerr(BN_F_BN_GF2M_MOD_SQRT, BN_R_INVALID_LENGTH);
+ goto err;
+ }
+ ret = BN_GF2m_mod_sqrt_arr(r, a, arr, ctx);
+ bn_check_top(r);
+ err:
+ OPENSSL_free(arr);
+ return ret;
+}
+
+/*
+ * Find r such that r^2 + r = a mod p. r could be a. If no r exists returns
+ * 0. Uses algorithms A.4.7 and A.4.6 from IEEE P1363.
+ */
+int BN_GF2m_mod_solve_quad_arr(BIGNUM *r, const BIGNUM *a_, const int p[],
+ BN_CTX *ctx)
+{
+ int ret = 0, count = 0, j;
+ BIGNUM *a, *z, *rho, *w, *w2, *tmp;
+
+ bn_check_top(a_);
+
+ if (!p[0]) {
+ /* reduction mod 1 => return 0 */
+ BN_zero(r);
+ return 1;
+ }
+
+ BN_CTX_start(ctx);
+ a = BN_CTX_get(ctx);
+ z = BN_CTX_get(ctx);
+ w = BN_CTX_get(ctx);
+ if (w == NULL)
+ goto err;
+
+ if (!BN_GF2m_mod_arr(a, a_, p))
+ goto err;
+
+ if (BN_is_zero(a)) {
+ BN_zero(r);
+ ret = 1;
+ goto err;
+ }
+
+ if (p[0] & 0x1) { /* m is odd */
+ /* compute half-trace of a */
+ if (!BN_copy(z, a))
+ goto err;
+ for (j = 1; j <= (p[0] - 1) / 2; j++) {
+ if (!BN_GF2m_mod_sqr_arr(z, z, p, ctx))
+ goto err;
+ if (!BN_GF2m_mod_sqr_arr(z, z, p, ctx))
+ goto err;
+ if (!BN_GF2m_add(z, z, a))
+ goto err;
+ }
+
+ } else { /* m is even */
+
+ rho = BN_CTX_get(ctx);
+ w2 = BN_CTX_get(ctx);
+ tmp = BN_CTX_get(ctx);
+ if (tmp == NULL)
+ goto err;
+ do {
+ if (!BN_priv_rand(rho, p[0], BN_RAND_TOP_ONE, BN_RAND_BOTTOM_ANY))
+ goto err;
+ if (!BN_GF2m_mod_arr(rho, rho, p))
+ goto err;
+ BN_zero(z);
+ if (!BN_copy(w, rho))
+ goto err;
+ for (j = 1; j <= p[0] - 1; j++) {
+ if (!BN_GF2m_mod_sqr_arr(z, z, p, ctx))
+ goto err;
+ if (!BN_GF2m_mod_sqr_arr(w2, w, p, ctx))
+ goto err;
+ if (!BN_GF2m_mod_mul_arr(tmp, w2, a, p, ctx))
+ goto err;
+ if (!BN_GF2m_add(z, z, tmp))
+ goto err;
+ if (!BN_GF2m_add(w, w2, rho))
+ goto err;
+ }
+ count++;
+ } while (BN_is_zero(w) && (count < MAX_ITERATIONS));
+ if (BN_is_zero(w)) {
+ BNerr(BN_F_BN_GF2M_MOD_SOLVE_QUAD_ARR, BN_R_TOO_MANY_ITERATIONS);
+ goto err;
+ }
+ }
+
+ if (!BN_GF2m_mod_sqr_arr(w, z, p, ctx))
+ goto err;
+ if (!BN_GF2m_add(w, z, w))
+ goto err;
+ if (BN_GF2m_cmp(w, a)) {
+ BNerr(BN_F_BN_GF2M_MOD_SOLVE_QUAD_ARR, BN_R_NO_SOLUTION);
+ goto err;
+ }
+
+ if (!BN_copy(r, z))
+ goto err;
+ bn_check_top(r);
+
+ ret = 1;
+
+ err:
+ BN_CTX_end(ctx);
+ return ret;
+}
+
+/*
+ * Find r such that r^2 + r = a mod p. r could be a. If no r exists returns
+ * 0. This function calls down to the BN_GF2m_mod_solve_quad_arr
+ * implementation; this wrapper function is only provided for convenience;
+ * for best performance, use the BN_GF2m_mod_solve_quad_arr function.
+ */
+int BN_GF2m_mod_solve_quad(BIGNUM *r, const BIGNUM *a, const BIGNUM *p,
+ BN_CTX *ctx)
+{
+ int ret = 0;
+ const int max = BN_num_bits(p) + 1;
+ int *arr = NULL;
+ bn_check_top(a);
+ bn_check_top(p);
+ if ((arr = OPENSSL_malloc(sizeof(*arr) * max)) == NULL)
+ goto err;
+ ret = BN_GF2m_poly2arr(p, arr, max);
+ if (!ret || ret > max) {
+ BNerr(BN_F_BN_GF2M_MOD_SOLVE_QUAD, BN_R_INVALID_LENGTH);
+ goto err;
+ }
+ ret = BN_GF2m_mod_solve_quad_arr(r, a, arr, ctx);
+ bn_check_top(r);
+ err:
+ OPENSSL_free(arr);
+ return ret;
+}
+
+/*
+ * Convert the bit-string representation of a polynomial ( \sum_{i=0}^n a_i *
+ * x^i) into an array of integers corresponding to the bits with non-zero
+ * coefficient. Array is terminated with -1. Up to max elements of the array
+ * will be filled. Return value is total number of array elements that would
+ * be filled if array was large enough.
+ */
+int BN_GF2m_poly2arr(const BIGNUM *a, int p[], int max)
+{
+ int i, j, k = 0;
+ BN_ULONG mask;
+
+ if (BN_is_zero(a))
+ return 0;
+
+ for (i = a->top - 1; i >= 0; i--) {
+ if (!a->d[i])
+ /* skip word if a->d[i] == 0 */
+ continue;
+ mask = BN_TBIT;
+ for (j = BN_BITS2 - 1; j >= 0; j--) {
+ if (a->d[i] & mask) {
+ if (k < max)
+ p[k] = BN_BITS2 * i + j;
+ k++;
+ }
+ mask >>= 1;
+ }
+ }
+
+ if (k < max) {
+ p[k] = -1;
+ k++;
+ }
+
+ return k;
+}
+
+/*
+ * Convert the coefficient array representation of a polynomial to a
+ * bit-string. The array must be terminated by -1.
+ */
+int BN_GF2m_arr2poly(const int p[], BIGNUM *a)
+{
+ int i;
+
+ bn_check_top(a);
+ BN_zero(a);
+ for (i = 0; p[i] != -1; i++) {
+ if (BN_set_bit(a, p[i]) == 0)
+ return 0;
+ }
+ bn_check_top(a);
+
+ return 1;
+}
+
+#endif
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_intern.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_intern.c
new file mode 100644
index 0000000..147b4fa
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_intern.c
@@ -0,0 +1,199 @@
+/*
+ * Copyright 2014-2018 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include "internal/cryptlib.h"
+#include "bn_local.h"
+
+/*
+ * Determine the modified width-(w+1) Non-Adjacent Form (wNAF) of 'scalar'.
+ * This is an array r[] of values that are either zero or odd with an
+ * absolute value less than 2^w satisfying
+ * scalar = \sum_j r[j]*2^j
+ * where at most one of any w+1 consecutive digits is non-zero
+ * with the exception that the most significant digit may be only
+ * w-1 zeros away from that next non-zero digit.
+ */
+signed char *bn_compute_wNAF(const BIGNUM *scalar, int w, size_t *ret_len)
+{
+ int window_val;
+ signed char *r = NULL;
+ int sign = 1;
+ int bit, next_bit, mask;
+ size_t len = 0, j;
+
+ if (BN_is_zero(scalar)) {
+ r = OPENSSL_malloc(1);
+ if (r == NULL) {
+ BNerr(BN_F_BN_COMPUTE_WNAF, ERR_R_MALLOC_FAILURE);
+ goto err;
+ }
+ r[0] = 0;
+ *ret_len = 1;
+ return r;
+ }
+
+ if (w <= 0 || w > 7) { /* 'signed char' can represent integers with
+ * absolute values less than 2^7 */
+ BNerr(BN_F_BN_COMPUTE_WNAF, ERR_R_INTERNAL_ERROR);
+ goto err;
+ }
+ bit = 1 << w; /* at most 128 */
+ next_bit = bit << 1; /* at most 256 */
+ mask = next_bit - 1; /* at most 255 */
+
+ if (BN_is_negative(scalar)) {
+ sign = -1;
+ }
+
+ if (scalar->d == NULL || scalar->top == 0) {
+ BNerr(BN_F_BN_COMPUTE_WNAF, ERR_R_INTERNAL_ERROR);
+ goto err;
+ }
+
+ len = BN_num_bits(scalar);
+ r = OPENSSL_malloc(len + 1); /*
+ * Modified wNAF may be one digit longer than binary representation
+ * (*ret_len will be set to the actual length, i.e. at most
+ * BN_num_bits(scalar) + 1)
+ */
+ if (r == NULL) {
+ BNerr(BN_F_BN_COMPUTE_WNAF, ERR_R_MALLOC_FAILURE);
+ goto err;
+ }
+ window_val = scalar->d[0] & mask;
+ j = 0;
+ while ((window_val != 0) || (j + w + 1 < len)) { /* if j+w+1 >= len,
+ * window_val will not
+ * increase */
+ int digit = 0;
+
+ /* 0 <= window_val <= 2^(w+1) */
+
+ if (window_val & 1) {
+ /* 0 < window_val < 2^(w+1) */
+
+ if (window_val & bit) {
+ digit = window_val - next_bit; /* -2^w < digit < 0 */
+
+#if 1 /* modified wNAF */
+ if (j + w + 1 >= len) {
+ /*
+ * Special case for generating modified wNAFs:
+ * no new bits will be added into window_val,
+ * so using a positive digit here will decrease
+ * the total length of the representation
+ */
+
+ digit = window_val & (mask >> 1); /* 0 < digit < 2^w */
+ }
+#endif
+ } else {
+ digit = window_val; /* 0 < digit < 2^w */
+ }
+
+ if (digit <= -bit || digit >= bit || !(digit & 1)) {
+ BNerr(BN_F_BN_COMPUTE_WNAF, ERR_R_INTERNAL_ERROR);
+ goto err;
+ }
+
+ window_val -= digit;
+
+ /*
+ * now window_val is 0 or 2^(w+1) in standard wNAF generation;
+ * for modified window NAFs, it may also be 2^w
+ */
+ if (window_val != 0 && window_val != next_bit
+ && window_val != bit) {
+ BNerr(BN_F_BN_COMPUTE_WNAF, ERR_R_INTERNAL_ERROR);
+ goto err;
+ }
+ }
+
+ r[j++] = sign * digit;
+
+ window_val >>= 1;
+ window_val += bit * BN_is_bit_set(scalar, j + w);
+
+ if (window_val > next_bit) {
+ BNerr(BN_F_BN_COMPUTE_WNAF, ERR_R_INTERNAL_ERROR);
+ goto err;
+ }
+ }
+
+ if (j > len + 1) {
+ BNerr(BN_F_BN_COMPUTE_WNAF, ERR_R_INTERNAL_ERROR);
+ goto err;
+ }
+ *ret_len = j;
+ return r;
+
+ err:
+ OPENSSL_free(r);
+ return NULL;
+}
+
+int bn_get_top(const BIGNUM *a)
+{
+ return a->top;
+}
+
+int bn_get_dmax(const BIGNUM *a)
+{
+ return a->dmax;
+}
+
+void bn_set_all_zero(BIGNUM *a)
+{
+ int i;
+
+ for (i = a->top; i < a->dmax; i++)
+ a->d[i] = 0;
+}
+
+int bn_copy_words(BN_ULONG *out, const BIGNUM *in, int size)
+{
+ if (in->top > size)
+ return 0;
+
+ memset(out, 0, sizeof(*out) * size);
+ if (in->d != NULL)
+ memcpy(out, in->d, sizeof(*out) * in->top);
+ return 1;
+}
+
+BN_ULONG *bn_get_words(const BIGNUM *a)
+{
+ return a->d;
+}
+
+void bn_set_static_words(BIGNUM *a, const BN_ULONG *words, int size)
+{
+ /*
+ * |const| qualifier omission is compensated by BN_FLG_STATIC_DATA
+ * flag, which effectively means "read-only data".
+ */
+ a->d = (BN_ULONG *)words;
+ a->dmax = a->top = size;
+ a->neg = 0;
+ a->flags |= BN_FLG_STATIC_DATA;
+ bn_correct_top(a);
+}
+
+int bn_set_words(BIGNUM *a, const BN_ULONG *words, int num_words)
+{
+ if (bn_wexpand(a, num_words) == NULL) {
+ BNerr(BN_F_BN_SET_WORDS, ERR_R_MALLOC_FAILURE);
+ return 0;
+ }
+
+ memcpy(a->d, words, sizeof(BN_ULONG) * num_words);
+ a->top = num_words;
+ bn_correct_top(a);
+ return 1;
+}
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_kron.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_kron.c
new file mode 100644
index 0000000..c1e09d2
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_kron.c
@@ -0,0 +1,140 @@
+/*
+ * Copyright 2000-2016 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include "internal/cryptlib.h"
+#include "bn_local.h"
+
+/* least significant word */
+#define BN_lsw(n) (((n)->top == 0) ? (BN_ULONG) 0 : (n)->d[0])
+
+/* Returns -2 for errors because both -1 and 0 are valid results. */
+int BN_kronecker(const BIGNUM *a, const BIGNUM *b, BN_CTX *ctx)
+{
+ int i;
+ int ret = -2; /* avoid 'uninitialized' warning */
+ int err = 0;
+ BIGNUM *A, *B, *tmp;
+ /*-
+ * In 'tab', only odd-indexed entries are relevant:
+ * For any odd BIGNUM n,
+ * tab[BN_lsw(n) & 7]
+ * is $(-1)^{(n^2-1)/8}$ (using TeX notation).
+ * Note that the sign of n does not matter.
+ */
+ static const int tab[8] = { 0, 1, 0, -1, 0, -1, 0, 1 };
+
+ bn_check_top(a);
+ bn_check_top(b);
+
+ BN_CTX_start(ctx);
+ A = BN_CTX_get(ctx);
+ B = BN_CTX_get(ctx);
+ if (B == NULL)
+ goto end;
+
+ err = !BN_copy(A, a);
+ if (err)
+ goto end;
+ err = !BN_copy(B, b);
+ if (err)
+ goto end;
+
+ /*
+ * Kronecker symbol, implemented according to Henri Cohen,
+ * "A Course in Computational Algebraic Number Theory"
+ * (algorithm 1.4.10).
+ */
+
+ /* Cohen's step 1: */
+
+ if (BN_is_zero(B)) {
+ ret = BN_abs_is_word(A, 1);
+ goto end;
+ }
+
+ /* Cohen's step 2: */
+
+ if (!BN_is_odd(A) && !BN_is_odd(B)) {
+ ret = 0;
+ goto end;
+ }
+
+ /* now B is non-zero */
+ i = 0;
+ while (!BN_is_bit_set(B, i))
+ i++;
+ err = !BN_rshift(B, B, i);
+ if (err)
+ goto end;
+ if (i & 1) {
+ /* i is odd */
+ /* (thus B was even, thus A must be odd!) */
+
+ /* set 'ret' to $(-1)^{(A^2-1)/8}$ */
+ ret = tab[BN_lsw(A) & 7];
+ } else {
+ /* i is even */
+ ret = 1;
+ }
+
+ if (B->neg) {
+ B->neg = 0;
+ if (A->neg)
+ ret = -ret;
+ }
+
+ /*
+ * now B is positive and odd, so what remains to be done is to compute
+ * the Jacobi symbol (A/B) and multiply it by 'ret'
+ */
+
+ while (1) {
+ /* Cohen's step 3: */
+
+ /* B is positive and odd */
+
+ if (BN_is_zero(A)) {
+ ret = BN_is_one(B) ? ret : 0;
+ goto end;
+ }
+
+ /* now A is non-zero */
+ i = 0;
+ while (!BN_is_bit_set(A, i))
+ i++;
+ err = !BN_rshift(A, A, i);
+ if (err)
+ goto end;
+ if (i & 1) {
+ /* i is odd */
+ /* multiply 'ret' by $(-1)^{(B^2-1)/8}$ */
+ ret = ret * tab[BN_lsw(B) & 7];
+ }
+
+ /* Cohen's step 4: */
+ /* multiply 'ret' by $(-1)^{(A-1)(B-1)/4}$ */
+ if ((A->neg ? ~BN_lsw(A) : BN_lsw(A)) & BN_lsw(B) & 2)
+ ret = -ret;
+
+ /* (A, B) := (B mod |A|, |A|) */
+ err = !BN_nnmod(B, B, A, ctx);
+ if (err)
+ goto end;
+ tmp = A;
+ A = B;
+ B = tmp;
+ tmp->neg = 0;
+ }
+ end:
+ BN_CTX_end(ctx);
+ if (err)
+ return -2;
+ else
+ return ret;
+}
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_lib.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_lib.c
new file mode 100644
index 0000000..eb4a318
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_lib.c
@@ -0,0 +1,1021 @@
+/*
+ * Copyright 1995-2020 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include <assert.h>
+#include <limits.h>
+#include "internal/cryptlib.h"
+#include "bn_local.h"
+#include <openssl/opensslconf.h>
+#include "internal/constant_time.h"
+
+/* This stuff appears to be completely unused, so is deprecated */
+#if OPENSSL_API_COMPAT < 0x00908000L
+/*-
+ * For a 32 bit machine
+ * 2 - 4 == 128
+ * 3 - 8 == 256
+ * 4 - 16 == 512
+ * 5 - 32 == 1024
+ * 6 - 64 == 2048
+ * 7 - 128 == 4096
+ * 8 - 256 == 8192
+ */
+static int bn_limit_bits = 0;
+static int bn_limit_num = 8; /* (1<<bn_limit_bits) */
+static int bn_limit_bits_low = 0;
+static int bn_limit_num_low = 8; /* (1<<bn_limit_bits_low) */
+static int bn_limit_bits_high = 0;
+static int bn_limit_num_high = 8; /* (1<<bn_limit_bits_high) */
+static int bn_limit_bits_mont = 0;
+static int bn_limit_num_mont = 8; /* (1<<bn_limit_bits_mont) */
+
+void BN_set_params(int mult, int high, int low, int mont)
+{
+ if (mult >= 0) {
+ if (mult > (int)(sizeof(int) * 8) - 1)
+ mult = sizeof(int) * 8 - 1;
+ bn_limit_bits = mult;
+ bn_limit_num = 1 << mult;
+ }
+ if (high >= 0) {
+ if (high > (int)(sizeof(int) * 8) - 1)
+ high = sizeof(int) * 8 - 1;
+ bn_limit_bits_high = high;
+ bn_limit_num_high = 1 << high;
+ }
+ if (low >= 0) {
+ if (low > (int)(sizeof(int) * 8) - 1)
+ low = sizeof(int) * 8 - 1;
+ bn_limit_bits_low = low;
+ bn_limit_num_low = 1 << low;
+ }
+ if (mont >= 0) {
+ if (mont > (int)(sizeof(int) * 8) - 1)
+ mont = sizeof(int) * 8 - 1;
+ bn_limit_bits_mont = mont;
+ bn_limit_num_mont = 1 << mont;
+ }
+}
+
+int BN_get_params(int which)
+{
+ if (which == 0)
+ return bn_limit_bits;
+ else if (which == 1)
+ return bn_limit_bits_high;
+ else if (which == 2)
+ return bn_limit_bits_low;
+ else if (which == 3)
+ return bn_limit_bits_mont;
+ else
+ return 0;
+}
+#endif
+
+const BIGNUM *BN_value_one(void)
+{
+ static const BN_ULONG data_one = 1L;
+ static const BIGNUM const_one =
+ { (BN_ULONG *)&data_one, 1, 1, 0, BN_FLG_STATIC_DATA };
+
+ return &const_one;
+}
+
+/*
+ * Old Visual Studio ARM compiler miscompiles BN_num_bits_word()
+ * https://mta.openssl.org/pipermail/openssl-users/2018-August/008465.html
+ */
+#if defined(_MSC_VER) && defined(_ARM_) && defined(_WIN32_WCE) \
+ && _MSC_VER>=1400 && _MSC_VER<1501
+# define MS_BROKEN_BN_num_bits_word
+# pragma optimize("", off)
+#endif
+int BN_num_bits_word(BN_ULONG l)
+{
+ BN_ULONG x, mask;
+ int bits = (l != 0);
+
+#if BN_BITS2 > 32
+ x = l >> 32;
+ mask = (0 - x) & BN_MASK2;
+ mask = (0 - (mask >> (BN_BITS2 - 1)));
+ bits += 32 & mask;
+ l ^= (x ^ l) & mask;
+#endif
+
+ x = l >> 16;
+ mask = (0 - x) & BN_MASK2;
+ mask = (0 - (mask >> (BN_BITS2 - 1)));
+ bits += 16 & mask;
+ l ^= (x ^ l) & mask;
+
+ x = l >> 8;
+ mask = (0 - x) & BN_MASK2;
+ mask = (0 - (mask >> (BN_BITS2 - 1)));
+ bits += 8 & mask;
+ l ^= (x ^ l) & mask;
+
+ x = l >> 4;
+ mask = (0 - x) & BN_MASK2;
+ mask = (0 - (mask >> (BN_BITS2 - 1)));
+ bits += 4 & mask;
+ l ^= (x ^ l) & mask;
+
+ x = l >> 2;
+ mask = (0 - x) & BN_MASK2;
+ mask = (0 - (mask >> (BN_BITS2 - 1)));
+ bits += 2 & mask;
+ l ^= (x ^ l) & mask;
+
+ x = l >> 1;
+ mask = (0 - x) & BN_MASK2;
+ mask = (0 - (mask >> (BN_BITS2 - 1)));
+ bits += 1 & mask;
+
+ return bits;
+}
+#ifdef MS_BROKEN_BN_num_bits_word
+# pragma optimize("", on)
+#endif
+
+/*
+ * This function still leaks `a->dmax`: it's caller's responsibility to
+ * expand the input `a` in advance to a public length.
+ */
+static ossl_inline
+int bn_num_bits_consttime(const BIGNUM *a)
+{
+ int j, ret;
+ unsigned int mask, past_i;
+ int i = a->top - 1;
+ bn_check_top(a);
+
+ for (j = 0, past_i = 0, ret = 0; j < a->dmax; j++) {
+ mask = constant_time_eq_int(i, j); /* 0xff..ff if i==j, 0x0 otherwise */
+
+ ret += BN_BITS2 & (~mask & ~past_i);
+ ret += BN_num_bits_word(a->d[j]) & mask;
+
+ past_i |= mask; /* past_i will become 0xff..ff after i==j */
+ }
+
+ /*
+ * if BN_is_zero(a) => i is -1 and ret contains garbage, so we mask the
+ * final result.
+ */
+ mask = ~(constant_time_eq_int(i, ((int)-1)));
+
+ return ret & mask;
+}
+
+int BN_num_bits(const BIGNUM *a)
+{
+ int i = a->top - 1;
+ bn_check_top(a);
+
+ if (a->flags & BN_FLG_CONSTTIME) {
+ /*
+ * We assume that BIGNUMs flagged as CONSTTIME have also been expanded
+ * so that a->dmax is not leaking secret information.
+ *
+ * In other words, it's the caller's responsibility to ensure `a` has
+ * been preallocated in advance to a public length if we hit this
+ * branch.
+ *
+ */
+ return bn_num_bits_consttime(a);
+ }
+
+ if (BN_is_zero(a))
+ return 0;
+
+ return ((i * BN_BITS2) + BN_num_bits_word(a->d[i]));
+}
+
+static void bn_free_d(BIGNUM *a, int clear)
+{
+ if (BN_get_flags(a, BN_FLG_SECURE))
+ OPENSSL_secure_clear_free(a->d, a->dmax * sizeof(a->d[0]));
+ else if (clear != 0)
+ OPENSSL_clear_free(a->d, a->dmax * sizeof(a->d[0]));
+ else
+ OPENSSL_free(a->d);
+}
+
+
+void BN_clear_free(BIGNUM *a)
+{
+ if (a == NULL)
+ return;
+ if (a->d != NULL && !BN_get_flags(a, BN_FLG_STATIC_DATA))
+ bn_free_d(a, 1);
+ if (BN_get_flags(a, BN_FLG_MALLOCED)) {
+ OPENSSL_cleanse(a, sizeof(*a));
+ OPENSSL_free(a);
+ }
+}
+
+void BN_free(BIGNUM *a)
+{
+ if (a == NULL)
+ return;
+ if (!BN_get_flags(a, BN_FLG_STATIC_DATA))
+ bn_free_d(a, 0);
+ if (a->flags & BN_FLG_MALLOCED)
+ OPENSSL_free(a);
+}
+
+void bn_init(BIGNUM *a)
+{
+ static BIGNUM nilbn;
+
+ *a = nilbn;
+ bn_check_top(a);
+}
+
+BIGNUM *BN_new(void)
+{
+ BIGNUM *ret;
+
+ if ((ret = OPENSSL_zalloc(sizeof(*ret))) == NULL) {
+ BNerr(BN_F_BN_NEW, ERR_R_MALLOC_FAILURE);
+ return NULL;
+ }
+ ret->flags = BN_FLG_MALLOCED;
+ bn_check_top(ret);
+ return ret;
+}
+
+ BIGNUM *BN_secure_new(void)
+ {
+ BIGNUM *ret = BN_new();
+ if (ret != NULL)
+ ret->flags |= BN_FLG_SECURE;
+ return ret;
+ }
+
+/* This is used by bn_expand2() */
+/* The caller MUST check that words > b->dmax before calling this */
+static BN_ULONG *bn_expand_internal(const BIGNUM *b, int words)
+{
+ BN_ULONG *a = NULL;
+
+ if (words > (INT_MAX / (4 * BN_BITS2))) {
+ BNerr(BN_F_BN_EXPAND_INTERNAL, BN_R_BIGNUM_TOO_LONG);
+ return NULL;
+ }
+ if (BN_get_flags(b, BN_FLG_STATIC_DATA)) {
+ BNerr(BN_F_BN_EXPAND_INTERNAL, BN_R_EXPAND_ON_STATIC_BIGNUM_DATA);
+ return NULL;
+ }
+ if (BN_get_flags(b, BN_FLG_SECURE))
+ a = OPENSSL_secure_zalloc(words * sizeof(*a));
+ else
+ a = OPENSSL_zalloc(words * sizeof(*a));
+ if (a == NULL) {
+ BNerr(BN_F_BN_EXPAND_INTERNAL, ERR_R_MALLOC_FAILURE);
+ return NULL;
+ }
+
+ assert(b->top <= words);
+ if (b->top > 0)
+ memcpy(a, b->d, sizeof(*a) * b->top);
+
+ return a;
+}
+
+/*
+ * This is an internal function that should not be used in applications. It
+ * ensures that 'b' has enough room for a 'words' word number and initialises
+ * any unused part of b->d with leading zeros. It is mostly used by the
+ * various BIGNUM routines. If there is an error, NULL is returned. If not,
+ * 'b' is returned.
+ */
+
+BIGNUM *bn_expand2(BIGNUM *b, int words)
+{
+ if (words > b->dmax) {
+ BN_ULONG *a = bn_expand_internal(b, words);
+ if (!a)
+ return NULL;
+ if (b->d != NULL)
+ bn_free_d(b, 1);
+ b->d = a;
+ b->dmax = words;
+ }
+
+ return b;
+}
+
+BIGNUM *BN_dup(const BIGNUM *a)
+{
+ BIGNUM *t;
+
+ if (a == NULL)
+ return NULL;
+ bn_check_top(a);
+
+ t = BN_get_flags(a, BN_FLG_SECURE) ? BN_secure_new() : BN_new();
+ if (t == NULL)
+ return NULL;
+ if (!BN_copy(t, a)) {
+ BN_free(t);
+ return NULL;
+ }
+ bn_check_top(t);
+ return t;
+}
+
+BIGNUM *BN_copy(BIGNUM *a, const BIGNUM *b)
+{
+ int bn_words;
+
+ bn_check_top(b);
+
+ bn_words = BN_get_flags(b, BN_FLG_CONSTTIME) ? b->dmax : b->top;
+
+ if (a == b)
+ return a;
+ if (bn_wexpand(a, bn_words) == NULL)
+ return NULL;
+
+ if (b->top > 0)
+ memcpy(a->d, b->d, sizeof(b->d[0]) * bn_words);
+
+ a->neg = b->neg;
+ a->top = b->top;
+ a->flags |= b->flags & BN_FLG_FIXED_TOP;
+ bn_check_top(a);
+ return a;
+}
+
+#define FLAGS_DATA(flags) ((flags) & (BN_FLG_STATIC_DATA \
+ | BN_FLG_CONSTTIME \
+ | BN_FLG_SECURE \
+ | BN_FLG_FIXED_TOP))
+#define FLAGS_STRUCT(flags) ((flags) & (BN_FLG_MALLOCED))
+
+void BN_swap(BIGNUM *a, BIGNUM *b)
+{
+ int flags_old_a, flags_old_b;
+ BN_ULONG *tmp_d;
+ int tmp_top, tmp_dmax, tmp_neg;
+
+ bn_check_top(a);
+ bn_check_top(b);
+
+ flags_old_a = a->flags;
+ flags_old_b = b->flags;
+
+ tmp_d = a->d;
+ tmp_top = a->top;
+ tmp_dmax = a->dmax;
+ tmp_neg = a->neg;
+
+ a->d = b->d;
+ a->top = b->top;
+ a->dmax = b->dmax;
+ a->neg = b->neg;
+
+ b->d = tmp_d;
+ b->top = tmp_top;
+ b->dmax = tmp_dmax;
+ b->neg = tmp_neg;
+
+ a->flags = FLAGS_STRUCT(flags_old_a) | FLAGS_DATA(flags_old_b);
+ b->flags = FLAGS_STRUCT(flags_old_b) | FLAGS_DATA(flags_old_a);
+ bn_check_top(a);
+ bn_check_top(b);
+}
+
+void BN_clear(BIGNUM *a)
+{
+ if (a == NULL)
+ return;
+ bn_check_top(a);
+ if (a->d != NULL)
+ OPENSSL_cleanse(a->d, sizeof(*a->d) * a->dmax);
+ a->neg = 0;
+ a->top = 0;
+ a->flags &= ~BN_FLG_FIXED_TOP;
+}
+
+BN_ULONG BN_get_word(const BIGNUM *a)
+{
+ if (a->top > 1)
+ return BN_MASK2;
+ else if (a->top == 1)
+ return a->d[0];
+ /* a->top == 0 */
+ return 0;
+}
+
+int BN_set_word(BIGNUM *a, BN_ULONG w)
+{
+ bn_check_top(a);
+ if (bn_expand(a, (int)sizeof(BN_ULONG) * 8) == NULL)
+ return 0;
+ a->neg = 0;
+ a->d[0] = w;
+ a->top = (w ? 1 : 0);
+ a->flags &= ~BN_FLG_FIXED_TOP;
+ bn_check_top(a);
+ return 1;
+}
+
+BIGNUM *BN_bin2bn(const unsigned char *s, int len, BIGNUM *ret)
+{
+ unsigned int i, m;
+ unsigned int n;
+ BN_ULONG l;
+ BIGNUM *bn = NULL;
+
+ if (ret == NULL)
+ ret = bn = BN_new();
+ if (ret == NULL)
+ return NULL;
+ bn_check_top(ret);
+ /* Skip leading zero's. */
+ for ( ; len > 0 && *s == 0; s++, len--)
+ continue;
+ n = len;
+ if (n == 0) {
+ ret->top = 0;
+ return ret;
+ }
+ i = ((n - 1) / BN_BYTES) + 1;
+ m = ((n - 1) % (BN_BYTES));
+ if (bn_wexpand(ret, (int)i) == NULL) {
+ BN_free(bn);
+ return NULL;
+ }
+ ret->top = i;
+ ret->neg = 0;
+ l = 0;
+ while (n--) {
+ l = (l << 8L) | *(s++);
+ if (m-- == 0) {
+ ret->d[--i] = l;
+ l = 0;
+ m = BN_BYTES - 1;
+ }
+ }
+ /*
+ * need to call this due to clear byte at top if avoiding having the top
+ * bit set (-ve number)
+ */
+ bn_correct_top(ret);
+ return ret;
+}
+
+typedef enum {big, little} endianess_t;
+
+/* ignore negative */
+static
+int bn2binpad(const BIGNUM *a, unsigned char *to, int tolen, endianess_t endianess)
+{
+ int n;
+ size_t i, lasti, j, atop, mask;
+ BN_ULONG l;
+
+ /*
+ * In case |a| is fixed-top, BN_num_bytes can return bogus length,
+ * but it's assumed that fixed-top inputs ought to be "nominated"
+ * even for padded output, so it works out...
+ */
+ n = BN_num_bytes(a);
+ if (tolen == -1) {
+ tolen = n;
+ } else if (tolen < n) { /* uncommon/unlike case */
+ BIGNUM temp = *a;
+
+ bn_correct_top(&temp);
+ n = BN_num_bytes(&temp);
+ if (tolen < n)
+ return -1;
+ }
+
+ /* Swipe through whole available data and don't give away padded zero. */
+ atop = a->dmax * BN_BYTES;
+ if (atop == 0) {
+ OPENSSL_cleanse(to, tolen);
+ return tolen;
+ }
+
+ lasti = atop - 1;
+ atop = a->top * BN_BYTES;
+ if (endianess == big)
+ to += tolen; /* start from the end of the buffer */
+ for (i = 0, j = 0; j < (size_t)tolen; j++) {
+ unsigned char val;
+ l = a->d[i / BN_BYTES];
+ mask = 0 - ((j - atop) >> (8 * sizeof(i) - 1));
+ val = (unsigned char)(l >> (8 * (i % BN_BYTES)) & mask);
+ if (endianess == big)
+ *--to = val;
+ else
+ *to++ = val;
+ i += (i - lasti) >> (8 * sizeof(i) - 1); /* stay on last limb */
+ }
+
+ return tolen;
+}
+
+int BN_bn2binpad(const BIGNUM *a, unsigned char *to, int tolen)
+{
+ if (tolen < 0)
+ return -1;
+ return bn2binpad(a, to, tolen, big);
+}
+
+int BN_bn2bin(const BIGNUM *a, unsigned char *to)
+{
+ return bn2binpad(a, to, -1, big);
+}
+
+BIGNUM *BN_lebin2bn(const unsigned char *s, int len, BIGNUM *ret)
+{
+ unsigned int i, m;
+ unsigned int n;
+ BN_ULONG l;
+ BIGNUM *bn = NULL;
+
+ if (ret == NULL)
+ ret = bn = BN_new();
+ if (ret == NULL)
+ return NULL;
+ bn_check_top(ret);
+ s += len;
+ /* Skip trailing zeroes. */
+ for ( ; len > 0 && s[-1] == 0; s--, len--)
+ continue;
+ n = len;
+ if (n == 0) {
+ ret->top = 0;
+ return ret;
+ }
+ i = ((n - 1) / BN_BYTES) + 1;
+ m = ((n - 1) % (BN_BYTES));
+ if (bn_wexpand(ret, (int)i) == NULL) {
+ BN_free(bn);
+ return NULL;
+ }
+ ret->top = i;
+ ret->neg = 0;
+ l = 0;
+ while (n--) {
+ s--;
+ l = (l << 8L) | *s;
+ if (m-- == 0) {
+ ret->d[--i] = l;
+ l = 0;
+ m = BN_BYTES - 1;
+ }
+ }
+ /*
+ * need to call this due to clear byte at top if avoiding having the top
+ * bit set (-ve number)
+ */
+ bn_correct_top(ret);
+ return ret;
+}
+
+int BN_bn2lebinpad(const BIGNUM *a, unsigned char *to, int tolen)
+{
+ if (tolen < 0)
+ return -1;
+ return bn2binpad(a, to, tolen, little);
+}
+
+int BN_ucmp(const BIGNUM *a, const BIGNUM *b)
+{
+ int i;
+ BN_ULONG t1, t2, *ap, *bp;
+
+ bn_check_top(a);
+ bn_check_top(b);
+
+ i = a->top - b->top;
+ if (i != 0)
+ return i;
+ ap = a->d;
+ bp = b->d;
+ for (i = a->top - 1; i >= 0; i--) {
+ t1 = ap[i];
+ t2 = bp[i];
+ if (t1 != t2)
+ return ((t1 > t2) ? 1 : -1);
+ }
+ return 0;
+}
+
+int BN_cmp(const BIGNUM *a, const BIGNUM *b)
+{
+ int i;
+ int gt, lt;
+ BN_ULONG t1, t2;
+
+ if ((a == NULL) || (b == NULL)) {
+ if (a != NULL)
+ return -1;
+ else if (b != NULL)
+ return 1;
+ else
+ return 0;
+ }
+
+ bn_check_top(a);
+ bn_check_top(b);
+
+ if (a->neg != b->neg) {
+ if (a->neg)
+ return -1;
+ else
+ return 1;
+ }
+ if (a->neg == 0) {
+ gt = 1;
+ lt = -1;
+ } else {
+ gt = -1;
+ lt = 1;
+ }
+
+ if (a->top > b->top)
+ return gt;
+ if (a->top < b->top)
+ return lt;
+ for (i = a->top - 1; i >= 0; i--) {
+ t1 = a->d[i];
+ t2 = b->d[i];
+ if (t1 > t2)
+ return gt;
+ if (t1 < t2)
+ return lt;
+ }
+ return 0;
+}
+
+int BN_set_bit(BIGNUM *a, int n)
+{
+ int i, j, k;
+
+ if (n < 0)
+ return 0;
+
+ i = n / BN_BITS2;
+ j = n % BN_BITS2;
+ if (a->top <= i) {
+ if (bn_wexpand(a, i + 1) == NULL)
+ return 0;
+ for (k = a->top; k < i + 1; k++)
+ a->d[k] = 0;
+ a->top = i + 1;
+ a->flags &= ~BN_FLG_FIXED_TOP;
+ }
+
+ a->d[i] |= (((BN_ULONG)1) << j);
+ bn_check_top(a);
+ return 1;
+}
+
+int BN_clear_bit(BIGNUM *a, int n)
+{
+ int i, j;
+
+ bn_check_top(a);
+ if (n < 0)
+ return 0;
+
+ i = n / BN_BITS2;
+ j = n % BN_BITS2;
+ if (a->top <= i)
+ return 0;
+
+ a->d[i] &= (~(((BN_ULONG)1) << j));
+ bn_correct_top(a);
+ return 1;
+}
+
+int BN_is_bit_set(const BIGNUM *a, int n)
+{
+ int i, j;
+
+ bn_check_top(a);
+ if (n < 0)
+ return 0;
+ i = n / BN_BITS2;
+ j = n % BN_BITS2;
+ if (a->top <= i)
+ return 0;
+ return (int)(((a->d[i]) >> j) & ((BN_ULONG)1));
+}
+
+int BN_mask_bits(BIGNUM *a, int n)
+{
+ int b, w;
+
+ bn_check_top(a);
+ if (n < 0)
+ return 0;
+
+ w = n / BN_BITS2;
+ b = n % BN_BITS2;
+ if (w >= a->top)
+ return 0;
+ if (b == 0)
+ a->top = w;
+ else {
+ a->top = w + 1;
+ a->d[w] &= ~(BN_MASK2 << b);
+ }
+ bn_correct_top(a);
+ return 1;
+}
+
+void BN_set_negative(BIGNUM *a, int b)
+{
+ if (b && !BN_is_zero(a))
+ a->neg = 1;
+ else
+ a->neg = 0;
+}
+
+int bn_cmp_words(const BN_ULONG *a, const BN_ULONG *b, int n)
+{
+ int i;
+ BN_ULONG aa, bb;
+
+ if (n == 0)
+ return 0;
+
+ aa = a[n - 1];
+ bb = b[n - 1];
+ if (aa != bb)
+ return ((aa > bb) ? 1 : -1);
+ for (i = n - 2; i >= 0; i--) {
+ aa = a[i];
+ bb = b[i];
+ if (aa != bb)
+ return ((aa > bb) ? 1 : -1);
+ }
+ return 0;
+}
+
+/*
+ * Here follows a specialised variants of bn_cmp_words(). It has the
+ * capability of performing the operation on arrays of different sizes. The
+ * sizes of those arrays is expressed through cl, which is the common length
+ * ( basically, min(len(a),len(b)) ), and dl, which is the delta between the
+ * two lengths, calculated as len(a)-len(b). All lengths are the number of
+ * BN_ULONGs...
+ */
+
+int bn_cmp_part_words(const BN_ULONG *a, const BN_ULONG *b, int cl, int dl)
+{
+ int n, i;
+ n = cl - 1;
+
+ if (dl < 0) {
+ for (i = dl; i < 0; i++) {
+ if (b[n - i] != 0)
+ return -1; /* a < b */
+ }
+ }
+ if (dl > 0) {
+ for (i = dl; i > 0; i--) {
+ if (a[n + i] != 0)
+ return 1; /* a > b */
+ }
+ }
+ return bn_cmp_words(a, b, cl);
+}
+
+/*-
+ * Constant-time conditional swap of a and b.
+ * a and b are swapped if condition is not 0.
+ * nwords is the number of words to swap.
+ * Assumes that at least nwords are allocated in both a and b.
+ * Assumes that no more than nwords are used by either a or b.
+ */
+void BN_consttime_swap(BN_ULONG condition, BIGNUM *a, BIGNUM *b, int nwords)
+{
+ BN_ULONG t;
+ int i;
+
+ if (a == b)
+ return;
+
+ bn_wcheck_size(a, nwords);
+ bn_wcheck_size(b, nwords);
+
+ condition = ((~condition & ((condition - 1))) >> (BN_BITS2 - 1)) - 1;
+
+ t = (a->top ^ b->top) & condition;
+ a->top ^= t;
+ b->top ^= t;
+
+ t = (a->neg ^ b->neg) & condition;
+ a->neg ^= t;
+ b->neg ^= t;
+
+ /*-
+ * BN_FLG_STATIC_DATA: indicates that data may not be written to. Intention
+ * is actually to treat it as it's read-only data, and some (if not most)
+ * of it does reside in read-only segment. In other words observation of
+ * BN_FLG_STATIC_DATA in BN_consttime_swap should be treated as fatal
+ * condition. It would either cause SEGV or effectively cause data
+ * corruption.
+ *
+ * BN_FLG_MALLOCED: refers to BN structure itself, and hence must be
+ * preserved.
+ *
+ * BN_FLG_SECURE: must be preserved, because it determines how x->d was
+ * allocated and hence how to free it.
+ *
+ * BN_FLG_CONSTTIME: sufficient to mask and swap
+ *
+ * BN_FLG_FIXED_TOP: indicates that we haven't called bn_correct_top() on
+ * the data, so the d array may be padded with additional 0 values (i.e.
+ * top could be greater than the minimal value that it could be). We should
+ * be swapping it
+ */
+
+#define BN_CONSTTIME_SWAP_FLAGS (BN_FLG_CONSTTIME | BN_FLG_FIXED_TOP)
+
+ t = ((a->flags ^ b->flags) & BN_CONSTTIME_SWAP_FLAGS) & condition;
+ a->flags ^= t;
+ b->flags ^= t;
+
+ /* conditionally swap the data */
+ for (i = 0; i < nwords; i++) {
+ t = (a->d[i] ^ b->d[i]) & condition;
+ a->d[i] ^= t;
+ b->d[i] ^= t;
+ }
+}
+
+#undef BN_CONSTTIME_SWAP_FLAGS
+
+/* Bits of security, see SP800-57 */
+
+int BN_security_bits(int L, int N)
+{
+ int secbits, bits;
+ if (L >= 15360)
+ secbits = 256;
+ else if (L >= 7680)
+ secbits = 192;
+ else if (L >= 3072)
+ secbits = 128;
+ else if (L >= 2048)
+ secbits = 112;
+ else if (L >= 1024)
+ secbits = 80;
+ else
+ return 0;
+ if (N == -1)
+ return secbits;
+ bits = N / 2;
+ if (bits < 80)
+ return 0;
+ return bits >= secbits ? secbits : bits;
+}
+
+void BN_zero_ex(BIGNUM *a)
+{
+ a->neg = 0;
+ a->top = 0;
+ a->flags &= ~BN_FLG_FIXED_TOP;
+}
+
+int BN_abs_is_word(const BIGNUM *a, const BN_ULONG w)
+{
+ return ((a->top == 1) && (a->d[0] == w)) || ((w == 0) && (a->top == 0));
+}
+
+int BN_is_zero(const BIGNUM *a)
+{
+ return a->top == 0;
+}
+
+int BN_is_one(const BIGNUM *a)
+{
+ return BN_abs_is_word(a, 1) && !a->neg;
+}
+
+int BN_is_word(const BIGNUM *a, const BN_ULONG w)
+{
+ return BN_abs_is_word(a, w) && (!w || !a->neg);
+}
+
+int BN_is_odd(const BIGNUM *a)
+{
+ return (a->top > 0) && (a->d[0] & 1);
+}
+
+int BN_is_negative(const BIGNUM *a)
+{
+ return (a->neg != 0);
+}
+
+int BN_to_montgomery(BIGNUM *r, const BIGNUM *a, BN_MONT_CTX *mont,
+ BN_CTX *ctx)
+{
+ return BN_mod_mul_montgomery(r, a, &(mont->RR), mont, ctx);
+}
+
+void BN_with_flags(BIGNUM *dest, const BIGNUM *b, int flags)
+{
+ dest->d = b->d;
+ dest->top = b->top;
+ dest->dmax = b->dmax;
+ dest->neg = b->neg;
+ dest->flags = ((dest->flags & BN_FLG_MALLOCED)
+ | (b->flags & ~BN_FLG_MALLOCED)
+ | BN_FLG_STATIC_DATA | flags);
+}
+
+BN_GENCB *BN_GENCB_new(void)
+{
+ BN_GENCB *ret;
+
+ if ((ret = OPENSSL_malloc(sizeof(*ret))) == NULL) {
+ BNerr(BN_F_BN_GENCB_NEW, ERR_R_MALLOC_FAILURE);
+ return NULL;
+ }
+
+ return ret;
+}
+
+void BN_GENCB_free(BN_GENCB *cb)
+{
+ if (cb == NULL)
+ return;
+ OPENSSL_free(cb);
+}
+
+void BN_set_flags(BIGNUM *b, int n)
+{
+ b->flags |= n;
+}
+
+int BN_get_flags(const BIGNUM *b, int n)
+{
+ return b->flags & n;
+}
+
+/* Populate a BN_GENCB structure with an "old"-style callback */
+void BN_GENCB_set_old(BN_GENCB *gencb, void (*callback) (int, int, void *),
+ void *cb_arg)
+{
+ BN_GENCB *tmp_gencb = gencb;
+ tmp_gencb->ver = 1;
+ tmp_gencb->arg = cb_arg;
+ tmp_gencb->cb.cb_1 = callback;
+}
+
+/* Populate a BN_GENCB structure with a "new"-style callback */
+void BN_GENCB_set(BN_GENCB *gencb, int (*callback) (int, int, BN_GENCB *),
+ void *cb_arg)
+{
+ BN_GENCB *tmp_gencb = gencb;
+ tmp_gencb->ver = 2;
+ tmp_gencb->arg = cb_arg;
+ tmp_gencb->cb.cb_2 = callback;
+}
+
+void *BN_GENCB_get_arg(BN_GENCB *cb)
+{
+ return cb->arg;
+}
+
+BIGNUM *bn_wexpand(BIGNUM *a, int words)
+{
+ return (words <= a->dmax) ? a : bn_expand2(a, words);
+}
+
+void bn_correct_top(BIGNUM *a)
+{
+ BN_ULONG *ftl;
+ int tmp_top = a->top;
+
+ if (tmp_top > 0) {
+ for (ftl = &(a->d[tmp_top]); tmp_top > 0; tmp_top--) {
+ ftl--;
+ if (*ftl != 0)
+ break;
+ }
+ a->top = tmp_top;
+ }
+ if (a->top == 0)
+ a->neg = 0;
+ a->flags &= ~BN_FLG_FIXED_TOP;
+ bn_pollute(a);
+}
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_local.h b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_local.h
new file mode 100644
index 0000000..8ad69cc
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_local.h
@@ -0,0 +1,668 @@
+/*
+ * Copyright 1995-2020 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#ifndef OSSL_CRYPTO_BN_LOCAL_H
+# define OSSL_CRYPTO_BN_LOCAL_H
+
+/*
+ * The EDK2 build doesn't use bn_conf.h; it sets THIRTY_TWO_BIT or
+ * SIXTY_FOUR_BIT in its own environment since it doesn't re-run our
+ * Configure script and needs to support both 32-bit and 64-bit.
+ */
+# include <openssl/opensslconf.h>
+
+# if !defined(OPENSSL_SYS_UEFI)
+# include "crypto/bn_conf.h"
+# endif
+
+# include "crypto/bn.h"
+
+/*
+ * These preprocessor symbols control various aspects of the bignum headers
+ * and library code. They're not defined by any "normal" configuration, as
+ * they are intended for development and testing purposes. NB: defining all
+ * three can be useful for debugging application code as well as openssl
+ * itself. BN_DEBUG - turn on various debugging alterations to the bignum
+ * code BN_DEBUG_RAND - uses random poisoning of unused words to trip up
+ * mismanagement of bignum internals. You must also define BN_DEBUG.
+ */
+/* #define BN_DEBUG */
+/* #define BN_DEBUG_RAND */
+
+# ifndef OPENSSL_SMALL_FOOTPRINT
+# define BN_MUL_COMBA
+# define BN_SQR_COMBA
+# define BN_RECURSION
+# endif
+
+/*
+ * This next option uses the C libraries (2 word)/(1 word) function. If it is
+ * not defined, I use my C version (which is slower). The reason for this
+ * flag is that when the particular C compiler library routine is used, and
+ * the library is linked with a different compiler, the library is missing.
+ * This mostly happens when the library is built with gcc and then linked
+ * using normal cc. This would be a common occurrence because gcc normally
+ * produces code that is 2 times faster than system compilers for the big
+ * number stuff. For machines with only one compiler (or shared libraries),
+ * this should be on. Again this in only really a problem on machines using
+ * "long long's", are 32bit, and are not using my assembler code.
+ */
+# if defined(OPENSSL_SYS_MSDOS) || defined(OPENSSL_SYS_WINDOWS) || \
+ defined(OPENSSL_SYS_WIN32) || defined(linux)
+# define BN_DIV2W
+# endif
+
+/*
+ * 64-bit processor with LP64 ABI
+ */
+# ifdef SIXTY_FOUR_BIT_LONG
+# define BN_ULLONG unsigned long long
+# define BN_BITS4 32
+# define BN_MASK2 (0xffffffffffffffffL)
+# define BN_MASK2l (0xffffffffL)
+# define BN_MASK2h (0xffffffff00000000L)
+# define BN_MASK2h1 (0xffffffff80000000L)
+# define BN_DEC_CONV (10000000000000000000UL)
+# define BN_DEC_NUM 19
+# define BN_DEC_FMT1 "%lu"
+# define BN_DEC_FMT2 "%019lu"
+# endif
+
+/*
+ * 64-bit processor other than LP64 ABI
+ */
+# ifdef SIXTY_FOUR_BIT
+# undef BN_LLONG
+# undef BN_ULLONG
+# define BN_BITS4 32
+# define BN_MASK2 (0xffffffffffffffffLL)
+# define BN_MASK2l (0xffffffffL)
+# define BN_MASK2h (0xffffffff00000000LL)
+# define BN_MASK2h1 (0xffffffff80000000LL)
+# define BN_DEC_CONV (10000000000000000000ULL)
+# define BN_DEC_NUM 19
+# define BN_DEC_FMT1 "%llu"
+# define BN_DEC_FMT2 "%019llu"
+# endif
+
+# ifdef THIRTY_TWO_BIT
+# ifdef BN_LLONG
+# if defined(_WIN32) && !defined(__GNUC__)
+# define BN_ULLONG unsigned __int64
+# else
+# define BN_ULLONG unsigned long long
+# endif
+# endif
+# define BN_BITS4 16
+# define BN_MASK2 (0xffffffffL)
+# define BN_MASK2l (0xffff)
+# define BN_MASK2h1 (0xffff8000L)
+# define BN_MASK2h (0xffff0000L)
+# define BN_DEC_CONV (1000000000L)
+# define BN_DEC_NUM 9
+# define BN_DEC_FMT1 "%u"
+# define BN_DEC_FMT2 "%09u"
+# endif
+
+
+/*-
+ * Bignum consistency macros
+ * There is one "API" macro, bn_fix_top(), for stripping leading zeroes from
+ * bignum data after direct manipulations on the data. There is also an
+ * "internal" macro, bn_check_top(), for verifying that there are no leading
+ * zeroes. Unfortunately, some auditing is required due to the fact that
+ * bn_fix_top() has become an overabused duct-tape because bignum data is
+ * occasionally passed around in an inconsistent state. So the following
+ * changes have been made to sort this out;
+ * - bn_fix_top()s implementation has been moved to bn_correct_top()
+ * - if BN_DEBUG isn't defined, bn_fix_top() maps to bn_correct_top(), and
+ * bn_check_top() is as before.
+ * - if BN_DEBUG *is* defined;
+ * - bn_check_top() tries to pollute unused words even if the bignum 'top' is
+ * consistent. (ed: only if BN_DEBUG_RAND is defined)
+ * - bn_fix_top() maps to bn_check_top() rather than "fixing" anything.
+ * The idea is to have debug builds flag up inconsistent bignums when they
+ * occur. If that occurs in a bn_fix_top(), we examine the code in question; if
+ * the use of bn_fix_top() was appropriate (ie. it follows directly after code
+ * that manipulates the bignum) it is converted to bn_correct_top(), and if it
+ * was not appropriate, we convert it permanently to bn_check_top() and track
+ * down the cause of the bug. Eventually, no internal code should be using the
+ * bn_fix_top() macro. External applications and libraries should try this with
+ * their own code too, both in terms of building against the openssl headers
+ * with BN_DEBUG defined *and* linking with a version of OpenSSL built with it
+ * defined. This not only improves external code, it provides more test
+ * coverage for openssl's own code.
+ */
+
+# ifdef BN_DEBUG
+/*
+ * The new BN_FLG_FIXED_TOP flag marks vectors that were not treated with
+ * bn_correct_top, in other words such vectors are permitted to have zeros
+ * in most significant limbs. Such vectors are used internally to achieve
+ * execution time invariance for critical operations with private keys.
+ * It's BN_DEBUG-only flag, because user application is not supposed to
+ * observe it anyway. Moreover, optimizing compiler would actually remove
+ * all operations manipulating the bit in question in non-BN_DEBUG build.
+ */
+# define BN_FLG_FIXED_TOP 0x10000
+# ifdef BN_DEBUG_RAND
+# define bn_pollute(a) \
+ do { \
+ const BIGNUM *_bnum1 = (a); \
+ if (_bnum1->top < _bnum1->dmax) { \
+ unsigned char _tmp_char; \
+ /* We cast away const without the compiler knowing, any \
+ * *genuinely* constant variables that aren't mutable \
+ * wouldn't be constructed with top!=dmax. */ \
+ BN_ULONG *_not_const; \
+ memcpy(&_not_const, &_bnum1->d, sizeof(_not_const)); \
+ RAND_bytes(&_tmp_char, 1); /* Debug only - safe to ignore error return */\
+ memset(_not_const + _bnum1->top, _tmp_char, \
+ sizeof(*_not_const) * (_bnum1->dmax - _bnum1->top)); \
+ } \
+ } while(0)
+# else
+# define bn_pollute(a)
+# endif
+# define bn_check_top(a) \
+ do { \
+ const BIGNUM *_bnum2 = (a); \
+ if (_bnum2 != NULL) { \
+ int _top = _bnum2->top; \
+ (void)ossl_assert((_top == 0 && !_bnum2->neg) || \
+ (_top && ((_bnum2->flags & BN_FLG_FIXED_TOP) \
+ || _bnum2->d[_top - 1] != 0))); \
+ bn_pollute(_bnum2); \
+ } \
+ } while(0)
+
+# define bn_fix_top(a) bn_check_top(a)
+
+# define bn_check_size(bn, bits) bn_wcheck_size(bn, ((bits+BN_BITS2-1))/BN_BITS2)
+# define bn_wcheck_size(bn, words) \
+ do { \
+ const BIGNUM *_bnum2 = (bn); \
+ assert((words) <= (_bnum2)->dmax && \
+ (words) >= (_bnum2)->top); \
+ /* avoid unused variable warning with NDEBUG */ \
+ (void)(_bnum2); \
+ } while(0)
+
+# else /* !BN_DEBUG */
+
+# define BN_FLG_FIXED_TOP 0
+# define bn_pollute(a)
+# define bn_check_top(a)
+# define bn_fix_top(a) bn_correct_top(a)
+# define bn_check_size(bn, bits)
+# define bn_wcheck_size(bn, words)
+
+# endif
+
+BN_ULONG bn_mul_add_words(BN_ULONG *rp, const BN_ULONG *ap, int num,
+ BN_ULONG w);
+BN_ULONG bn_mul_words(BN_ULONG *rp, const BN_ULONG *ap, int num, BN_ULONG w);
+void bn_sqr_words(BN_ULONG *rp, const BN_ULONG *ap, int num);
+BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d);
+BN_ULONG bn_add_words(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
+ int num);
+BN_ULONG bn_sub_words(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
+ int num);
+
+struct bignum_st {
+ BN_ULONG *d; /* Pointer to an array of 'BN_BITS2' bit
+ * chunks. */
+ int top; /* Index of last used d +1. */
+ /* The next are internal book keeping for bn_expand. */
+ int dmax; /* Size of the d array. */
+ int neg; /* one if the number is negative */
+ int flags;
+};
+
+/* Used for montgomery multiplication */
+struct bn_mont_ctx_st {
+ int ri; /* number of bits in R */
+ BIGNUM RR; /* used to convert to montgomery form,
+ possibly zero-padded */
+ BIGNUM N; /* The modulus */
+ BIGNUM Ni; /* R*(1/R mod N) - N*Ni = 1 (Ni is only
+ * stored for bignum algorithm) */
+ BN_ULONG n0[2]; /* least significant word(s) of Ni; (type
+ * changed with 0.9.9, was "BN_ULONG n0;"
+ * before) */
+ int flags;
+};
+
+/*
+ * Used for reciprocal division/mod functions It cannot be shared between
+ * threads
+ */
+struct bn_recp_ctx_st {
+ BIGNUM N; /* the divisor */
+ BIGNUM Nr; /* the reciprocal */
+ int num_bits;
+ int shift;
+ int flags;
+};
+
+/* Used for slow "generation" functions. */
+struct bn_gencb_st {
+ unsigned int ver; /* To handle binary (in)compatibility */
+ void *arg; /* callback-specific data */
+ union {
+ /* if (ver==1) - handles old style callbacks */
+ void (*cb_1) (int, int, void *);
+ /* if (ver==2) - new callback style */
+ int (*cb_2) (int, int, BN_GENCB *);
+ } cb;
+};
+
+/*-
+ * BN_window_bits_for_exponent_size -- macro for sliding window mod_exp functions
+ *
+ *
+ * For window size 'w' (w >= 2) and a random 'b' bits exponent,
+ * the number of multiplications is a constant plus on average
+ *
+ * 2^(w-1) + (b-w)/(w+1);
+ *
+ * here 2^(w-1) is for precomputing the table (we actually need
+ * entries only for windows that have the lowest bit set), and
+ * (b-w)/(w+1) is an approximation for the expected number of
+ * w-bit windows, not counting the first one.
+ *
+ * Thus we should use
+ *
+ * w >= 6 if b > 671
+ * w = 5 if 671 > b > 239
+ * w = 4 if 239 > b > 79
+ * w = 3 if 79 > b > 23
+ * w <= 2 if 23 > b
+ *
+ * (with draws in between). Very small exponents are often selected
+ * with low Hamming weight, so we use w = 1 for b <= 23.
+ */
+# define BN_window_bits_for_exponent_size(b) \
+ ((b) > 671 ? 6 : \
+ (b) > 239 ? 5 : \
+ (b) > 79 ? 4 : \
+ (b) > 23 ? 3 : 1)
+
+/*
+ * BN_mod_exp_mont_consttime is based on the assumption that the L1 data cache
+ * line width of the target processor is at least the following value.
+ */
+# define MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH ( 64 )
+# define MOD_EXP_CTIME_MIN_CACHE_LINE_MASK (MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH - 1)
+
+/*
+ * Window sizes optimized for fixed window size modular exponentiation
+ * algorithm (BN_mod_exp_mont_consttime). To achieve the security goals of
+ * BN_mode_exp_mont_consttime, the maximum size of the window must not exceed
+ * log_2(MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH). Window size thresholds are
+ * defined for cache line sizes of 32 and 64, cache line sizes where
+ * log_2(32)=5 and log_2(64)=6 respectively. A window size of 7 should only be
+ * used on processors that have a 128 byte or greater cache line size.
+ */
+# if MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH == 64
+
+# define BN_window_bits_for_ctime_exponent_size(b) \
+ ((b) > 937 ? 6 : \
+ (b) > 306 ? 5 : \
+ (b) > 89 ? 4 : \
+ (b) > 22 ? 3 : 1)
+# define BN_MAX_WINDOW_BITS_FOR_CTIME_EXPONENT_SIZE (6)
+
+# elif MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH == 32
+
+# define BN_window_bits_for_ctime_exponent_size(b) \
+ ((b) > 306 ? 5 : \
+ (b) > 89 ? 4 : \
+ (b) > 22 ? 3 : 1)
+# define BN_MAX_WINDOW_BITS_FOR_CTIME_EXPONENT_SIZE (5)
+
+# endif
+
+/* Pentium pro 16,16,16,32,64 */
+/* Alpha 16,16,16,16.64 */
+# define BN_MULL_SIZE_NORMAL (16)/* 32 */
+# define BN_MUL_RECURSIVE_SIZE_NORMAL (16)/* 32 less than */
+# define BN_SQR_RECURSIVE_SIZE_NORMAL (16)/* 32 */
+# define BN_MUL_LOW_RECURSIVE_SIZE_NORMAL (32)/* 32 */
+# define BN_MONT_CTX_SET_SIZE_WORD (64)/* 32 */
+
+/*
+ * 2011-02-22 SMS. In various places, a size_t variable or a type cast to
+ * size_t was used to perform integer-only operations on pointers. This
+ * failed on VMS with 64-bit pointers (CC /POINTER_SIZE = 64) because size_t
+ * is still only 32 bits. What's needed in these cases is an integer type
+ * with the same size as a pointer, which size_t is not certain to be. The
+ * only fix here is VMS-specific.
+ */
+# if defined(OPENSSL_SYS_VMS)
+# if __INITIAL_POINTER_SIZE == 64
+# define PTR_SIZE_INT long long
+# else /* __INITIAL_POINTER_SIZE == 64 */
+# define PTR_SIZE_INT int
+# endif /* __INITIAL_POINTER_SIZE == 64 [else] */
+# elif !defined(PTR_SIZE_INT) /* defined(OPENSSL_SYS_VMS) */
+# define PTR_SIZE_INT size_t
+# endif /* defined(OPENSSL_SYS_VMS) [else] */
+
+# if !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM) && !defined(PEDANTIC)
+/*
+ * BN_UMULT_HIGH section.
+ * If the compiler doesn't support 2*N integer type, then you have to
+ * replace every N*N multiplication with 4 (N/2)*(N/2) accompanied by some
+ * shifts and additions which unavoidably results in severe performance
+ * penalties. Of course provided that the hardware is capable of producing
+ * 2*N result... That's when you normally start considering assembler
+ * implementation. However! It should be pointed out that some CPUs (e.g.,
+ * PowerPC, Alpha, and IA-64) provide *separate* instruction calculating
+ * the upper half of the product placing the result into a general
+ * purpose register. Now *if* the compiler supports inline assembler,
+ * then it's not impossible to implement the "bignum" routines (and have
+ * the compiler optimize 'em) exhibiting "native" performance in C. That's
+ * what BN_UMULT_HIGH macro is about:-) Note that more recent compilers do
+ * support 2*64 integer type, which is also used here.
+ */
+# if defined(__SIZEOF_INT128__) && __SIZEOF_INT128__==16 && \
+ (defined(SIXTY_FOUR_BIT) || defined(SIXTY_FOUR_BIT_LONG))
+# define BN_UMULT_HIGH(a,b) (((__uint128_t)(a)*(b))>>64)
+# define BN_UMULT_LOHI(low,high,a,b) ({ \
+ __uint128_t ret=(__uint128_t)(a)*(b); \
+ (high)=ret>>64; (low)=ret; })
+# elif defined(__alpha) && (defined(SIXTY_FOUR_BIT_LONG) || defined(SIXTY_FOUR_BIT))
+# if defined(__DECC)
+# include <c_asm.h>
+# define BN_UMULT_HIGH(a,b) (BN_ULONG)asm("umulh %a0,%a1,%v0",(a),(b))
+# elif defined(__GNUC__) && __GNUC__>=2
+# define BN_UMULT_HIGH(a,b) ({ \
+ register BN_ULONG ret; \
+ asm ("umulh %1,%2,%0" \
+ : "=r"(ret) \
+ : "r"(a), "r"(b)); \
+ ret; })
+# endif /* compiler */
+# elif defined(_ARCH_PPC64) && defined(SIXTY_FOUR_BIT_LONG)
+# if defined(__GNUC__) && __GNUC__>=2
+# define BN_UMULT_HIGH(a,b) ({ \
+ register BN_ULONG ret; \
+ asm ("mulhdu %0,%1,%2" \
+ : "=r"(ret) \
+ : "r"(a), "r"(b)); \
+ ret; })
+# endif /* compiler */
+# elif (defined(__x86_64) || defined(__x86_64__)) && \
+ (defined(SIXTY_FOUR_BIT_LONG) || defined(SIXTY_FOUR_BIT))
+# if defined(__GNUC__) && __GNUC__>=2
+# define BN_UMULT_HIGH(a,b) ({ \
+ register BN_ULONG ret,discard; \
+ asm ("mulq %3" \
+ : "=a"(discard),"=d"(ret) \
+ : "a"(a), "g"(b) \
+ : "cc"); \
+ ret; })
+# define BN_UMULT_LOHI(low,high,a,b) \
+ asm ("mulq %3" \
+ : "=a"(low),"=d"(high) \
+ : "a"(a),"g"(b) \
+ : "cc");
+# endif
+# elif (defined(_M_AMD64) || defined(_M_X64)) && defined(SIXTY_FOUR_BIT)
+# if defined(_MSC_VER) && _MSC_VER>=1400
+unsigned __int64 __umulh(unsigned __int64 a, unsigned __int64 b);
+unsigned __int64 _umul128(unsigned __int64 a, unsigned __int64 b,
+ unsigned __int64 *h);
+# pragma intrinsic(__umulh,_umul128)
+# define BN_UMULT_HIGH(a,b) __umulh((a),(b))
+# define BN_UMULT_LOHI(low,high,a,b) ((low)=_umul128((a),(b),&(high)))
+# endif
+# elif defined(__mips) && (defined(SIXTY_FOUR_BIT) || defined(SIXTY_FOUR_BIT_LONG))
+# if defined(__GNUC__) && __GNUC__>=2
+# define BN_UMULT_HIGH(a,b) ({ \
+ register BN_ULONG ret; \
+ asm ("dmultu %1,%2" \
+ : "=h"(ret) \
+ : "r"(a), "r"(b) : "l"); \
+ ret; })
+# define BN_UMULT_LOHI(low,high,a,b) \
+ asm ("dmultu %2,%3" \
+ : "=l"(low),"=h"(high) \
+ : "r"(a), "r"(b));
+# endif
+# elif defined(__aarch64__) && defined(SIXTY_FOUR_BIT_LONG)
+# if defined(__GNUC__) && __GNUC__>=2
+# define BN_UMULT_HIGH(a,b) ({ \
+ register BN_ULONG ret; \
+ asm ("umulh %0,%1,%2" \
+ : "=r"(ret) \
+ : "r"(a), "r"(b)); \
+ ret; })
+# endif
+# endif /* cpu */
+# endif /* OPENSSL_NO_ASM */
+
+# ifdef BN_DEBUG_RAND
+# define bn_clear_top2max(a) \
+ { \
+ int ind = (a)->dmax - (a)->top; \
+ BN_ULONG *ftl = &(a)->d[(a)->top-1]; \
+ for (; ind != 0; ind--) \
+ *(++ftl) = 0x0; \
+ }
+# else
+# define bn_clear_top2max(a)
+# endif
+
+# ifdef BN_LLONG
+/*******************************************************************
+ * Using the long long type, has to be twice as wide as BN_ULONG...
+ */
+# define Lw(t) (((BN_ULONG)(t))&BN_MASK2)
+# define Hw(t) (((BN_ULONG)((t)>>BN_BITS2))&BN_MASK2)
+
+# define mul_add(r,a,w,c) { \
+ BN_ULLONG t; \
+ t=(BN_ULLONG)w * (a) + (r) + (c); \
+ (r)= Lw(t); \
+ (c)= Hw(t); \
+ }
+
+# define mul(r,a,w,c) { \
+ BN_ULLONG t; \
+ t=(BN_ULLONG)w * (a) + (c); \
+ (r)= Lw(t); \
+ (c)= Hw(t); \
+ }
+
+# define sqr(r0,r1,a) { \
+ BN_ULLONG t; \
+ t=(BN_ULLONG)(a)*(a); \
+ (r0)=Lw(t); \
+ (r1)=Hw(t); \
+ }
+
+# elif defined(BN_UMULT_LOHI)
+# define mul_add(r,a,w,c) { \
+ BN_ULONG high,low,ret,tmp=(a); \
+ ret = (r); \
+ BN_UMULT_LOHI(low,high,w,tmp); \
+ ret += (c); \
+ (c) = (ret<(c))?1:0; \
+ (c) += high; \
+ ret += low; \
+ (c) += (ret<low)?1:0; \
+ (r) = ret; \
+ }
+
+# define mul(r,a,w,c) { \
+ BN_ULONG high,low,ret,ta=(a); \
+ BN_UMULT_LOHI(low,high,w,ta); \
+ ret = low + (c); \
+ (c) = high; \
+ (c) += (ret<low)?1:0; \
+ (r) = ret; \
+ }
+
+# define sqr(r0,r1,a) { \
+ BN_ULONG tmp=(a); \
+ BN_UMULT_LOHI(r0,r1,tmp,tmp); \
+ }
+
+# elif defined(BN_UMULT_HIGH)
+# define mul_add(r,a,w,c) { \
+ BN_ULONG high,low,ret,tmp=(a); \
+ ret = (r); \
+ high= BN_UMULT_HIGH(w,tmp); \
+ ret += (c); \
+ low = (w) * tmp; \
+ (c) = (ret<(c))?1:0; \
+ (c) += high; \
+ ret += low; \
+ (c) += (ret<low)?1:0; \
+ (r) = ret; \
+ }
+
+# define mul(r,a,w,c) { \
+ BN_ULONG high,low,ret,ta=(a); \
+ low = (w) * ta; \
+ high= BN_UMULT_HIGH(w,ta); \
+ ret = low + (c); \
+ (c) = high; \
+ (c) += (ret<low)?1:0; \
+ (r) = ret; \
+ }
+
+# define sqr(r0,r1,a) { \
+ BN_ULONG tmp=(a); \
+ (r0) = tmp * tmp; \
+ (r1) = BN_UMULT_HIGH(tmp,tmp); \
+ }
+
+# else
+/*************************************************************
+ * No long long type
+ */
+
+# define LBITS(a) ((a)&BN_MASK2l)
+# define HBITS(a) (((a)>>BN_BITS4)&BN_MASK2l)
+# define L2HBITS(a) (((a)<<BN_BITS4)&BN_MASK2)
+
+# define LLBITS(a) ((a)&BN_MASKl)
+# define LHBITS(a) (((a)>>BN_BITS2)&BN_MASKl)
+# define LL2HBITS(a) ((BN_ULLONG)((a)&BN_MASKl)<<BN_BITS2)
+
+# define mul64(l,h,bl,bh) \
+ { \
+ BN_ULONG m,m1,lt,ht; \
+ \
+ lt=l; \
+ ht=h; \
+ m =(bh)*(lt); \
+ lt=(bl)*(lt); \
+ m1=(bl)*(ht); \
+ ht =(bh)*(ht); \
+ m=(m+m1)&BN_MASK2; if (m < m1) ht+=L2HBITS((BN_ULONG)1); \
+ ht+=HBITS(m); \
+ m1=L2HBITS(m); \
+ lt=(lt+m1)&BN_MASK2; if (lt < m1) ht++; \
+ (l)=lt; \
+ (h)=ht; \
+ }
+
+# define sqr64(lo,ho,in) \
+ { \
+ BN_ULONG l,h,m; \
+ \
+ h=(in); \
+ l=LBITS(h); \
+ h=HBITS(h); \
+ m =(l)*(h); \
+ l*=l; \
+ h*=h; \
+ h+=(m&BN_MASK2h1)>>(BN_BITS4-1); \
+ m =(m&BN_MASK2l)<<(BN_BITS4+1); \
+ l=(l+m)&BN_MASK2; if (l < m) h++; \
+ (lo)=l; \
+ (ho)=h; \
+ }
+
+# define mul_add(r,a,bl,bh,c) { \
+ BN_ULONG l,h; \
+ \
+ h= (a); \
+ l=LBITS(h); \
+ h=HBITS(h); \
+ mul64(l,h,(bl),(bh)); \
+ \
+ /* non-multiply part */ \
+ l=(l+(c))&BN_MASK2; if (l < (c)) h++; \
+ (c)=(r); \
+ l=(l+(c))&BN_MASK2; if (l < (c)) h++; \
+ (c)=h&BN_MASK2; \
+ (r)=l; \
+ }
+
+# define mul(r,a,bl,bh,c) { \
+ BN_ULONG l,h; \
+ \
+ h= (a); \
+ l=LBITS(h); \
+ h=HBITS(h); \
+ mul64(l,h,(bl),(bh)); \
+ \
+ /* non-multiply part */ \
+ l+=(c); if ((l&BN_MASK2) < (c)) h++; \
+ (c)=h&BN_MASK2; \
+ (r)=l&BN_MASK2; \
+ }
+# endif /* !BN_LLONG */
+
+void BN_RECP_CTX_init(BN_RECP_CTX *recp);
+void BN_MONT_CTX_init(BN_MONT_CTX *ctx);
+
+void bn_init(BIGNUM *a);
+void bn_mul_normal(BN_ULONG *r, BN_ULONG *a, int na, BN_ULONG *b, int nb);
+void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b);
+void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b);
+void bn_sqr_normal(BN_ULONG *r, const BN_ULONG *a, int n, BN_ULONG *tmp);
+void bn_sqr_comba8(BN_ULONG *r, const BN_ULONG *a);
+void bn_sqr_comba4(BN_ULONG *r, const BN_ULONG *a);
+int bn_cmp_words(const BN_ULONG *a, const BN_ULONG *b, int n);
+int bn_cmp_part_words(const BN_ULONG *a, const BN_ULONG *b, int cl, int dl);
+void bn_mul_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2,
+ int dna, int dnb, BN_ULONG *t);
+void bn_mul_part_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b,
+ int n, int tna, int tnb, BN_ULONG *t);
+void bn_sqr_recursive(BN_ULONG *r, const BN_ULONG *a, int n2, BN_ULONG *t);
+void bn_mul_low_normal(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n);
+void bn_mul_low_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2,
+ BN_ULONG *t);
+BN_ULONG bn_sub_part_words(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b,
+ int cl, int dl);
+int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
+ const BN_ULONG *np, const BN_ULONG *n0, int num);
+
+BIGNUM *int_bn_mod_inverse(BIGNUM *in,
+ const BIGNUM *a, const BIGNUM *n, BN_CTX *ctx,
+ int *noinv);
+
+static ossl_inline BIGNUM *bn_expand(BIGNUM *a, int bits)
+{
+ if (bits > (INT_MAX - BN_BITS2 + 1))
+ return NULL;
+
+ if (((bits+BN_BITS2-1)/BN_BITS2) <= (a)->dmax)
+ return a;
+
+ return bn_expand2((a),(bits+BN_BITS2-1)/BN_BITS2);
+}
+
+#endif
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_mod.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_mod.c
new file mode 100644
index 0000000..f7d2e26
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_mod.c
@@ -0,0 +1,321 @@
+/*
+ * Copyright 1998-2018 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include "internal/cryptlib.h"
+#include "bn_local.h"
+
+int BN_nnmod(BIGNUM *r, const BIGNUM *m, const BIGNUM *d, BN_CTX *ctx)
+{
+ /*
+ * like BN_mod, but returns non-negative remainder (i.e., 0 <= r < |d|
+ * always holds)
+ */
+
+ if (!(BN_mod(r, m, d, ctx)))
+ return 0;
+ if (!r->neg)
+ return 1;
+ /* now -|d| < r < 0, so we have to set r := r + |d| */
+ return (d->neg ? BN_sub : BN_add) (r, r, d);
+}
+
+int BN_mod_add(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, const BIGNUM *m,
+ BN_CTX *ctx)
+{
+ if (!BN_add(r, a, b))
+ return 0;
+ return BN_nnmod(r, r, m, ctx);
+}
+
+/*
+ * BN_mod_add variant that may be used if both a and b are non-negative and
+ * less than m. The original algorithm was
+ *
+ * if (!BN_uadd(r, a, b))
+ * return 0;
+ * if (BN_ucmp(r, m) >= 0)
+ * return BN_usub(r, r, m);
+ *
+ * which is replaced with addition, subtracting modulus, and conditional
+ * move depending on whether or not subtraction borrowed.
+ */
+int bn_mod_add_fixed_top(BIGNUM *r, const BIGNUM *a, const BIGNUM *b,
+ const BIGNUM *m)
+{
+ size_t i, ai, bi, mtop = m->top;
+ BN_ULONG storage[1024 / BN_BITS2];
+ BN_ULONG carry, temp, mask, *rp, *tp = storage;
+ const BN_ULONG *ap, *bp;
+
+ if (bn_wexpand(r, mtop) == NULL)
+ return 0;
+
+ if (mtop > sizeof(storage) / sizeof(storage[0])
+ && (tp = OPENSSL_malloc(mtop * sizeof(BN_ULONG))) == NULL)
+ return 0;
+
+ ap = a->d != NULL ? a->d : tp;
+ bp = b->d != NULL ? b->d : tp;
+
+ for (i = 0, ai = 0, bi = 0, carry = 0; i < mtop;) {
+ mask = (BN_ULONG)0 - ((i - a->top) >> (8 * sizeof(i) - 1));
+ temp = ((ap[ai] & mask) + carry) & BN_MASK2;
+ carry = (temp < carry);
+
+ mask = (BN_ULONG)0 - ((i - b->top) >> (8 * sizeof(i) - 1));
+ tp[i] = ((bp[bi] & mask) + temp) & BN_MASK2;
+ carry += (tp[i] < temp);
+
+ i++;
+ ai += (i - a->dmax) >> (8 * sizeof(i) - 1);
+ bi += (i - b->dmax) >> (8 * sizeof(i) - 1);
+ }
+ rp = r->d;
+ carry -= bn_sub_words(rp, tp, m->d, mtop);
+ for (i = 0; i < mtop; i++) {
+ rp[i] = (carry & tp[i]) | (~carry & rp[i]);
+ ((volatile BN_ULONG *)tp)[i] = 0;
+ }
+ r->top = mtop;
+ r->flags |= BN_FLG_FIXED_TOP;
+ r->neg = 0;
+
+ if (tp != storage)
+ OPENSSL_free(tp);
+
+ return 1;
+}
+
+int BN_mod_add_quick(BIGNUM *r, const BIGNUM *a, const BIGNUM *b,
+ const BIGNUM *m)
+{
+ int ret = bn_mod_add_fixed_top(r, a, b, m);
+
+ if (ret)
+ bn_correct_top(r);
+
+ return ret;
+}
+
+int BN_mod_sub(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, const BIGNUM *m,
+ BN_CTX *ctx)
+{
+ if (!BN_sub(r, a, b))
+ return 0;
+ return BN_nnmod(r, r, m, ctx);
+}
+
+/*
+ * BN_mod_sub variant that may be used if both a and b are non-negative,
+ * a is less than m, while b is of same bit width as m. It's implemented
+ * as subtraction followed by two conditional additions.
+ *
+ * 0 <= a < m
+ * 0 <= b < 2^w < 2*m
+ *
+ * after subtraction
+ *
+ * -2*m < r = a - b < m
+ *
+ * Thus it takes up to two conditional additions to make |r| positive.
+ */
+int bn_mod_sub_fixed_top(BIGNUM *r, const BIGNUM *a, const BIGNUM *b,
+ const BIGNUM *m)
+{
+ size_t i, ai, bi, mtop = m->top;
+ BN_ULONG borrow, carry, ta, tb, mask, *rp;
+ const BN_ULONG *ap, *bp;
+
+ if (bn_wexpand(r, mtop) == NULL)
+ return 0;
+
+ rp = r->d;
+ ap = a->d != NULL ? a->d : rp;
+ bp = b->d != NULL ? b->d : rp;
+
+ for (i = 0, ai = 0, bi = 0, borrow = 0; i < mtop;) {
+ mask = (BN_ULONG)0 - ((i - a->top) >> (8 * sizeof(i) - 1));
+ ta = ap[ai] & mask;
+
+ mask = (BN_ULONG)0 - ((i - b->top) >> (8 * sizeof(i) - 1));
+ tb = bp[bi] & mask;
+ rp[i] = ta - tb - borrow;
+ if (ta != tb)
+ borrow = (ta < tb);
+
+ i++;
+ ai += (i - a->dmax) >> (8 * sizeof(i) - 1);
+ bi += (i - b->dmax) >> (8 * sizeof(i) - 1);
+ }
+ ap = m->d;
+ for (i = 0, mask = 0 - borrow, carry = 0; i < mtop; i++) {
+ ta = ((ap[i] & mask) + carry) & BN_MASK2;
+ carry = (ta < carry);
+ rp[i] = (rp[i] + ta) & BN_MASK2;
+ carry += (rp[i] < ta);
+ }
+ borrow -= carry;
+ for (i = 0, mask = 0 - borrow, carry = 0; i < mtop; i++) {
+ ta = ((ap[i] & mask) + carry) & BN_MASK2;
+ carry = (ta < carry);
+ rp[i] = (rp[i] + ta) & BN_MASK2;
+ carry += (rp[i] < ta);
+ }
+
+ r->top = mtop;
+ r->flags |= BN_FLG_FIXED_TOP;
+ r->neg = 0;
+
+ return 1;
+}
+
+/*
+ * BN_mod_sub variant that may be used if both a and b are non-negative and
+ * less than m
+ */
+int BN_mod_sub_quick(BIGNUM *r, const BIGNUM *a, const BIGNUM *b,
+ const BIGNUM *m)
+{
+ if (!BN_sub(r, a, b))
+ return 0;
+ if (r->neg)
+ return BN_add(r, r, m);
+ return 1;
+}
+
+/* slow but works */
+int BN_mod_mul(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, const BIGNUM *m,
+ BN_CTX *ctx)
+{
+ BIGNUM *t;
+ int ret = 0;
+
+ bn_check_top(a);
+ bn_check_top(b);
+ bn_check_top(m);
+
+ BN_CTX_start(ctx);
+ if ((t = BN_CTX_get(ctx)) == NULL)
+ goto err;
+ if (a == b) {
+ if (!BN_sqr(t, a, ctx))
+ goto err;
+ } else {
+ if (!BN_mul(t, a, b, ctx))
+ goto err;
+ }
+ if (!BN_nnmod(r, t, m, ctx))
+ goto err;
+ bn_check_top(r);
+ ret = 1;
+ err:
+ BN_CTX_end(ctx);
+ return ret;
+}
+
+int BN_mod_sqr(BIGNUM *r, const BIGNUM *a, const BIGNUM *m, BN_CTX *ctx)
+{
+ if (!BN_sqr(r, a, ctx))
+ return 0;
+ /* r->neg == 0, thus we don't need BN_nnmod */
+ return BN_mod(r, r, m, ctx);
+}
+
+int BN_mod_lshift1(BIGNUM *r, const BIGNUM *a, const BIGNUM *m, BN_CTX *ctx)
+{
+ if (!BN_lshift1(r, a))
+ return 0;
+ bn_check_top(r);
+ return BN_nnmod(r, r, m, ctx);
+}
+
+/*
+ * BN_mod_lshift1 variant that may be used if a is non-negative and less than
+ * m
+ */
+int BN_mod_lshift1_quick(BIGNUM *r, const BIGNUM *a, const BIGNUM *m)
+{
+ if (!BN_lshift1(r, a))
+ return 0;
+ bn_check_top(r);
+ if (BN_cmp(r, m) >= 0)
+ return BN_sub(r, r, m);
+ return 1;
+}
+
+int BN_mod_lshift(BIGNUM *r, const BIGNUM *a, int n, const BIGNUM *m,
+ BN_CTX *ctx)
+{
+ BIGNUM *abs_m = NULL;
+ int ret;
+
+ if (!BN_nnmod(r, a, m, ctx))
+ return 0;
+
+ if (m->neg) {
+ abs_m = BN_dup(m);
+ if (abs_m == NULL)
+ return 0;
+ abs_m->neg = 0;
+ }
+
+ ret = BN_mod_lshift_quick(r, r, n, (abs_m ? abs_m : m));
+ bn_check_top(r);
+
+ BN_free(abs_m);
+ return ret;
+}
+
+/*
+ * BN_mod_lshift variant that may be used if a is non-negative and less than
+ * m
+ */
+int BN_mod_lshift_quick(BIGNUM *r, const BIGNUM *a, int n, const BIGNUM *m)
+{
+ if (r != a) {
+ if (BN_copy(r, a) == NULL)
+ return 0;
+ }
+
+ while (n > 0) {
+ int max_shift;
+
+ /* 0 < r < m */
+ max_shift = BN_num_bits(m) - BN_num_bits(r);
+ /* max_shift >= 0 */
+
+ if (max_shift < 0) {
+ BNerr(BN_F_BN_MOD_LSHIFT_QUICK, BN_R_INPUT_NOT_REDUCED);
+ return 0;
+ }
+
+ if (max_shift > n)
+ max_shift = n;
+
+ if (max_shift) {
+ if (!BN_lshift(r, r, max_shift))
+ return 0;
+ n -= max_shift;
+ } else {
+ if (!BN_lshift1(r, r))
+ return 0;
+ --n;
+ }
+
+ /* BN_num_bits(r) <= BN_num_bits(m) */
+
+ if (BN_cmp(r, m) >= 0) {
+ if (!BN_sub(r, r, m))
+ return 0;
+ }
+ }
+ bn_check_top(r);
+
+ return 1;
+}
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_mont.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_mont.c
new file mode 100644
index 0000000..1e5045a
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_mont.c
@@ -0,0 +1,464 @@
+/*
+ * Copyright 1995-2018 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+/*
+ * Details about Montgomery multiplication algorithms can be found at
+ * http://security.ece.orst.edu/publications.html, e.g.
+ * http://security.ece.orst.edu/koc/papers/j37acmon.pdf and
+ * sections 3.8 and 4.2 in http://security.ece.orst.edu/koc/papers/r01rsasw.pdf
+ */
+
+#include "internal/cryptlib.h"
+#include "bn_local.h"
+
+#define MONT_WORD /* use the faster word-based algorithm */
+
+#ifdef MONT_WORD
+static int bn_from_montgomery_word(BIGNUM *ret, BIGNUM *r, BN_MONT_CTX *mont);
+#endif
+
+int BN_mod_mul_montgomery(BIGNUM *r, const BIGNUM *a, const BIGNUM *b,
+ BN_MONT_CTX *mont, BN_CTX *ctx)
+{
+ int ret = bn_mul_mont_fixed_top(r, a, b, mont, ctx);
+
+ bn_correct_top(r);
+ bn_check_top(r);
+
+ return ret;
+}
+
+int bn_mul_mont_fixed_top(BIGNUM *r, const BIGNUM *a, const BIGNUM *b,
+ BN_MONT_CTX *mont, BN_CTX *ctx)
+{
+ BIGNUM *tmp;
+ int ret = 0;
+ int num = mont->N.top;
+
+#if defined(OPENSSL_BN_ASM_MONT) && defined(MONT_WORD)
+ if (num > 1 && a->top == num && b->top == num) {
+ if (bn_wexpand(r, num) == NULL)
+ return 0;
+ if (bn_mul_mont(r->d, a->d, b->d, mont->N.d, mont->n0, num)) {
+ r->neg = a->neg ^ b->neg;
+ r->top = num;
+ r->flags |= BN_FLG_FIXED_TOP;
+ return 1;
+ }
+ }
+#endif
+
+ if ((a->top + b->top) > 2 * num)
+ return 0;
+
+ BN_CTX_start(ctx);
+ tmp = BN_CTX_get(ctx);
+ if (tmp == NULL)
+ goto err;
+
+ bn_check_top(tmp);
+ if (a == b) {
+ if (!bn_sqr_fixed_top(tmp, a, ctx))
+ goto err;
+ } else {
+ if (!bn_mul_fixed_top(tmp, a, b, ctx))
+ goto err;
+ }
+ /* reduce from aRR to aR */
+#ifdef MONT_WORD
+ if (!bn_from_montgomery_word(r, tmp, mont))
+ goto err;
+#else
+ if (!BN_from_montgomery(r, tmp, mont, ctx))
+ goto err;
+#endif
+ ret = 1;
+ err:
+ BN_CTX_end(ctx);
+ return ret;
+}
+
+#ifdef MONT_WORD
+static int bn_from_montgomery_word(BIGNUM *ret, BIGNUM *r, BN_MONT_CTX *mont)
+{
+ BIGNUM *n;
+ BN_ULONG *ap, *np, *rp, n0, v, carry;
+ int nl, max, i;
+ unsigned int rtop;
+
+ n = &(mont->N);
+ nl = n->top;
+ if (nl == 0) {
+ ret->top = 0;
+ return 1;
+ }
+
+ max = (2 * nl); /* carry is stored separately */
+ if (bn_wexpand(r, max) == NULL)
+ return 0;
+
+ r->neg ^= n->neg;
+ np = n->d;
+ rp = r->d;
+
+ /* clear the top words of T */
+ for (rtop = r->top, i = 0; i < max; i++) {
+ v = (BN_ULONG)0 - ((i - rtop) >> (8 * sizeof(rtop) - 1));
+ rp[i] &= v;
+ }
+
+ r->top = max;
+ r->flags |= BN_FLG_FIXED_TOP;
+ n0 = mont->n0[0];
+
+ /*
+ * Add multiples of |n| to |r| until R = 2^(nl * BN_BITS2) divides it. On
+ * input, we had |r| < |n| * R, so now |r| < 2 * |n| * R. Note that |r|
+ * includes |carry| which is stored separately.
+ */
+ for (carry = 0, i = 0; i < nl; i++, rp++) {
+ v = bn_mul_add_words(rp, np, nl, (rp[0] * n0) & BN_MASK2);
+ v = (v + carry + rp[nl]) & BN_MASK2;
+ carry |= (v != rp[nl]);
+ carry &= (v <= rp[nl]);
+ rp[nl] = v;
+ }
+
+ if (bn_wexpand(ret, nl) == NULL)
+ return 0;
+ ret->top = nl;
+ ret->flags |= BN_FLG_FIXED_TOP;
+ ret->neg = r->neg;
+
+ rp = ret->d;
+
+ /*
+ * Shift |nl| words to divide by R. We have |ap| < 2 * |n|. Note that |ap|
+ * includes |carry| which is stored separately.
+ */
+ ap = &(r->d[nl]);
+
+ carry -= bn_sub_words(rp, ap, np, nl);
+ /*
+ * |carry| is -1 if |ap| - |np| underflowed or zero if it did not. Note
+ * |carry| cannot be 1. That would imply the subtraction did not fit in
+ * |nl| words, and we know at most one subtraction is needed.
+ */
+ for (i = 0; i < nl; i++) {
+ rp[i] = (carry & ap[i]) | (~carry & rp[i]);
+ ap[i] = 0;
+ }
+
+ return 1;
+}
+#endif /* MONT_WORD */
+
+int BN_from_montgomery(BIGNUM *ret, const BIGNUM *a, BN_MONT_CTX *mont,
+ BN_CTX *ctx)
+{
+ int retn;
+
+ retn = bn_from_mont_fixed_top(ret, a, mont, ctx);
+ bn_correct_top(ret);
+ bn_check_top(ret);
+
+ return retn;
+}
+
+int bn_from_mont_fixed_top(BIGNUM *ret, const BIGNUM *a, BN_MONT_CTX *mont,
+ BN_CTX *ctx)
+{
+ int retn = 0;
+#ifdef MONT_WORD
+ BIGNUM *t;
+
+ BN_CTX_start(ctx);
+ if ((t = BN_CTX_get(ctx)) && BN_copy(t, a)) {
+ retn = bn_from_montgomery_word(ret, t, mont);
+ }
+ BN_CTX_end(ctx);
+#else /* !MONT_WORD */
+ BIGNUM *t1, *t2;
+
+ BN_CTX_start(ctx);
+ t1 = BN_CTX_get(ctx);
+ t2 = BN_CTX_get(ctx);
+ if (t2 == NULL)
+ goto err;
+
+ if (!BN_copy(t1, a))
+ goto err;
+ BN_mask_bits(t1, mont->ri);
+
+ if (!BN_mul(t2, t1, &mont->Ni, ctx))
+ goto err;
+ BN_mask_bits(t2, mont->ri);
+
+ if (!BN_mul(t1, t2, &mont->N, ctx))
+ goto err;
+ if (!BN_add(t2, a, t1))
+ goto err;
+ if (!BN_rshift(ret, t2, mont->ri))
+ goto err;
+
+ if (BN_ucmp(ret, &(mont->N)) >= 0) {
+ if (!BN_usub(ret, ret, &(mont->N)))
+ goto err;
+ }
+ retn = 1;
+ bn_check_top(ret);
+ err:
+ BN_CTX_end(ctx);
+#endif /* MONT_WORD */
+ return retn;
+}
+
+int bn_to_mont_fixed_top(BIGNUM *r, const BIGNUM *a, BN_MONT_CTX *mont,
+ BN_CTX *ctx)
+{
+ return bn_mul_mont_fixed_top(r, a, &(mont->RR), mont, ctx);
+}
+
+BN_MONT_CTX *BN_MONT_CTX_new(void)
+{
+ BN_MONT_CTX *ret;
+
+ if ((ret = OPENSSL_malloc(sizeof(*ret))) == NULL) {
+ BNerr(BN_F_BN_MONT_CTX_NEW, ERR_R_MALLOC_FAILURE);
+ return NULL;
+ }
+
+ BN_MONT_CTX_init(ret);
+ ret->flags = BN_FLG_MALLOCED;
+ return ret;
+}
+
+void BN_MONT_CTX_init(BN_MONT_CTX *ctx)
+{
+ ctx->ri = 0;
+ bn_init(&ctx->RR);
+ bn_init(&ctx->N);
+ bn_init(&ctx->Ni);
+ ctx->n0[0] = ctx->n0[1] = 0;
+ ctx->flags = 0;
+}
+
+void BN_MONT_CTX_free(BN_MONT_CTX *mont)
+{
+ if (mont == NULL)
+ return;
+ BN_clear_free(&mont->RR);
+ BN_clear_free(&mont->N);
+ BN_clear_free(&mont->Ni);
+ if (mont->flags & BN_FLG_MALLOCED)
+ OPENSSL_free(mont);
+}
+
+int BN_MONT_CTX_set(BN_MONT_CTX *mont, const BIGNUM *mod, BN_CTX *ctx)
+{
+ int i, ret = 0;
+ BIGNUM *Ri, *R;
+
+ if (BN_is_zero(mod))
+ return 0;
+
+ BN_CTX_start(ctx);
+ if ((Ri = BN_CTX_get(ctx)) == NULL)
+ goto err;
+ R = &(mont->RR); /* grab RR as a temp */
+ if (!BN_copy(&(mont->N), mod))
+ goto err; /* Set N */
+ if (BN_get_flags(mod, BN_FLG_CONSTTIME) != 0)
+ BN_set_flags(&(mont->N), BN_FLG_CONSTTIME);
+ mont->N.neg = 0;
+
+#ifdef MONT_WORD
+ {
+ BIGNUM tmod;
+ BN_ULONG buf[2];
+
+ bn_init(&tmod);
+ tmod.d = buf;
+ tmod.dmax = 2;
+ tmod.neg = 0;
+
+ if (BN_get_flags(mod, BN_FLG_CONSTTIME) != 0)
+ BN_set_flags(&tmod, BN_FLG_CONSTTIME);
+
+ mont->ri = (BN_num_bits(mod) + (BN_BITS2 - 1)) / BN_BITS2 * BN_BITS2;
+
+# if defined(OPENSSL_BN_ASM_MONT) && (BN_BITS2<=32)
+ /*
+ * Only certain BN_BITS2<=32 platforms actually make use of n0[1],
+ * and we could use the #else case (with a shorter R value) for the
+ * others. However, currently only the assembler files do know which
+ * is which.
+ */
+
+ BN_zero(R);
+ if (!(BN_set_bit(R, 2 * BN_BITS2)))
+ goto err;
+
+ tmod.top = 0;
+ if ((buf[0] = mod->d[0]))
+ tmod.top = 1;
+ if ((buf[1] = mod->top > 1 ? mod->d[1] : 0))
+ tmod.top = 2;
+
+ if (BN_is_one(&tmod))
+ BN_zero(Ri);
+ else if ((BN_mod_inverse(Ri, R, &tmod, ctx)) == NULL)
+ goto err;
+ if (!BN_lshift(Ri, Ri, 2 * BN_BITS2))
+ goto err; /* R*Ri */
+ if (!BN_is_zero(Ri)) {
+ if (!BN_sub_word(Ri, 1))
+ goto err;
+ } else { /* if N mod word size == 1 */
+
+ if (bn_expand(Ri, (int)sizeof(BN_ULONG) * 2) == NULL)
+ goto err;
+ /* Ri-- (mod double word size) */
+ Ri->neg = 0;
+ Ri->d[0] = BN_MASK2;
+ Ri->d[1] = BN_MASK2;
+ Ri->top = 2;
+ }
+ if (!BN_div(Ri, NULL, Ri, &tmod, ctx))
+ goto err;
+ /*
+ * Ni = (R*Ri-1)/N, keep only couple of least significant words:
+ */
+ mont->n0[0] = (Ri->top > 0) ? Ri->d[0] : 0;
+ mont->n0[1] = (Ri->top > 1) ? Ri->d[1] : 0;
+# else
+ BN_zero(R);
+ if (!(BN_set_bit(R, BN_BITS2)))
+ goto err; /* R */
+
+ buf[0] = mod->d[0]; /* tmod = N mod word size */
+ buf[1] = 0;
+ tmod.top = buf[0] != 0 ? 1 : 0;
+ /* Ri = R^-1 mod N */
+ if (BN_is_one(&tmod))
+ BN_zero(Ri);
+ else if ((BN_mod_inverse(Ri, R, &tmod, ctx)) == NULL)
+ goto err;
+ if (!BN_lshift(Ri, Ri, BN_BITS2))
+ goto err; /* R*Ri */
+ if (!BN_is_zero(Ri)) {
+ if (!BN_sub_word(Ri, 1))
+ goto err;
+ } else { /* if N mod word size == 1 */
+
+ if (!BN_set_word(Ri, BN_MASK2))
+ goto err; /* Ri-- (mod word size) */
+ }
+ if (!BN_div(Ri, NULL, Ri, &tmod, ctx))
+ goto err;
+ /*
+ * Ni = (R*Ri-1)/N, keep only least significant word:
+ */
+ mont->n0[0] = (Ri->top > 0) ? Ri->d[0] : 0;
+ mont->n0[1] = 0;
+# endif
+ }
+#else /* !MONT_WORD */
+ { /* bignum version */
+ mont->ri = BN_num_bits(&mont->N);
+ BN_zero(R);
+ if (!BN_set_bit(R, mont->ri))
+ goto err; /* R = 2^ri */
+ /* Ri = R^-1 mod N */
+ if ((BN_mod_inverse(Ri, R, &mont->N, ctx)) == NULL)
+ goto err;
+ if (!BN_lshift(Ri, Ri, mont->ri))
+ goto err; /* R*Ri */
+ if (!BN_sub_word(Ri, 1))
+ goto err;
+ /*
+ * Ni = (R*Ri-1) / N
+ */
+ if (!BN_div(&(mont->Ni), NULL, Ri, &mont->N, ctx))
+ goto err;
+ }
+#endif
+
+ /* setup RR for conversions */
+ BN_zero(&(mont->RR));
+ if (!BN_set_bit(&(mont->RR), mont->ri * 2))
+ goto err;
+ if (!BN_mod(&(mont->RR), &(mont->RR), &(mont->N), ctx))
+ goto err;
+
+ for (i = mont->RR.top, ret = mont->N.top; i < ret; i++)
+ mont->RR.d[i] = 0;
+ mont->RR.top = ret;
+ mont->RR.flags |= BN_FLG_FIXED_TOP;
+
+ ret = 1;
+ err:
+ BN_CTX_end(ctx);
+ return ret;
+}
+
+BN_MONT_CTX *BN_MONT_CTX_copy(BN_MONT_CTX *to, BN_MONT_CTX *from)
+{
+ if (to == from)
+ return to;
+
+ if (!BN_copy(&(to->RR), &(from->RR)))
+ return NULL;
+ if (!BN_copy(&(to->N), &(from->N)))
+ return NULL;
+ if (!BN_copy(&(to->Ni), &(from->Ni)))
+ return NULL;
+ to->ri = from->ri;
+ to->n0[0] = from->n0[0];
+ to->n0[1] = from->n0[1];
+ return to;
+}
+
+BN_MONT_CTX *BN_MONT_CTX_set_locked(BN_MONT_CTX **pmont, CRYPTO_RWLOCK *lock,
+ const BIGNUM *mod, BN_CTX *ctx)
+{
+ BN_MONT_CTX *ret;
+
+ CRYPTO_THREAD_read_lock(lock);
+ ret = *pmont;
+ CRYPTO_THREAD_unlock(lock);
+ if (ret)
+ return ret;
+
+ /*
+ * We don't want to serialise globally while doing our lazy-init math in
+ * BN_MONT_CTX_set. That punishes threads that are doing independent
+ * things. Instead, punish the case where more than one thread tries to
+ * lazy-init the same 'pmont', by having each do the lazy-init math work
+ * independently and only use the one from the thread that wins the race
+ * (the losers throw away the work they've done).
+ */
+ ret = BN_MONT_CTX_new();
+ if (ret == NULL)
+ return NULL;
+ if (!BN_MONT_CTX_set(ret, mod, ctx)) {
+ BN_MONT_CTX_free(ret);
+ return NULL;
+ }
+
+ /* The locked compare-and-set, after the local work is done. */
+ CRYPTO_THREAD_write_lock(lock);
+ if (*pmont) {
+ BN_MONT_CTX_free(ret);
+ ret = *pmont;
+ } else
+ *pmont = ret;
+ CRYPTO_THREAD_unlock(lock);
+ return ret;
+}
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_mpi.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_mpi.c
new file mode 100644
index 0000000..0902da5
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_mpi.c
@@ -0,0 +1,86 @@
+/*
+ * Copyright 1995-2020 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include <stdio.h>
+#include "internal/cryptlib.h"
+#include "bn_local.h"
+
+int BN_bn2mpi(const BIGNUM *a, unsigned char *d)
+{
+ int bits;
+ int num = 0;
+ int ext = 0;
+ long l;
+
+ bits = BN_num_bits(a);
+ num = (bits + 7) / 8;
+ if (bits > 0) {
+ ext = ((bits & 0x07) == 0);
+ }
+ if (d == NULL)
+ return (num + 4 + ext);
+
+ l = num + ext;
+ d[0] = (unsigned char)(l >> 24) & 0xff;
+ d[1] = (unsigned char)(l >> 16) & 0xff;
+ d[2] = (unsigned char)(l >> 8) & 0xff;
+ d[3] = (unsigned char)(l) & 0xff;
+ if (ext)
+ d[4] = 0;
+ num = BN_bn2bin(a, &(d[4 + ext]));
+ if (a->neg)
+ d[4] |= 0x80;
+ return (num + 4 + ext);
+}
+
+BIGNUM *BN_mpi2bn(const unsigned char *d, int n, BIGNUM *ain)
+{
+ long len;
+ int neg = 0;
+ BIGNUM *a = NULL;
+
+ if (n < 4 || (d[0] & 0x80) != 0) {
+ BNerr(BN_F_BN_MPI2BN, BN_R_INVALID_LENGTH);
+ return NULL;
+ }
+ len = ((long)d[0] << 24) | ((long)d[1] << 16) | ((int)d[2] << 8) | (int)
+ d[3];
+ if ((len + 4) != n) {
+ BNerr(BN_F_BN_MPI2BN, BN_R_ENCODING_ERROR);
+ return NULL;
+ }
+
+ if (ain == NULL)
+ a = BN_new();
+ else
+ a = ain;
+
+ if (a == NULL)
+ return NULL;
+
+ if (len == 0) {
+ a->neg = 0;
+ a->top = 0;
+ return a;
+ }
+ d += 4;
+ if ((*d) & 0x80)
+ neg = 1;
+ if (BN_bin2bn(d, (int)len, a) == NULL) {
+ if (ain == NULL)
+ BN_free(a);
+ return NULL;
+ }
+ a->neg = neg;
+ if (neg) {
+ BN_clear_bit(a, BN_num_bits(a) - 1);
+ }
+ bn_check_top(a);
+ return a;
+}
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_mul.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_mul.c
new file mode 100644
index 0000000..6743e7b
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_mul.c
@@ -0,0 +1,684 @@
+/*
+ * Copyright 1995-2018 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include <assert.h>
+#include "internal/cryptlib.h"
+#include "bn_local.h"
+
+#if defined(OPENSSL_NO_ASM) || !defined(OPENSSL_BN_ASM_PART_WORDS)
+/*
+ * Here follows specialised variants of bn_add_words() and bn_sub_words().
+ * They have the property performing operations on arrays of different sizes.
+ * The sizes of those arrays is expressed through cl, which is the common
+ * length ( basically, min(len(a),len(b)) ), and dl, which is the delta
+ * between the two lengths, calculated as len(a)-len(b). All lengths are the
+ * number of BN_ULONGs... For the operations that require a result array as
+ * parameter, it must have the length cl+abs(dl). These functions should
+ * probably end up in bn_asm.c as soon as there are assembler counterparts
+ * for the systems that use assembler files.
+ */
+
+BN_ULONG bn_sub_part_words(BN_ULONG *r,
+ const BN_ULONG *a, const BN_ULONG *b,
+ int cl, int dl)
+{
+ BN_ULONG c, t;
+
+ assert(cl >= 0);
+ c = bn_sub_words(r, a, b, cl);
+
+ if (dl == 0)
+ return c;
+
+ r += cl;
+ a += cl;
+ b += cl;
+
+ if (dl < 0) {
+ for (;;) {
+ t = b[0];
+ r[0] = (0 - t - c) & BN_MASK2;
+ if (t != 0)
+ c = 1;
+ if (++dl >= 0)
+ break;
+
+ t = b[1];
+ r[1] = (0 - t - c) & BN_MASK2;
+ if (t != 0)
+ c = 1;
+ if (++dl >= 0)
+ break;
+
+ t = b[2];
+ r[2] = (0 - t - c) & BN_MASK2;
+ if (t != 0)
+ c = 1;
+ if (++dl >= 0)
+ break;
+
+ t = b[3];
+ r[3] = (0 - t - c) & BN_MASK2;
+ if (t != 0)
+ c = 1;
+ if (++dl >= 0)
+ break;
+
+ b += 4;
+ r += 4;
+ }
+ } else {
+ int save_dl = dl;
+ while (c) {
+ t = a[0];
+ r[0] = (t - c) & BN_MASK2;
+ if (t != 0)
+ c = 0;
+ if (--dl <= 0)
+ break;
+
+ t = a[1];
+ r[1] = (t - c) & BN_MASK2;
+ if (t != 0)
+ c = 0;
+ if (--dl <= 0)
+ break;
+
+ t = a[2];
+ r[2] = (t - c) & BN_MASK2;
+ if (t != 0)
+ c = 0;
+ if (--dl <= 0)
+ break;
+
+ t = a[3];
+ r[3] = (t - c) & BN_MASK2;
+ if (t != 0)
+ c = 0;
+ if (--dl <= 0)
+ break;
+
+ save_dl = dl;
+ a += 4;
+ r += 4;
+ }
+ if (dl > 0) {
+ if (save_dl > dl) {
+ switch (save_dl - dl) {
+ case 1:
+ r[1] = a[1];
+ if (--dl <= 0)
+ break;
+ /* fall thru */
+ case 2:
+ r[2] = a[2];
+ if (--dl <= 0)
+ break;
+ /* fall thru */
+ case 3:
+ r[3] = a[3];
+ if (--dl <= 0)
+ break;
+ }
+ a += 4;
+ r += 4;
+ }
+ }
+ if (dl > 0) {
+ for (;;) {
+ r[0] = a[0];
+ if (--dl <= 0)
+ break;
+ r[1] = a[1];
+ if (--dl <= 0)
+ break;
+ r[2] = a[2];
+ if (--dl <= 0)
+ break;
+ r[3] = a[3];
+ if (--dl <= 0)
+ break;
+
+ a += 4;
+ r += 4;
+ }
+ }
+ }
+ return c;
+}
+#endif
+
+#ifdef BN_RECURSION
+/*
+ * Karatsuba recursive multiplication algorithm (cf. Knuth, The Art of
+ * Computer Programming, Vol. 2)
+ */
+
+/*-
+ * r is 2*n2 words in size,
+ * a and b are both n2 words in size.
+ * n2 must be a power of 2.
+ * We multiply and return the result.
+ * t must be 2*n2 words in size
+ * We calculate
+ * a[0]*b[0]
+ * a[0]*b[0]+a[1]*b[1]+(a[0]-a[1])*(b[1]-b[0])
+ * a[1]*b[1]
+ */
+/* dnX may not be positive, but n2/2+dnX has to be */
+void bn_mul_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2,
+ int dna, int dnb, BN_ULONG *t)
+{
+ int n = n2 / 2, c1, c2;
+ int tna = n + dna, tnb = n + dnb;
+ unsigned int neg, zero;
+ BN_ULONG ln, lo, *p;
+
+# ifdef BN_MUL_COMBA
+# if 0
+ if (n2 == 4) {
+ bn_mul_comba4(r, a, b);
+ return;
+ }
+# endif
+ /*
+ * Only call bn_mul_comba 8 if n2 == 8 and the two arrays are complete
+ * [steve]
+ */
+ if (n2 == 8 && dna == 0 && dnb == 0) {
+ bn_mul_comba8(r, a, b);
+ return;
+ }
+# endif /* BN_MUL_COMBA */
+ /* Else do normal multiply */
+ if (n2 < BN_MUL_RECURSIVE_SIZE_NORMAL) {
+ bn_mul_normal(r, a, n2 + dna, b, n2 + dnb);
+ if ((dna + dnb) < 0)
+ memset(&r[2 * n2 + dna + dnb], 0,
+ sizeof(BN_ULONG) * -(dna + dnb));
+ return;
+ }
+ /* r=(a[0]-a[1])*(b[1]-b[0]) */
+ c1 = bn_cmp_part_words(a, &(a[n]), tna, n - tna);
+ c2 = bn_cmp_part_words(&(b[n]), b, tnb, tnb - n);
+ zero = neg = 0;
+ switch (c1 * 3 + c2) {
+ case -4:
+ bn_sub_part_words(t, &(a[n]), a, tna, tna - n); /* - */
+ bn_sub_part_words(&(t[n]), b, &(b[n]), tnb, n - tnb); /* - */
+ break;
+ case -3:
+ zero = 1;
+ break;
+ case -2:
+ bn_sub_part_words(t, &(a[n]), a, tna, tna - n); /* - */
+ bn_sub_part_words(&(t[n]), &(b[n]), b, tnb, tnb - n); /* + */
+ neg = 1;
+ break;
+ case -1:
+ case 0:
+ case 1:
+ zero = 1;
+ break;
+ case 2:
+ bn_sub_part_words(t, a, &(a[n]), tna, n - tna); /* + */
+ bn_sub_part_words(&(t[n]), b, &(b[n]), tnb, n - tnb); /* - */
+ neg = 1;
+ break;
+ case 3:
+ zero = 1;
+ break;
+ case 4:
+ bn_sub_part_words(t, a, &(a[n]), tna, n - tna);
+ bn_sub_part_words(&(t[n]), &(b[n]), b, tnb, tnb - n);
+ break;
+ }
+
+# ifdef BN_MUL_COMBA
+ if (n == 4 && dna == 0 && dnb == 0) { /* XXX: bn_mul_comba4 could take
+ * extra args to do this well */
+ if (!zero)
+ bn_mul_comba4(&(t[n2]), t, &(t[n]));
+ else
+ memset(&t[n2], 0, sizeof(*t) * 8);
+
+ bn_mul_comba4(r, a, b);
+ bn_mul_comba4(&(r[n2]), &(a[n]), &(b[n]));
+ } else if (n == 8 && dna == 0 && dnb == 0) { /* XXX: bn_mul_comba8 could
+ * take extra args to do
+ * this well */
+ if (!zero)
+ bn_mul_comba8(&(t[n2]), t, &(t[n]));
+ else
+ memset(&t[n2], 0, sizeof(*t) * 16);
+
+ bn_mul_comba8(r, a, b);
+ bn_mul_comba8(&(r[n2]), &(a[n]), &(b[n]));
+ } else
+# endif /* BN_MUL_COMBA */
+ {
+ p = &(t[n2 * 2]);
+ if (!zero)
+ bn_mul_recursive(&(t[n2]), t, &(t[n]), n, 0, 0, p);
+ else
+ memset(&t[n2], 0, sizeof(*t) * n2);
+ bn_mul_recursive(r, a, b, n, 0, 0, p);
+ bn_mul_recursive(&(r[n2]), &(a[n]), &(b[n]), n, dna, dnb, p);
+ }
+
+ /*-
+ * t[32] holds (a[0]-a[1])*(b[1]-b[0]), c1 is the sign
+ * r[10] holds (a[0]*b[0])
+ * r[32] holds (b[1]*b[1])
+ */
+
+ c1 = (int)(bn_add_words(t, r, &(r[n2]), n2));
+
+ if (neg) { /* if t[32] is negative */
+ c1 -= (int)(bn_sub_words(&(t[n2]), t, &(t[n2]), n2));
+ } else {
+ /* Might have a carry */
+ c1 += (int)(bn_add_words(&(t[n2]), &(t[n2]), t, n2));
+ }
+
+ /*-
+ * t[32] holds (a[0]-a[1])*(b[1]-b[0])+(a[0]*b[0])+(a[1]*b[1])
+ * r[10] holds (a[0]*b[0])
+ * r[32] holds (b[1]*b[1])
+ * c1 holds the carry bits
+ */
+ c1 += (int)(bn_add_words(&(r[n]), &(r[n]), &(t[n2]), n2));
+ if (c1) {
+ p = &(r[n + n2]);
+ lo = *p;
+ ln = (lo + c1) & BN_MASK2;
+ *p = ln;
+
+ /*
+ * The overflow will stop before we over write words we should not
+ * overwrite
+ */
+ if (ln < (BN_ULONG)c1) {
+ do {
+ p++;
+ lo = *p;
+ ln = (lo + 1) & BN_MASK2;
+ *p = ln;
+ } while (ln == 0);
+ }
+ }
+}
+
+/*
+ * n+tn is the word length t needs to be n*4 is size, as does r
+ */
+/* tnX may not be negative but less than n */
+void bn_mul_part_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n,
+ int tna, int tnb, BN_ULONG *t)
+{
+ int i, j, n2 = n * 2;
+ int c1, c2, neg;
+ BN_ULONG ln, lo, *p;
+
+ if (n < 8) {
+ bn_mul_normal(r, a, n + tna, b, n + tnb);
+ return;
+ }
+
+ /* r=(a[0]-a[1])*(b[1]-b[0]) */
+ c1 = bn_cmp_part_words(a, &(a[n]), tna, n - tna);
+ c2 = bn_cmp_part_words(&(b[n]), b, tnb, tnb - n);
+ neg = 0;
+ switch (c1 * 3 + c2) {
+ case -4:
+ bn_sub_part_words(t, &(a[n]), a, tna, tna - n); /* - */
+ bn_sub_part_words(&(t[n]), b, &(b[n]), tnb, n - tnb); /* - */
+ break;
+ case -3:
+ case -2:
+ bn_sub_part_words(t, &(a[n]), a, tna, tna - n); /* - */
+ bn_sub_part_words(&(t[n]), &(b[n]), b, tnb, tnb - n); /* + */
+ neg = 1;
+ break;
+ case -1:
+ case 0:
+ case 1:
+ case 2:
+ bn_sub_part_words(t, a, &(a[n]), tna, n - tna); /* + */
+ bn_sub_part_words(&(t[n]), b, &(b[n]), tnb, n - tnb); /* - */
+ neg = 1;
+ break;
+ case 3:
+ case 4:
+ bn_sub_part_words(t, a, &(a[n]), tna, n - tna);
+ bn_sub_part_words(&(t[n]), &(b[n]), b, tnb, tnb - n);
+ break;
+ }
+ /*
+ * The zero case isn't yet implemented here. The speedup would probably
+ * be negligible.
+ */
+# if 0
+ if (n == 4) {
+ bn_mul_comba4(&(t[n2]), t, &(t[n]));
+ bn_mul_comba4(r, a, b);
+ bn_mul_normal(&(r[n2]), &(a[n]), tn, &(b[n]), tn);
+ memset(&r[n2 + tn * 2], 0, sizeof(*r) * (n2 - tn * 2));
+ } else
+# endif
+ if (n == 8) {
+ bn_mul_comba8(&(t[n2]), t, &(t[n]));
+ bn_mul_comba8(r, a, b);
+ bn_mul_normal(&(r[n2]), &(a[n]), tna, &(b[n]), tnb);
+ memset(&r[n2 + tna + tnb], 0, sizeof(*r) * (n2 - tna - tnb));
+ } else {
+ p = &(t[n2 * 2]);
+ bn_mul_recursive(&(t[n2]), t, &(t[n]), n, 0, 0, p);
+ bn_mul_recursive(r, a, b, n, 0, 0, p);
+ i = n / 2;
+ /*
+ * If there is only a bottom half to the number, just do it
+ */
+ if (tna > tnb)
+ j = tna - i;
+ else
+ j = tnb - i;
+ if (j == 0) {
+ bn_mul_recursive(&(r[n2]), &(a[n]), &(b[n]),
+ i, tna - i, tnb - i, p);
+ memset(&r[n2 + i * 2], 0, sizeof(*r) * (n2 - i * 2));
+ } else if (j > 0) { /* eg, n == 16, i == 8 and tn == 11 */
+ bn_mul_part_recursive(&(r[n2]), &(a[n]), &(b[n]),
+ i, tna - i, tnb - i, p);
+ memset(&(r[n2 + tna + tnb]), 0,
+ sizeof(BN_ULONG) * (n2 - tna - tnb));
+ } else { /* (j < 0) eg, n == 16, i == 8 and tn == 5 */
+
+ memset(&r[n2], 0, sizeof(*r) * n2);
+ if (tna < BN_MUL_RECURSIVE_SIZE_NORMAL
+ && tnb < BN_MUL_RECURSIVE_SIZE_NORMAL) {
+ bn_mul_normal(&(r[n2]), &(a[n]), tna, &(b[n]), tnb);
+ } else {
+ for (;;) {
+ i /= 2;
+ /*
+ * these simplified conditions work exclusively because
+ * difference between tna and tnb is 1 or 0
+ */
+ if (i < tna || i < tnb) {
+ bn_mul_part_recursive(&(r[n2]),
+ &(a[n]), &(b[n]),
+ i, tna - i, tnb - i, p);
+ break;
+ } else if (i == tna || i == tnb) {
+ bn_mul_recursive(&(r[n2]),
+ &(a[n]), &(b[n]),
+ i, tna - i, tnb - i, p);
+ break;
+ }
+ }
+ }
+ }
+ }
+
+ /*-
+ * t[32] holds (a[0]-a[1])*(b[1]-b[0]), c1 is the sign
+ * r[10] holds (a[0]*b[0])
+ * r[32] holds (b[1]*b[1])
+ */
+
+ c1 = (int)(bn_add_words(t, r, &(r[n2]), n2));
+
+ if (neg) { /* if t[32] is negative */
+ c1 -= (int)(bn_sub_words(&(t[n2]), t, &(t[n2]), n2));
+ } else {
+ /* Might have a carry */
+ c1 += (int)(bn_add_words(&(t[n2]), &(t[n2]), t, n2));
+ }
+
+ /*-
+ * t[32] holds (a[0]-a[1])*(b[1]-b[0])+(a[0]*b[0])+(a[1]*b[1])
+ * r[10] holds (a[0]*b[0])
+ * r[32] holds (b[1]*b[1])
+ * c1 holds the carry bits
+ */
+ c1 += (int)(bn_add_words(&(r[n]), &(r[n]), &(t[n2]), n2));
+ if (c1) {
+ p = &(r[n + n2]);
+ lo = *p;
+ ln = (lo + c1) & BN_MASK2;
+ *p = ln;
+
+ /*
+ * The overflow will stop before we over write words we should not
+ * overwrite
+ */
+ if (ln < (BN_ULONG)c1) {
+ do {
+ p++;
+ lo = *p;
+ ln = (lo + 1) & BN_MASK2;
+ *p = ln;
+ } while (ln == 0);
+ }
+ }
+}
+
+/*-
+ * a and b must be the same size, which is n2.
+ * r needs to be n2 words and t needs to be n2*2
+ */
+void bn_mul_low_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2,
+ BN_ULONG *t)
+{
+ int n = n2 / 2;
+
+ bn_mul_recursive(r, a, b, n, 0, 0, &(t[0]));
+ if (n >= BN_MUL_LOW_RECURSIVE_SIZE_NORMAL) {
+ bn_mul_low_recursive(&(t[0]), &(a[0]), &(b[n]), n, &(t[n2]));
+ bn_add_words(&(r[n]), &(r[n]), &(t[0]), n);
+ bn_mul_low_recursive(&(t[0]), &(a[n]), &(b[0]), n, &(t[n2]));
+ bn_add_words(&(r[n]), &(r[n]), &(t[0]), n);
+ } else {
+ bn_mul_low_normal(&(t[0]), &(a[0]), &(b[n]), n);
+ bn_mul_low_normal(&(t[n]), &(a[n]), &(b[0]), n);
+ bn_add_words(&(r[n]), &(r[n]), &(t[0]), n);
+ bn_add_words(&(r[n]), &(r[n]), &(t[n]), n);
+ }
+}
+#endif /* BN_RECURSION */
+
+int BN_mul(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, BN_CTX *ctx)
+{
+ int ret = bn_mul_fixed_top(r, a, b, ctx);
+
+ bn_correct_top(r);
+ bn_check_top(r);
+
+ return ret;
+}
+
+int bn_mul_fixed_top(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, BN_CTX *ctx)
+{
+ int ret = 0;
+ int top, al, bl;
+ BIGNUM *rr;
+#if defined(BN_MUL_COMBA) || defined(BN_RECURSION)
+ int i;
+#endif
+#ifdef BN_RECURSION
+ BIGNUM *t = NULL;
+ int j = 0, k;
+#endif
+
+ bn_check_top(a);
+ bn_check_top(b);
+ bn_check_top(r);
+
+ al = a->top;
+ bl = b->top;
+
+ if ((al == 0) || (bl == 0)) {
+ BN_zero(r);
+ return 1;
+ }
+ top = al + bl;
+
+ BN_CTX_start(ctx);
+ if ((r == a) || (r == b)) {
+ if ((rr = BN_CTX_get(ctx)) == NULL)
+ goto err;
+ } else
+ rr = r;
+
+#if defined(BN_MUL_COMBA) || defined(BN_RECURSION)
+ i = al - bl;
+#endif
+#ifdef BN_MUL_COMBA
+ if (i == 0) {
+# if 0
+ if (al == 4) {
+ if (bn_wexpand(rr, 8) == NULL)
+ goto err;
+ rr->top = 8;
+ bn_mul_comba4(rr->d, a->d, b->d);
+ goto end;
+ }
+# endif
+ if (al == 8) {
+ if (bn_wexpand(rr, 16) == NULL)
+ goto err;
+ rr->top = 16;
+ bn_mul_comba8(rr->d, a->d, b->d);
+ goto end;
+ }
+ }
+#endif /* BN_MUL_COMBA */
+#ifdef BN_RECURSION
+ if ((al >= BN_MULL_SIZE_NORMAL) && (bl >= BN_MULL_SIZE_NORMAL)) {
+ if (i >= -1 && i <= 1) {
+ /*
+ * Find out the power of two lower or equal to the longest of the
+ * two numbers
+ */
+ if (i >= 0) {
+ j = BN_num_bits_word((BN_ULONG)al);
+ }
+ if (i == -1) {
+ j = BN_num_bits_word((BN_ULONG)bl);
+ }
+ j = 1 << (j - 1);
+ assert(j <= al || j <= bl);
+ k = j + j;
+ t = BN_CTX_get(ctx);
+ if (t == NULL)
+ goto err;
+ if (al > j || bl > j) {
+ if (bn_wexpand(t, k * 4) == NULL)
+ goto err;
+ if (bn_wexpand(rr, k * 4) == NULL)
+ goto err;
+ bn_mul_part_recursive(rr->d, a->d, b->d,
+ j, al - j, bl - j, t->d);
+ } else { /* al <= j || bl <= j */
+
+ if (bn_wexpand(t, k * 2) == NULL)
+ goto err;
+ if (bn_wexpand(rr, k * 2) == NULL)
+ goto err;
+ bn_mul_recursive(rr->d, a->d, b->d, j, al - j, bl - j, t->d);
+ }
+ rr->top = top;
+ goto end;
+ }
+ }
+#endif /* BN_RECURSION */
+ if (bn_wexpand(rr, top) == NULL)
+ goto err;
+ rr->top = top;
+ bn_mul_normal(rr->d, a->d, al, b->d, bl);
+
+#if defined(BN_MUL_COMBA) || defined(BN_RECURSION)
+ end:
+#endif
+ rr->neg = a->neg ^ b->neg;
+ rr->flags |= BN_FLG_FIXED_TOP;
+ if (r != rr && BN_copy(r, rr) == NULL)
+ goto err;
+
+ ret = 1;
+ err:
+ bn_check_top(r);
+ BN_CTX_end(ctx);
+ return ret;
+}
+
+void bn_mul_normal(BN_ULONG *r, BN_ULONG *a, int na, BN_ULONG *b, int nb)
+{
+ BN_ULONG *rr;
+
+ if (na < nb) {
+ int itmp;
+ BN_ULONG *ltmp;
+
+ itmp = na;
+ na = nb;
+ nb = itmp;
+ ltmp = a;
+ a = b;
+ b = ltmp;
+
+ }
+ rr = &(r[na]);
+ if (nb <= 0) {
+ (void)bn_mul_words(r, a, na, 0);
+ return;
+ } else
+ rr[0] = bn_mul_words(r, a, na, b[0]);
+
+ for (;;) {
+ if (--nb <= 0)
+ return;
+ rr[1] = bn_mul_add_words(&(r[1]), a, na, b[1]);
+ if (--nb <= 0)
+ return;
+ rr[2] = bn_mul_add_words(&(r[2]), a, na, b[2]);
+ if (--nb <= 0)
+ return;
+ rr[3] = bn_mul_add_words(&(r[3]), a, na, b[3]);
+ if (--nb <= 0)
+ return;
+ rr[4] = bn_mul_add_words(&(r[4]), a, na, b[4]);
+ rr += 4;
+ r += 4;
+ b += 4;
+ }
+}
+
+void bn_mul_low_normal(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
+{
+ bn_mul_words(r, a, n, b[0]);
+
+ for (;;) {
+ if (--n <= 0)
+ return;
+ bn_mul_add_words(&(r[1]), a, n, b[1]);
+ if (--n <= 0)
+ return;
+ bn_mul_add_words(&(r[2]), a, n, b[2]);
+ if (--n <= 0)
+ return;
+ bn_mul_add_words(&(r[3]), a, n, b[3]);
+ if (--n <= 0)
+ return;
+ bn_mul_add_words(&(r[4]), a, n, b[4]);
+ r += 4;
+ b += 4;
+ }
+}
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_nist.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_nist.c
new file mode 100644
index 0000000..325dc22
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_nist.c
@@ -0,0 +1,1239 @@
+/*
+ * Copyright 2002-2018 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include "bn_local.h"
+#include "internal/cryptlib.h"
+
+#define BN_NIST_192_TOP (192+BN_BITS2-1)/BN_BITS2
+#define BN_NIST_224_TOP (224+BN_BITS2-1)/BN_BITS2
+#define BN_NIST_256_TOP (256+BN_BITS2-1)/BN_BITS2
+#define BN_NIST_384_TOP (384+BN_BITS2-1)/BN_BITS2
+#define BN_NIST_521_TOP (521+BN_BITS2-1)/BN_BITS2
+
+/* pre-computed tables are "carry-less" values of modulus*(i+1) */
+#if BN_BITS2 == 64
+static const BN_ULONG _nist_p_192[][BN_NIST_192_TOP] = {
+ {0xFFFFFFFFFFFFFFFFULL, 0xFFFFFFFFFFFFFFFEULL, 0xFFFFFFFFFFFFFFFFULL},
+ {0xFFFFFFFFFFFFFFFEULL, 0xFFFFFFFFFFFFFFFDULL, 0xFFFFFFFFFFFFFFFFULL},
+ {0xFFFFFFFFFFFFFFFDULL, 0xFFFFFFFFFFFFFFFCULL, 0xFFFFFFFFFFFFFFFFULL}
+};
+
+static const BN_ULONG _nist_p_192_sqr[] = {
+ 0x0000000000000001ULL, 0x0000000000000002ULL, 0x0000000000000001ULL,
+ 0xFFFFFFFFFFFFFFFEULL, 0xFFFFFFFFFFFFFFFDULL, 0xFFFFFFFFFFFFFFFFULL
+};
+
+static const BN_ULONG _nist_p_224[][BN_NIST_224_TOP] = {
+ {0x0000000000000001ULL, 0xFFFFFFFF00000000ULL,
+ 0xFFFFFFFFFFFFFFFFULL, 0x00000000FFFFFFFFULL},
+ {0x0000000000000002ULL, 0xFFFFFFFE00000000ULL,
+ 0xFFFFFFFFFFFFFFFFULL, 0x00000001FFFFFFFFULL} /* this one is
+ * "carry-full" */
+};
+
+static const BN_ULONG _nist_p_224_sqr[] = {
+ 0x0000000000000001ULL, 0xFFFFFFFE00000000ULL,
+ 0xFFFFFFFFFFFFFFFFULL, 0x0000000200000000ULL,
+ 0x0000000000000000ULL, 0xFFFFFFFFFFFFFFFEULL,
+ 0xFFFFFFFFFFFFFFFFULL
+};
+
+static const BN_ULONG _nist_p_256[][BN_NIST_256_TOP] = {
+ {0xFFFFFFFFFFFFFFFFULL, 0x00000000FFFFFFFFULL,
+ 0x0000000000000000ULL, 0xFFFFFFFF00000001ULL},
+ {0xFFFFFFFFFFFFFFFEULL, 0x00000001FFFFFFFFULL,
+ 0x0000000000000000ULL, 0xFFFFFFFE00000002ULL},
+ {0xFFFFFFFFFFFFFFFDULL, 0x00000002FFFFFFFFULL,
+ 0x0000000000000000ULL, 0xFFFFFFFD00000003ULL},
+ {0xFFFFFFFFFFFFFFFCULL, 0x00000003FFFFFFFFULL,
+ 0x0000000000000000ULL, 0xFFFFFFFC00000004ULL},
+ {0xFFFFFFFFFFFFFFFBULL, 0x00000004FFFFFFFFULL,
+ 0x0000000000000000ULL, 0xFFFFFFFB00000005ULL},
+};
+
+static const BN_ULONG _nist_p_256_sqr[] = {
+ 0x0000000000000001ULL, 0xFFFFFFFE00000000ULL,
+ 0xFFFFFFFFFFFFFFFFULL, 0x00000001FFFFFFFEULL,
+ 0x00000001FFFFFFFEULL, 0x00000001FFFFFFFEULL,
+ 0xFFFFFFFE00000001ULL, 0xFFFFFFFE00000002ULL
+};
+
+static const BN_ULONG _nist_p_384[][BN_NIST_384_TOP] = {
+ {0x00000000FFFFFFFFULL, 0xFFFFFFFF00000000ULL, 0xFFFFFFFFFFFFFFFEULL,
+ 0xFFFFFFFFFFFFFFFFULL, 0xFFFFFFFFFFFFFFFFULL, 0xFFFFFFFFFFFFFFFFULL},
+ {0x00000001FFFFFFFEULL, 0xFFFFFFFE00000000ULL, 0xFFFFFFFFFFFFFFFDULL,
+ 0xFFFFFFFFFFFFFFFFULL, 0xFFFFFFFFFFFFFFFFULL, 0xFFFFFFFFFFFFFFFFULL},
+ {0x00000002FFFFFFFDULL, 0xFFFFFFFD00000000ULL, 0xFFFFFFFFFFFFFFFCULL,
+ 0xFFFFFFFFFFFFFFFFULL, 0xFFFFFFFFFFFFFFFFULL, 0xFFFFFFFFFFFFFFFFULL},
+ {0x00000003FFFFFFFCULL, 0xFFFFFFFC00000000ULL, 0xFFFFFFFFFFFFFFFBULL,
+ 0xFFFFFFFFFFFFFFFFULL, 0xFFFFFFFFFFFFFFFFULL, 0xFFFFFFFFFFFFFFFFULL},
+ {0x00000004FFFFFFFBULL, 0xFFFFFFFB00000000ULL, 0xFFFFFFFFFFFFFFFAULL,
+ 0xFFFFFFFFFFFFFFFFULL, 0xFFFFFFFFFFFFFFFFULL, 0xFFFFFFFFFFFFFFFFULL},
+};
+
+static const BN_ULONG _nist_p_384_sqr[] = {
+ 0xFFFFFFFE00000001ULL, 0x0000000200000000ULL, 0xFFFFFFFE00000000ULL,
+ 0x0000000200000000ULL, 0x0000000000000001ULL, 0x0000000000000000ULL,
+ 0x00000001FFFFFFFEULL, 0xFFFFFFFE00000000ULL, 0xFFFFFFFFFFFFFFFDULL,
+ 0xFFFFFFFFFFFFFFFFULL, 0xFFFFFFFFFFFFFFFFULL, 0xFFFFFFFFFFFFFFFFULL
+};
+
+static const BN_ULONG _nist_p_521[] =
+ { 0xFFFFFFFFFFFFFFFFULL, 0xFFFFFFFFFFFFFFFFULL,
+ 0xFFFFFFFFFFFFFFFFULL, 0xFFFFFFFFFFFFFFFFULL,
+ 0xFFFFFFFFFFFFFFFFULL, 0xFFFFFFFFFFFFFFFFULL,
+ 0xFFFFFFFFFFFFFFFFULL, 0xFFFFFFFFFFFFFFFFULL,
+ 0x00000000000001FFULL
+};
+
+static const BN_ULONG _nist_p_521_sqr[] = {
+ 0x0000000000000001ULL, 0x0000000000000000ULL, 0x0000000000000000ULL,
+ 0x0000000000000000ULL, 0x0000000000000000ULL, 0x0000000000000000ULL,
+ 0x0000000000000000ULL, 0x0000000000000000ULL, 0xFFFFFFFFFFFFFC00ULL,
+ 0xFFFFFFFFFFFFFFFFULL, 0xFFFFFFFFFFFFFFFFULL, 0xFFFFFFFFFFFFFFFFULL,
+ 0xFFFFFFFFFFFFFFFFULL, 0xFFFFFFFFFFFFFFFFULL, 0xFFFFFFFFFFFFFFFFULL,
+ 0xFFFFFFFFFFFFFFFFULL, 0x000000000003FFFFULL
+};
+#elif BN_BITS2 == 32
+static const BN_ULONG _nist_p_192[][BN_NIST_192_TOP] = {
+ {0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFE, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF},
+ {0xFFFFFFFE, 0xFFFFFFFF, 0xFFFFFFFD, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF},
+ {0xFFFFFFFD, 0xFFFFFFFF, 0xFFFFFFFC, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF}
+};
+
+static const BN_ULONG _nist_p_192_sqr[] = {
+ 0x00000001, 0x00000000, 0x00000002, 0x00000000, 0x00000001, 0x00000000,
+ 0xFFFFFFFE, 0xFFFFFFFF, 0xFFFFFFFD, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF
+};
+
+static const BN_ULONG _nist_p_224[][BN_NIST_224_TOP] = {
+ {0x00000001, 0x00000000, 0x00000000, 0xFFFFFFFF,
+ 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF},
+ {0x00000002, 0x00000000, 0x00000000, 0xFFFFFFFE,
+ 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF}
+};
+
+static const BN_ULONG _nist_p_224_sqr[] = {
+ 0x00000001, 0x00000000, 0x00000000, 0xFFFFFFFE,
+ 0xFFFFFFFF, 0xFFFFFFFF, 0x00000000, 0x00000002,
+ 0x00000000, 0x00000000, 0xFFFFFFFE, 0xFFFFFFFF,
+ 0xFFFFFFFF, 0xFFFFFFFF
+};
+
+static const BN_ULONG _nist_p_256[][BN_NIST_256_TOP] = {
+ {0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000001, 0xFFFFFFFF},
+ {0xFFFFFFFE, 0xFFFFFFFF, 0xFFFFFFFF, 0x00000001,
+ 0x00000000, 0x00000000, 0x00000002, 0xFFFFFFFE},
+ {0xFFFFFFFD, 0xFFFFFFFF, 0xFFFFFFFF, 0x00000002,
+ 0x00000000, 0x00000000, 0x00000003, 0xFFFFFFFD},
+ {0xFFFFFFFC, 0xFFFFFFFF, 0xFFFFFFFF, 0x00000003,
+ 0x00000000, 0x00000000, 0x00000004, 0xFFFFFFFC},
+ {0xFFFFFFFB, 0xFFFFFFFF, 0xFFFFFFFF, 0x00000004,
+ 0x00000000, 0x00000000, 0x00000005, 0xFFFFFFFB},
+};
+
+static const BN_ULONG _nist_p_256_sqr[] = {
+ 0x00000001, 0x00000000, 0x00000000, 0xFFFFFFFE,
+ 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFE, 0x00000001,
+ 0xFFFFFFFE, 0x00000001, 0xFFFFFFFE, 0x00000001,
+ 0x00000001, 0xFFFFFFFE, 0x00000002, 0xFFFFFFFE
+};
+
+static const BN_ULONG _nist_p_384[][BN_NIST_384_TOP] = {
+ {0xFFFFFFFF, 0x00000000, 0x00000000, 0xFFFFFFFF, 0xFFFFFFFE, 0xFFFFFFFF,
+ 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF},
+ {0xFFFFFFFE, 0x00000001, 0x00000000, 0xFFFFFFFE, 0xFFFFFFFD, 0xFFFFFFFF,
+ 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF},
+ {0xFFFFFFFD, 0x00000002, 0x00000000, 0xFFFFFFFD, 0xFFFFFFFC, 0xFFFFFFFF,
+ 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF},
+ {0xFFFFFFFC, 0x00000003, 0x00000000, 0xFFFFFFFC, 0xFFFFFFFB, 0xFFFFFFFF,
+ 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF},
+ {0xFFFFFFFB, 0x00000004, 0x00000000, 0xFFFFFFFB, 0xFFFFFFFA, 0xFFFFFFFF,
+ 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF},
+};
+
+static const BN_ULONG _nist_p_384_sqr[] = {
+ 0x00000001, 0xFFFFFFFE, 0x00000000, 0x00000002, 0x00000000, 0xFFFFFFFE,
+ 0x00000000, 0x00000002, 0x00000001, 0x00000000, 0x00000000, 0x00000000,
+ 0xFFFFFFFE, 0x00000001, 0x00000000, 0xFFFFFFFE, 0xFFFFFFFD, 0xFFFFFFFF,
+ 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF
+};
+
+static const BN_ULONG _nist_p_521[] = { 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF,
+ 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF,
+ 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF,
+ 0xFFFFFFFF, 0x000001FF
+};
+
+static const BN_ULONG _nist_p_521_sqr[] = {
+ 0x00000001, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0xFFFFFC00, 0xFFFFFFFF,
+ 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF,
+ 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF,
+ 0xFFFFFFFF, 0xFFFFFFFF, 0x0003FFFF
+};
+#else
+# error "unsupported BN_BITS2"
+#endif
+
+static const BIGNUM _bignum_nist_p_192 = {
+ (BN_ULONG *)_nist_p_192[0],
+ BN_NIST_192_TOP,
+ BN_NIST_192_TOP,
+ 0,
+ BN_FLG_STATIC_DATA
+};
+
+static const BIGNUM _bignum_nist_p_224 = {
+ (BN_ULONG *)_nist_p_224[0],
+ BN_NIST_224_TOP,
+ BN_NIST_224_TOP,
+ 0,
+ BN_FLG_STATIC_DATA
+};
+
+static const BIGNUM _bignum_nist_p_256 = {
+ (BN_ULONG *)_nist_p_256[0],
+ BN_NIST_256_TOP,
+ BN_NIST_256_TOP,
+ 0,
+ BN_FLG_STATIC_DATA
+};
+
+static const BIGNUM _bignum_nist_p_384 = {
+ (BN_ULONG *)_nist_p_384[0],
+ BN_NIST_384_TOP,
+ BN_NIST_384_TOP,
+ 0,
+ BN_FLG_STATIC_DATA
+};
+
+static const BIGNUM _bignum_nist_p_521 = {
+ (BN_ULONG *)_nist_p_521,
+ BN_NIST_521_TOP,
+ BN_NIST_521_TOP,
+ 0,
+ BN_FLG_STATIC_DATA
+};
+
+const BIGNUM *BN_get0_nist_prime_192(void)
+{
+ return &_bignum_nist_p_192;
+}
+
+const BIGNUM *BN_get0_nist_prime_224(void)
+{
+ return &_bignum_nist_p_224;
+}
+
+const BIGNUM *BN_get0_nist_prime_256(void)
+{
+ return &_bignum_nist_p_256;
+}
+
+const BIGNUM *BN_get0_nist_prime_384(void)
+{
+ return &_bignum_nist_p_384;
+}
+
+const BIGNUM *BN_get0_nist_prime_521(void)
+{
+ return &_bignum_nist_p_521;
+}
+
+static void nist_cp_bn_0(BN_ULONG *dst, const BN_ULONG *src, int top, int max)
+{
+ int i;
+
+#ifdef BN_DEBUG
+ (void)ossl_assert(top <= max);
+#endif
+ for (i = 0; i < top; i++)
+ dst[i] = src[i];
+ for (; i < max; i++)
+ dst[i] = 0;
+}
+
+static void nist_cp_bn(BN_ULONG *dst, const BN_ULONG *src, int top)
+{
+ int i;
+
+ for (i = 0; i < top; i++)
+ dst[i] = src[i];
+}
+
+#if BN_BITS2 == 64
+# define bn_cp_64(to, n, from, m) (to)[n] = (m>=0)?((from)[m]):0;
+# define bn_64_set_0(to, n) (to)[n] = (BN_ULONG)0;
+/*
+ * two following macros are implemented under assumption that they
+ * are called in a sequence with *ascending* n, i.e. as they are...
+ */
+# define bn_cp_32_naked(to, n, from, m) (((n)&1)?(to[(n)/2]|=((m)&1)?(from[(m)/2]&BN_MASK2h):(from[(m)/2]<<32))\
+ :(to[(n)/2] =((m)&1)?(from[(m)/2]>>32):(from[(m)/2]&BN_MASK2l)))
+# define bn_32_set_0(to, n) (((n)&1)?(to[(n)/2]&=BN_MASK2l):(to[(n)/2]=0));
+# define bn_cp_32(to,n,from,m) ((m)>=0)?bn_cp_32_naked(to,n,from,m):bn_32_set_0(to,n)
+# if defined(L_ENDIAN)
+# if defined(__arch64__)
+# define NIST_INT64 long
+# else
+# define NIST_INT64 long long
+# endif
+# endif
+#else
+# define bn_cp_64(to, n, from, m) \
+ { \
+ bn_cp_32(to, (n)*2, from, (m)*2); \
+ bn_cp_32(to, (n)*2+1, from, (m)*2+1); \
+ }
+# define bn_64_set_0(to, n) \
+ { \
+ bn_32_set_0(to, (n)*2); \
+ bn_32_set_0(to, (n)*2+1); \
+ }
+# define bn_cp_32(to, n, from, m) (to)[n] = (m>=0)?((from)[m]):0;
+# define bn_32_set_0(to, n) (to)[n] = (BN_ULONG)0;
+# if defined(_WIN32) && !defined(__GNUC__)
+# define NIST_INT64 __int64
+# elif defined(BN_LLONG)
+# define NIST_INT64 long long
+# endif
+#endif /* BN_BITS2 != 64 */
+
+#define nist_set_192(to, from, a1, a2, a3) \
+ { \
+ bn_cp_64(to, 0, from, (a3) - 3) \
+ bn_cp_64(to, 1, from, (a2) - 3) \
+ bn_cp_64(to, 2, from, (a1) - 3) \
+ }
+
+int BN_nist_mod_192(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
+ BN_CTX *ctx)
+{
+ int top = a->top, i;
+ int carry;
+ register BN_ULONG *r_d, *a_d = a->d;
+ union {
+ BN_ULONG bn[BN_NIST_192_TOP];
+ unsigned int ui[BN_NIST_192_TOP * sizeof(BN_ULONG) /
+ sizeof(unsigned int)];
+ } buf;
+ BN_ULONG c_d[BN_NIST_192_TOP], *res;
+ PTR_SIZE_INT mask;
+ static const BIGNUM _bignum_nist_p_192_sqr = {
+ (BN_ULONG *)_nist_p_192_sqr,
+ OSSL_NELEM(_nist_p_192_sqr),
+ OSSL_NELEM(_nist_p_192_sqr),
+ 0, BN_FLG_STATIC_DATA
+ };
+
+ field = &_bignum_nist_p_192; /* just to make sure */
+
+ if (BN_is_negative(a) || BN_ucmp(a, &_bignum_nist_p_192_sqr) >= 0)
+ return BN_nnmod(r, a, field, ctx);
+
+ i = BN_ucmp(field, a);
+ if (i == 0) {
+ BN_zero(r);
+ return 1;
+ } else if (i > 0)
+ return (r == a) ? 1 : (BN_copy(r, a) != NULL);
+
+ if (r != a) {
+ if (!bn_wexpand(r, BN_NIST_192_TOP))
+ return 0;
+ r_d = r->d;
+ nist_cp_bn(r_d, a_d, BN_NIST_192_TOP);
+ } else
+ r_d = a_d;
+
+ nist_cp_bn_0(buf.bn, a_d + BN_NIST_192_TOP, top - BN_NIST_192_TOP,
+ BN_NIST_192_TOP);
+
+#if defined(NIST_INT64)
+ {
+ NIST_INT64 acc; /* accumulator */
+ unsigned int *rp = (unsigned int *)r_d;
+ const unsigned int *bp = (const unsigned int *)buf.ui;
+
+ acc = rp[0];
+ acc += bp[3 * 2 - 6];
+ acc += bp[5 * 2 - 6];
+ rp[0] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[1];
+ acc += bp[3 * 2 - 5];
+ acc += bp[5 * 2 - 5];
+ rp[1] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[2];
+ acc += bp[3 * 2 - 6];
+ acc += bp[4 * 2 - 6];
+ acc += bp[5 * 2 - 6];
+ rp[2] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[3];
+ acc += bp[3 * 2 - 5];
+ acc += bp[4 * 2 - 5];
+ acc += bp[5 * 2 - 5];
+ rp[3] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[4];
+ acc += bp[4 * 2 - 6];
+ acc += bp[5 * 2 - 6];
+ rp[4] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[5];
+ acc += bp[4 * 2 - 5];
+ acc += bp[5 * 2 - 5];
+ rp[5] = (unsigned int)acc;
+
+ carry = (int)(acc >> 32);
+ }
+#else
+ {
+ BN_ULONG t_d[BN_NIST_192_TOP];
+
+ nist_set_192(t_d, buf.bn, 0, 3, 3);
+ carry = (int)bn_add_words(r_d, r_d, t_d, BN_NIST_192_TOP);
+ nist_set_192(t_d, buf.bn, 4, 4, 0);
+ carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_192_TOP);
+ nist_set_192(t_d, buf.bn, 5, 5, 5)
+ carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_192_TOP);
+ }
+#endif
+ if (carry > 0)
+ carry =
+ (int)bn_sub_words(r_d, r_d, _nist_p_192[carry - 1],
+ BN_NIST_192_TOP);
+ else
+ carry = 1;
+
+ /*
+ * we need 'if (carry==0 || result>=modulus) result-=modulus;'
+ * as comparison implies subtraction, we can write
+ * 'tmp=result-modulus; if (!carry || !borrow) result=tmp;'
+ * this is what happens below, but without explicit if:-) a.
+ */
+ mask =
+ 0 - (PTR_SIZE_INT) bn_sub_words(c_d, r_d, _nist_p_192[0],
+ BN_NIST_192_TOP);
+ mask &= 0 - (PTR_SIZE_INT) carry;
+ res = c_d;
+ res = (BN_ULONG *)
+ (((PTR_SIZE_INT) res & ~mask) | ((PTR_SIZE_INT) r_d & mask));
+ nist_cp_bn(r_d, res, BN_NIST_192_TOP);
+ r->top = BN_NIST_192_TOP;
+ bn_correct_top(r);
+
+ return 1;
+}
+
+typedef BN_ULONG (*bn_addsub_f) (BN_ULONG *, const BN_ULONG *,
+ const BN_ULONG *, int);
+
+#define nist_set_224(to, from, a1, a2, a3, a4, a5, a6, a7) \
+ { \
+ bn_cp_32(to, 0, from, (a7) - 7) \
+ bn_cp_32(to, 1, from, (a6) - 7) \
+ bn_cp_32(to, 2, from, (a5) - 7) \
+ bn_cp_32(to, 3, from, (a4) - 7) \
+ bn_cp_32(to, 4, from, (a3) - 7) \
+ bn_cp_32(to, 5, from, (a2) - 7) \
+ bn_cp_32(to, 6, from, (a1) - 7) \
+ }
+
+int BN_nist_mod_224(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
+ BN_CTX *ctx)
+{
+ int top = a->top, i;
+ int carry;
+ BN_ULONG *r_d, *a_d = a->d;
+ union {
+ BN_ULONG bn[BN_NIST_224_TOP];
+ unsigned int ui[BN_NIST_224_TOP * sizeof(BN_ULONG) /
+ sizeof(unsigned int)];
+ } buf;
+ BN_ULONG c_d[BN_NIST_224_TOP], *res;
+ PTR_SIZE_INT mask;
+ union {
+ bn_addsub_f f;
+ PTR_SIZE_INT p;
+ } u;
+ static const BIGNUM _bignum_nist_p_224_sqr = {
+ (BN_ULONG *)_nist_p_224_sqr,
+ OSSL_NELEM(_nist_p_224_sqr),
+ OSSL_NELEM(_nist_p_224_sqr),
+ 0, BN_FLG_STATIC_DATA
+ };
+
+ field = &_bignum_nist_p_224; /* just to make sure */
+
+ if (BN_is_negative(a) || BN_ucmp(a, &_bignum_nist_p_224_sqr) >= 0)
+ return BN_nnmod(r, a, field, ctx);
+
+ i = BN_ucmp(field, a);
+ if (i == 0) {
+ BN_zero(r);
+ return 1;
+ } else if (i > 0)
+ return (r == a) ? 1 : (BN_copy(r, a) != NULL);
+
+ if (r != a) {
+ if (!bn_wexpand(r, BN_NIST_224_TOP))
+ return 0;
+ r_d = r->d;
+ nist_cp_bn(r_d, a_d, BN_NIST_224_TOP);
+ } else
+ r_d = a_d;
+
+#if BN_BITS2==64
+ /* copy upper 256 bits of 448 bit number ... */
+ nist_cp_bn_0(c_d, a_d + (BN_NIST_224_TOP - 1),
+ top - (BN_NIST_224_TOP - 1), BN_NIST_224_TOP);
+ /* ... and right shift by 32 to obtain upper 224 bits */
+ nist_set_224(buf.bn, c_d, 14, 13, 12, 11, 10, 9, 8);
+ /* truncate lower part to 224 bits too */
+ r_d[BN_NIST_224_TOP - 1] &= BN_MASK2l;
+#else
+ nist_cp_bn_0(buf.bn, a_d + BN_NIST_224_TOP, top - BN_NIST_224_TOP,
+ BN_NIST_224_TOP);
+#endif
+
+#if defined(NIST_INT64) && BN_BITS2!=64
+ {
+ NIST_INT64 acc; /* accumulator */
+ unsigned int *rp = (unsigned int *)r_d;
+ const unsigned int *bp = (const unsigned int *)buf.ui;
+
+ acc = rp[0];
+ acc -= bp[7 - 7];
+ acc -= bp[11 - 7];
+ rp[0] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[1];
+ acc -= bp[8 - 7];
+ acc -= bp[12 - 7];
+ rp[1] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[2];
+ acc -= bp[9 - 7];
+ acc -= bp[13 - 7];
+ rp[2] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[3];
+ acc += bp[7 - 7];
+ acc += bp[11 - 7];
+ acc -= bp[10 - 7];
+ rp[3] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[4];
+ acc += bp[8 - 7];
+ acc += bp[12 - 7];
+ acc -= bp[11 - 7];
+ rp[4] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[5];
+ acc += bp[9 - 7];
+ acc += bp[13 - 7];
+ acc -= bp[12 - 7];
+ rp[5] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[6];
+ acc += bp[10 - 7];
+ acc -= bp[13 - 7];
+ rp[6] = (unsigned int)acc;
+
+ carry = (int)(acc >> 32);
+# if BN_BITS2==64
+ rp[7] = carry;
+# endif
+ }
+#else
+ {
+ BN_ULONG t_d[BN_NIST_224_TOP];
+
+ nist_set_224(t_d, buf.bn, 10, 9, 8, 7, 0, 0, 0);
+ carry = (int)bn_add_words(r_d, r_d, t_d, BN_NIST_224_TOP);
+ nist_set_224(t_d, buf.bn, 0, 13, 12, 11, 0, 0, 0);
+ carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_224_TOP);
+ nist_set_224(t_d, buf.bn, 13, 12, 11, 10, 9, 8, 7);
+ carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_224_TOP);
+ nist_set_224(t_d, buf.bn, 0, 0, 0, 0, 13, 12, 11);
+ carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_224_TOP);
+
+# if BN_BITS2==64
+ carry = (int)(r_d[BN_NIST_224_TOP - 1] >> 32);
+# endif
+ }
+#endif
+ u.f = bn_sub_words;
+ if (carry > 0) {
+ carry =
+ (int)bn_sub_words(r_d, r_d, _nist_p_224[carry - 1],
+ BN_NIST_224_TOP);
+#if BN_BITS2==64
+ carry = (int)(~(r_d[BN_NIST_224_TOP - 1] >> 32)) & 1;
+#endif
+ } else if (carry < 0) {
+ /*
+ * it's a bit more complicated logic in this case. if bn_add_words
+ * yields no carry, then result has to be adjusted by unconditionally
+ * *adding* the modulus. but if it does, then result has to be
+ * compared to the modulus and conditionally adjusted by
+ * *subtracting* the latter.
+ */
+ carry =
+ (int)bn_add_words(r_d, r_d, _nist_p_224[-carry - 1],
+ BN_NIST_224_TOP);
+ mask = 0 - (PTR_SIZE_INT) carry;
+ u.p = ((PTR_SIZE_INT) bn_sub_words & mask) |
+ ((PTR_SIZE_INT) bn_add_words & ~mask);
+ } else
+ carry = 1;
+
+ /* otherwise it's effectively same as in BN_nist_mod_192... */
+ mask =
+ 0 - (PTR_SIZE_INT) (*u.f) (c_d, r_d, _nist_p_224[0], BN_NIST_224_TOP);
+ mask &= 0 - (PTR_SIZE_INT) carry;
+ res = c_d;
+ res = (BN_ULONG *)(((PTR_SIZE_INT) res & ~mask) |
+ ((PTR_SIZE_INT) r_d & mask));
+ nist_cp_bn(r_d, res, BN_NIST_224_TOP);
+ r->top = BN_NIST_224_TOP;
+ bn_correct_top(r);
+
+ return 1;
+}
+
+#define nist_set_256(to, from, a1, a2, a3, a4, a5, a6, a7, a8) \
+ { \
+ bn_cp_32(to, 0, from, (a8) - 8) \
+ bn_cp_32(to, 1, from, (a7) - 8) \
+ bn_cp_32(to, 2, from, (a6) - 8) \
+ bn_cp_32(to, 3, from, (a5) - 8) \
+ bn_cp_32(to, 4, from, (a4) - 8) \
+ bn_cp_32(to, 5, from, (a3) - 8) \
+ bn_cp_32(to, 6, from, (a2) - 8) \
+ bn_cp_32(to, 7, from, (a1) - 8) \
+ }
+
+int BN_nist_mod_256(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
+ BN_CTX *ctx)
+{
+ int i, top = a->top;
+ int carry = 0;
+ register BN_ULONG *a_d = a->d, *r_d;
+ union {
+ BN_ULONG bn[BN_NIST_256_TOP];
+ unsigned int ui[BN_NIST_256_TOP * sizeof(BN_ULONG) /
+ sizeof(unsigned int)];
+ } buf;
+ BN_ULONG c_d[BN_NIST_256_TOP], *res;
+ PTR_SIZE_INT mask;
+ union {
+ bn_addsub_f f;
+ PTR_SIZE_INT p;
+ } u;
+ static const BIGNUM _bignum_nist_p_256_sqr = {
+ (BN_ULONG *)_nist_p_256_sqr,
+ OSSL_NELEM(_nist_p_256_sqr),
+ OSSL_NELEM(_nist_p_256_sqr),
+ 0, BN_FLG_STATIC_DATA
+ };
+
+ field = &_bignum_nist_p_256; /* just to make sure */
+
+ if (BN_is_negative(a) || BN_ucmp(a, &_bignum_nist_p_256_sqr) >= 0)
+ return BN_nnmod(r, a, field, ctx);
+
+ i = BN_ucmp(field, a);
+ if (i == 0) {
+ BN_zero(r);
+ return 1;
+ } else if (i > 0)
+ return (r == a) ? 1 : (BN_copy(r, a) != NULL);
+
+ if (r != a) {
+ if (!bn_wexpand(r, BN_NIST_256_TOP))
+ return 0;
+ r_d = r->d;
+ nist_cp_bn(r_d, a_d, BN_NIST_256_TOP);
+ } else
+ r_d = a_d;
+
+ nist_cp_bn_0(buf.bn, a_d + BN_NIST_256_TOP, top - BN_NIST_256_TOP,
+ BN_NIST_256_TOP);
+
+#if defined(NIST_INT64)
+ {
+ NIST_INT64 acc; /* accumulator */
+ unsigned int *rp = (unsigned int *)r_d;
+ const unsigned int *bp = (const unsigned int *)buf.ui;
+
+ acc = rp[0];
+ acc += bp[8 - 8];
+ acc += bp[9 - 8];
+ acc -= bp[11 - 8];
+ acc -= bp[12 - 8];
+ acc -= bp[13 - 8];
+ acc -= bp[14 - 8];
+ rp[0] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[1];
+ acc += bp[9 - 8];
+ acc += bp[10 - 8];
+ acc -= bp[12 - 8];
+ acc -= bp[13 - 8];
+ acc -= bp[14 - 8];
+ acc -= bp[15 - 8];
+ rp[1] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[2];
+ acc += bp[10 - 8];
+ acc += bp[11 - 8];
+ acc -= bp[13 - 8];
+ acc -= bp[14 - 8];
+ acc -= bp[15 - 8];
+ rp[2] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[3];
+ acc += bp[11 - 8];
+ acc += bp[11 - 8];
+ acc += bp[12 - 8];
+ acc += bp[12 - 8];
+ acc += bp[13 - 8];
+ acc -= bp[15 - 8];
+ acc -= bp[8 - 8];
+ acc -= bp[9 - 8];
+ rp[3] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[4];
+ acc += bp[12 - 8];
+ acc += bp[12 - 8];
+ acc += bp[13 - 8];
+ acc += bp[13 - 8];
+ acc += bp[14 - 8];
+ acc -= bp[9 - 8];
+ acc -= bp[10 - 8];
+ rp[4] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[5];
+ acc += bp[13 - 8];
+ acc += bp[13 - 8];
+ acc += bp[14 - 8];
+ acc += bp[14 - 8];
+ acc += bp[15 - 8];
+ acc -= bp[10 - 8];
+ acc -= bp[11 - 8];
+ rp[5] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[6];
+ acc += bp[14 - 8];
+ acc += bp[14 - 8];
+ acc += bp[15 - 8];
+ acc += bp[15 - 8];
+ acc += bp[14 - 8];
+ acc += bp[13 - 8];
+ acc -= bp[8 - 8];
+ acc -= bp[9 - 8];
+ rp[6] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[7];
+ acc += bp[15 - 8];
+ acc += bp[15 - 8];
+ acc += bp[15 - 8];
+ acc += bp[8 - 8];
+ acc -= bp[10 - 8];
+ acc -= bp[11 - 8];
+ acc -= bp[12 - 8];
+ acc -= bp[13 - 8];
+ rp[7] = (unsigned int)acc;
+
+ carry = (int)(acc >> 32);
+ }
+#else
+ {
+ BN_ULONG t_d[BN_NIST_256_TOP];
+
+ /*
+ * S1
+ */
+ nist_set_256(t_d, buf.bn, 15, 14, 13, 12, 11, 0, 0, 0);
+ /*
+ * S2
+ */
+ nist_set_256(c_d, buf.bn, 0, 15, 14, 13, 12, 0, 0, 0);
+ carry = (int)bn_add_words(t_d, t_d, c_d, BN_NIST_256_TOP);
+ /* left shift */
+ {
+ register BN_ULONG *ap, t, c;
+ ap = t_d;
+ c = 0;
+ for (i = BN_NIST_256_TOP; i != 0; --i) {
+ t = *ap;
+ *(ap++) = ((t << 1) | c) & BN_MASK2;
+ c = (t & BN_TBIT) ? 1 : 0;
+ }
+ carry <<= 1;
+ carry |= c;
+ }
+ carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_256_TOP);
+ /*
+ * S3
+ */
+ nist_set_256(t_d, buf.bn, 15, 14, 0, 0, 0, 10, 9, 8);
+ carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_256_TOP);
+ /*
+ * S4
+ */
+ nist_set_256(t_d, buf.bn, 8, 13, 15, 14, 13, 11, 10, 9);
+ carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_256_TOP);
+ /*
+ * D1
+ */
+ nist_set_256(t_d, buf.bn, 10, 8, 0, 0, 0, 13, 12, 11);
+ carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_256_TOP);
+ /*
+ * D2
+ */
+ nist_set_256(t_d, buf.bn, 11, 9, 0, 0, 15, 14, 13, 12);
+ carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_256_TOP);
+ /*
+ * D3
+ */
+ nist_set_256(t_d, buf.bn, 12, 0, 10, 9, 8, 15, 14, 13);
+ carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_256_TOP);
+ /*
+ * D4
+ */
+ nist_set_256(t_d, buf.bn, 13, 0, 11, 10, 9, 0, 15, 14);
+ carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_256_TOP);
+
+ }
+#endif
+ /* see BN_nist_mod_224 for explanation */
+ u.f = bn_sub_words;
+ if (carry > 0)
+ carry =
+ (int)bn_sub_words(r_d, r_d, _nist_p_256[carry - 1],
+ BN_NIST_256_TOP);
+ else if (carry < 0) {
+ carry =
+ (int)bn_add_words(r_d, r_d, _nist_p_256[-carry - 1],
+ BN_NIST_256_TOP);
+ mask = 0 - (PTR_SIZE_INT) carry;
+ u.p = ((PTR_SIZE_INT) bn_sub_words & mask) |
+ ((PTR_SIZE_INT) bn_add_words & ~mask);
+ } else
+ carry = 1;
+
+ mask =
+ 0 - (PTR_SIZE_INT) (*u.f) (c_d, r_d, _nist_p_256[0], BN_NIST_256_TOP);
+ mask &= 0 - (PTR_SIZE_INT) carry;
+ res = c_d;
+ res = (BN_ULONG *)(((PTR_SIZE_INT) res & ~mask) |
+ ((PTR_SIZE_INT) r_d & mask));
+ nist_cp_bn(r_d, res, BN_NIST_256_TOP);
+ r->top = BN_NIST_256_TOP;
+ bn_correct_top(r);
+
+ return 1;
+}
+
+#define nist_set_384(to,from,a1,a2,a3,a4,a5,a6,a7,a8,a9,a10,a11,a12) \
+ { \
+ bn_cp_32(to, 0, from, (a12) - 12) \
+ bn_cp_32(to, 1, from, (a11) - 12) \
+ bn_cp_32(to, 2, from, (a10) - 12) \
+ bn_cp_32(to, 3, from, (a9) - 12) \
+ bn_cp_32(to, 4, from, (a8) - 12) \
+ bn_cp_32(to, 5, from, (a7) - 12) \
+ bn_cp_32(to, 6, from, (a6) - 12) \
+ bn_cp_32(to, 7, from, (a5) - 12) \
+ bn_cp_32(to, 8, from, (a4) - 12) \
+ bn_cp_32(to, 9, from, (a3) - 12) \
+ bn_cp_32(to, 10, from, (a2) - 12) \
+ bn_cp_32(to, 11, from, (a1) - 12) \
+ }
+
+int BN_nist_mod_384(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
+ BN_CTX *ctx)
+{
+ int i, top = a->top;
+ int carry = 0;
+ register BN_ULONG *r_d, *a_d = a->d;
+ union {
+ BN_ULONG bn[BN_NIST_384_TOP];
+ unsigned int ui[BN_NIST_384_TOP * sizeof(BN_ULONG) /
+ sizeof(unsigned int)];
+ } buf;
+ BN_ULONG c_d[BN_NIST_384_TOP], *res;
+ PTR_SIZE_INT mask;
+ union {
+ bn_addsub_f f;
+ PTR_SIZE_INT p;
+ } u;
+ static const BIGNUM _bignum_nist_p_384_sqr = {
+ (BN_ULONG *)_nist_p_384_sqr,
+ OSSL_NELEM(_nist_p_384_sqr),
+ OSSL_NELEM(_nist_p_384_sqr),
+ 0, BN_FLG_STATIC_DATA
+ };
+
+ field = &_bignum_nist_p_384; /* just to make sure */
+
+ if (BN_is_negative(a) || BN_ucmp(a, &_bignum_nist_p_384_sqr) >= 0)
+ return BN_nnmod(r, a, field, ctx);
+
+ i = BN_ucmp(field, a);
+ if (i == 0) {
+ BN_zero(r);
+ return 1;
+ } else if (i > 0)
+ return (r == a) ? 1 : (BN_copy(r, a) != NULL);
+
+ if (r != a) {
+ if (!bn_wexpand(r, BN_NIST_384_TOP))
+ return 0;
+ r_d = r->d;
+ nist_cp_bn(r_d, a_d, BN_NIST_384_TOP);
+ } else
+ r_d = a_d;
+
+ nist_cp_bn_0(buf.bn, a_d + BN_NIST_384_TOP, top - BN_NIST_384_TOP,
+ BN_NIST_384_TOP);
+
+#if defined(NIST_INT64)
+ {
+ NIST_INT64 acc; /* accumulator */
+ unsigned int *rp = (unsigned int *)r_d;
+ const unsigned int *bp = (const unsigned int *)buf.ui;
+
+ acc = rp[0];
+ acc += bp[12 - 12];
+ acc += bp[21 - 12];
+ acc += bp[20 - 12];
+ acc -= bp[23 - 12];
+ rp[0] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[1];
+ acc += bp[13 - 12];
+ acc += bp[22 - 12];
+ acc += bp[23 - 12];
+ acc -= bp[12 - 12];
+ acc -= bp[20 - 12];
+ rp[1] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[2];
+ acc += bp[14 - 12];
+ acc += bp[23 - 12];
+ acc -= bp[13 - 12];
+ acc -= bp[21 - 12];
+ rp[2] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[3];
+ acc += bp[15 - 12];
+ acc += bp[12 - 12];
+ acc += bp[20 - 12];
+ acc += bp[21 - 12];
+ acc -= bp[14 - 12];
+ acc -= bp[22 - 12];
+ acc -= bp[23 - 12];
+ rp[3] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[4];
+ acc += bp[21 - 12];
+ acc += bp[21 - 12];
+ acc += bp[16 - 12];
+ acc += bp[13 - 12];
+ acc += bp[12 - 12];
+ acc += bp[20 - 12];
+ acc += bp[22 - 12];
+ acc -= bp[15 - 12];
+ acc -= bp[23 - 12];
+ acc -= bp[23 - 12];
+ rp[4] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[5];
+ acc += bp[22 - 12];
+ acc += bp[22 - 12];
+ acc += bp[17 - 12];
+ acc += bp[14 - 12];
+ acc += bp[13 - 12];
+ acc += bp[21 - 12];
+ acc += bp[23 - 12];
+ acc -= bp[16 - 12];
+ rp[5] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[6];
+ acc += bp[23 - 12];
+ acc += bp[23 - 12];
+ acc += bp[18 - 12];
+ acc += bp[15 - 12];
+ acc += bp[14 - 12];
+ acc += bp[22 - 12];
+ acc -= bp[17 - 12];
+ rp[6] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[7];
+ acc += bp[19 - 12];
+ acc += bp[16 - 12];
+ acc += bp[15 - 12];
+ acc += bp[23 - 12];
+ acc -= bp[18 - 12];
+ rp[7] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[8];
+ acc += bp[20 - 12];
+ acc += bp[17 - 12];
+ acc += bp[16 - 12];
+ acc -= bp[19 - 12];
+ rp[8] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[9];
+ acc += bp[21 - 12];
+ acc += bp[18 - 12];
+ acc += bp[17 - 12];
+ acc -= bp[20 - 12];
+ rp[9] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[10];
+ acc += bp[22 - 12];
+ acc += bp[19 - 12];
+ acc += bp[18 - 12];
+ acc -= bp[21 - 12];
+ rp[10] = (unsigned int)acc;
+ acc >>= 32;
+
+ acc += rp[11];
+ acc += bp[23 - 12];
+ acc += bp[20 - 12];
+ acc += bp[19 - 12];
+ acc -= bp[22 - 12];
+ rp[11] = (unsigned int)acc;
+
+ carry = (int)(acc >> 32);
+ }
+#else
+ {
+ BN_ULONG t_d[BN_NIST_384_TOP];
+
+ /*
+ * S1
+ */
+ nist_set_256(t_d, buf.bn, 0, 0, 0, 0, 0, 23 - 4, 22 - 4, 21 - 4);
+ /* left shift */
+ {
+ register BN_ULONG *ap, t, c;
+ ap = t_d;
+ c = 0;
+ for (i = 3; i != 0; --i) {
+ t = *ap;
+ *(ap++) = ((t << 1) | c) & BN_MASK2;
+ c = (t & BN_TBIT) ? 1 : 0;
+ }
+ *ap = c;
+ }
+ carry =
+ (int)bn_add_words(r_d + (128 / BN_BITS2), r_d + (128 / BN_BITS2),
+ t_d, BN_NIST_256_TOP);
+ /*
+ * S2
+ */
+ carry += (int)bn_add_words(r_d, r_d, buf.bn, BN_NIST_384_TOP);
+ /*
+ * S3
+ */
+ nist_set_384(t_d, buf.bn, 20, 19, 18, 17, 16, 15, 14, 13, 12, 23, 22,
+ 21);
+ carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_384_TOP);
+ /*
+ * S4
+ */
+ nist_set_384(t_d, buf.bn, 19, 18, 17, 16, 15, 14, 13, 12, 20, 0, 23,
+ 0);
+ carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_384_TOP);
+ /*
+ * S5
+ */
+ nist_set_384(t_d, buf.bn, 0, 0, 0, 0, 23, 22, 21, 20, 0, 0, 0, 0);
+ carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_384_TOP);
+ /*
+ * S6
+ */
+ nist_set_384(t_d, buf.bn, 0, 0, 0, 0, 0, 0, 23, 22, 21, 0, 0, 20);
+ carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_384_TOP);
+ /*
+ * D1
+ */
+ nist_set_384(t_d, buf.bn, 22, 21, 20, 19, 18, 17, 16, 15, 14, 13, 12,
+ 23);
+ carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_384_TOP);
+ /*
+ * D2
+ */
+ nist_set_384(t_d, buf.bn, 0, 0, 0, 0, 0, 0, 0, 23, 22, 21, 20, 0);
+ carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_384_TOP);
+ /*
+ * D3
+ */
+ nist_set_384(t_d, buf.bn, 0, 0, 0, 0, 0, 0, 0, 23, 23, 0, 0, 0);
+ carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_384_TOP);
+
+ }
+#endif
+ /* see BN_nist_mod_224 for explanation */
+ u.f = bn_sub_words;
+ if (carry > 0)
+ carry =
+ (int)bn_sub_words(r_d, r_d, _nist_p_384[carry - 1],
+ BN_NIST_384_TOP);
+ else if (carry < 0) {
+ carry =
+ (int)bn_add_words(r_d, r_d, _nist_p_384[-carry - 1],
+ BN_NIST_384_TOP);
+ mask = 0 - (PTR_SIZE_INT) carry;
+ u.p = ((PTR_SIZE_INT) bn_sub_words & mask) |
+ ((PTR_SIZE_INT) bn_add_words & ~mask);
+ } else
+ carry = 1;
+
+ mask =
+ 0 - (PTR_SIZE_INT) (*u.f) (c_d, r_d, _nist_p_384[0], BN_NIST_384_TOP);
+ mask &= 0 - (PTR_SIZE_INT) carry;
+ res = c_d;
+ res = (BN_ULONG *)(((PTR_SIZE_INT) res & ~mask) |
+ ((PTR_SIZE_INT) r_d & mask));
+ nist_cp_bn(r_d, res, BN_NIST_384_TOP);
+ r->top = BN_NIST_384_TOP;
+ bn_correct_top(r);
+
+ return 1;
+}
+
+#define BN_NIST_521_RSHIFT (521%BN_BITS2)
+#define BN_NIST_521_LSHIFT (BN_BITS2-BN_NIST_521_RSHIFT)
+#define BN_NIST_521_TOP_MASK ((BN_ULONG)BN_MASK2>>BN_NIST_521_LSHIFT)
+
+int BN_nist_mod_521(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
+ BN_CTX *ctx)
+{
+ int top = a->top, i;
+ BN_ULONG *r_d, *a_d = a->d, t_d[BN_NIST_521_TOP], val, tmp, *res;
+ PTR_SIZE_INT mask;
+ static const BIGNUM _bignum_nist_p_521_sqr = {
+ (BN_ULONG *)_nist_p_521_sqr,
+ OSSL_NELEM(_nist_p_521_sqr),
+ OSSL_NELEM(_nist_p_521_sqr),
+ 0, BN_FLG_STATIC_DATA
+ };
+
+ field = &_bignum_nist_p_521; /* just to make sure */
+
+ if (BN_is_negative(a) || BN_ucmp(a, &_bignum_nist_p_521_sqr) >= 0)
+ return BN_nnmod(r, a, field, ctx);
+
+ i = BN_ucmp(field, a);
+ if (i == 0) {
+ BN_zero(r);
+ return 1;
+ } else if (i > 0)
+ return (r == a) ? 1 : (BN_copy(r, a) != NULL);
+
+ if (r != a) {
+ if (!bn_wexpand(r, BN_NIST_521_TOP))
+ return 0;
+ r_d = r->d;
+ nist_cp_bn(r_d, a_d, BN_NIST_521_TOP);
+ } else
+ r_d = a_d;
+
+ /* upper 521 bits, copy ... */
+ nist_cp_bn_0(t_d, a_d + (BN_NIST_521_TOP - 1),
+ top - (BN_NIST_521_TOP - 1), BN_NIST_521_TOP);
+ /* ... and right shift */
+ for (val = t_d[0], i = 0; i < BN_NIST_521_TOP - 1; i++) {
+#if 0
+ /*
+ * MSC ARM compiler [version 2013, presumably even earlier,
+ * much earlier] miscompiles this code, but not one in
+ * #else section. See RT#3541.
+ */
+ tmp = val >> BN_NIST_521_RSHIFT;
+ val = t_d[i + 1];
+ t_d[i] = (tmp | val << BN_NIST_521_LSHIFT) & BN_MASK2;
+#else
+ t_d[i] = (val >> BN_NIST_521_RSHIFT |
+ (tmp = t_d[i + 1]) << BN_NIST_521_LSHIFT) & BN_MASK2;
+ val = tmp;
+#endif
+ }
+ t_d[i] = val >> BN_NIST_521_RSHIFT;
+ /* lower 521 bits */
+ r_d[i] &= BN_NIST_521_TOP_MASK;
+
+ bn_add_words(r_d, r_d, t_d, BN_NIST_521_TOP);
+ mask =
+ 0 - (PTR_SIZE_INT) bn_sub_words(t_d, r_d, _nist_p_521,
+ BN_NIST_521_TOP);
+ res = t_d;
+ res = (BN_ULONG *)(((PTR_SIZE_INT) res & ~mask) |
+ ((PTR_SIZE_INT) r_d & mask));
+ nist_cp_bn(r_d, res, BN_NIST_521_TOP);
+ r->top = BN_NIST_521_TOP;
+ bn_correct_top(r);
+
+ return 1;
+}
+
+int (*BN_nist_mod_func(const BIGNUM *p)) (BIGNUM *r, const BIGNUM *a,
+ const BIGNUM *field, BN_CTX *ctx) {
+ if (BN_ucmp(&_bignum_nist_p_192, p) == 0)
+ return BN_nist_mod_192;
+ if (BN_ucmp(&_bignum_nist_p_224, p) == 0)
+ return BN_nist_mod_224;
+ if (BN_ucmp(&_bignum_nist_p_256, p) == 0)
+ return BN_nist_mod_256;
+ if (BN_ucmp(&_bignum_nist_p_384, p) == 0)
+ return BN_nist_mod_384;
+ if (BN_ucmp(&_bignum_nist_p_521, p) == 0)
+ return BN_nist_mod_521;
+ return 0;
+}
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_prime.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_prime.c
new file mode 100644
index 0000000..d0cf377
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_prime.c
@@ -0,0 +1,391 @@
+/*
+ * Copyright 1995-2020 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include <stdio.h>
+#include <time.h>
+#include "internal/cryptlib.h"
+#include "bn_local.h"
+
+/*
+ * The quick sieve algorithm approach to weeding out primes is Philip
+ * Zimmermann's, as implemented in PGP. I have had a read of his comments
+ * and implemented my own version.
+ */
+#include "bn_prime.h"
+
+static int witness(BIGNUM *w, const BIGNUM *a, const BIGNUM *a1,
+ const BIGNUM *a1_odd, int k, BN_CTX *ctx,
+ BN_MONT_CTX *mont);
+static int probable_prime(BIGNUM *rnd, int bits, int safe, prime_t *mods);
+static int probable_prime_dh(BIGNUM *rnd, int bits, int safe, prime_t *mods,
+ const BIGNUM *add, const BIGNUM *rem,
+ BN_CTX *ctx);
+
+#define square(x) ((BN_ULONG)(x) * (BN_ULONG)(x))
+
+int BN_GENCB_call(BN_GENCB *cb, int a, int b)
+{
+ /* No callback means continue */
+ if (!cb)
+ return 1;
+ switch (cb->ver) {
+ case 1:
+ /* Deprecated-style callbacks */
+ if (!cb->cb.cb_1)
+ return 1;
+ cb->cb.cb_1(a, b, cb->arg);
+ return 1;
+ case 2:
+ /* New-style callbacks */
+ return cb->cb.cb_2(a, b, cb);
+ default:
+ break;
+ }
+ /* Unrecognised callback type */
+ return 0;
+}
+
+int BN_generate_prime_ex(BIGNUM *ret, int bits, int safe,
+ const BIGNUM *add, const BIGNUM *rem, BN_GENCB *cb)
+{
+ BIGNUM *t;
+ int found = 0;
+ int i, j, c1 = 0;
+ BN_CTX *ctx = NULL;
+ prime_t *mods = NULL;
+ int checks = BN_prime_checks_for_size(bits);
+
+ if (bits < 2) {
+ /* There are no prime numbers this small. */
+ BNerr(BN_F_BN_GENERATE_PRIME_EX, BN_R_BITS_TOO_SMALL);
+ return 0;
+ } else if (add == NULL && safe && bits < 6 && bits != 3) {
+ /*
+ * The smallest safe prime (7) is three bits.
+ * But the following two safe primes with less than 6 bits (11, 23)
+ * are unreachable for BN_rand with BN_RAND_TOP_TWO.
+ */
+ BNerr(BN_F_BN_GENERATE_PRIME_EX, BN_R_BITS_TOO_SMALL);
+ return 0;
+ }
+
+ mods = OPENSSL_zalloc(sizeof(*mods) * NUMPRIMES);
+ if (mods == NULL)
+ goto err;
+
+ ctx = BN_CTX_new();
+ if (ctx == NULL)
+ goto err;
+ BN_CTX_start(ctx);
+ t = BN_CTX_get(ctx);
+ if (t == NULL)
+ goto err;
+ loop:
+ /* make a random number and set the top and bottom bits */
+ if (add == NULL) {
+ if (!probable_prime(ret, bits, safe, mods))
+ goto err;
+ } else {
+ if (!probable_prime_dh(ret, bits, safe, mods, add, rem, ctx))
+ goto err;
+ }
+
+ if (!BN_GENCB_call(cb, 0, c1++))
+ /* aborted */
+ goto err;
+
+ if (!safe) {
+ i = BN_is_prime_fasttest_ex(ret, checks, ctx, 0, cb);
+ if (i == -1)
+ goto err;
+ if (i == 0)
+ goto loop;
+ } else {
+ /*
+ * for "safe prime" generation, check that (p-1)/2 is prime. Since a
+ * prime is odd, We just need to divide by 2
+ */
+ if (!BN_rshift1(t, ret))
+ goto err;
+
+ for (i = 0; i < checks; i++) {
+ j = BN_is_prime_fasttest_ex(ret, 1, ctx, 0, cb);
+ if (j == -1)
+ goto err;
+ if (j == 0)
+ goto loop;
+
+ j = BN_is_prime_fasttest_ex(t, 1, ctx, 0, cb);
+ if (j == -1)
+ goto err;
+ if (j == 0)
+ goto loop;
+
+ if (!BN_GENCB_call(cb, 2, c1 - 1))
+ goto err;
+ /* We have a safe prime test pass */
+ }
+ }
+ /* we have a prime :-) */
+ found = 1;
+ err:
+ OPENSSL_free(mods);
+ BN_CTX_end(ctx);
+ BN_CTX_free(ctx);
+ bn_check_top(ret);
+ return found;
+}
+
+int BN_is_prime_ex(const BIGNUM *a, int checks, BN_CTX *ctx_passed,
+ BN_GENCB *cb)
+{
+ return BN_is_prime_fasttest_ex(a, checks, ctx_passed, 0, cb);
+}
+
+int BN_is_prime_fasttest_ex(const BIGNUM *a, int checks, BN_CTX *ctx_passed,
+ int do_trial_division, BN_GENCB *cb)
+{
+ int i, j, ret = -1;
+ int k;
+ BN_CTX *ctx = NULL;
+ BIGNUM *A1, *A1_odd, *A3, *check; /* taken from ctx */
+ BN_MONT_CTX *mont = NULL;
+
+ /* Take care of the really small primes 2 & 3 */
+ if (BN_is_word(a, 2) || BN_is_word(a, 3))
+ return 1;
+
+ /* Check odd and bigger than 1 */
+ if (!BN_is_odd(a) || BN_cmp(a, BN_value_one()) <= 0)
+ return 0;
+
+ if (checks == BN_prime_checks)
+ checks = BN_prime_checks_for_size(BN_num_bits(a));
+
+ /* first look for small factors */
+ if (do_trial_division) {
+ for (i = 1; i < NUMPRIMES; i++) {
+ BN_ULONG mod = BN_mod_word(a, primes[i]);
+ if (mod == (BN_ULONG)-1)
+ goto err;
+ if (mod == 0)
+ return BN_is_word(a, primes[i]);
+ }
+ if (!BN_GENCB_call(cb, 1, -1))
+ goto err;
+ }
+
+ if (ctx_passed != NULL)
+ ctx = ctx_passed;
+ else if ((ctx = BN_CTX_new()) == NULL)
+ goto err;
+ BN_CTX_start(ctx);
+
+ A1 = BN_CTX_get(ctx);
+ A3 = BN_CTX_get(ctx);
+ A1_odd = BN_CTX_get(ctx);
+ check = BN_CTX_get(ctx);
+ if (check == NULL)
+ goto err;
+
+ /* compute A1 := a - 1 */
+ if (!BN_copy(A1, a) || !BN_sub_word(A1, 1))
+ goto err;
+ /* compute A3 := a - 3 */
+ if (!BN_copy(A3, a) || !BN_sub_word(A3, 3))
+ goto err;
+
+ /* write A1 as A1_odd * 2^k */
+ k = 1;
+ while (!BN_is_bit_set(A1, k))
+ k++;
+ if (!BN_rshift(A1_odd, A1, k))
+ goto err;
+
+ /* Montgomery setup for computations mod a */
+ mont = BN_MONT_CTX_new();
+ if (mont == NULL)
+ goto err;
+ if (!BN_MONT_CTX_set(mont, a, ctx))
+ goto err;
+
+ for (i = 0; i < checks; i++) {
+ /* 1 < check < a-1 */
+ if (!BN_priv_rand_range(check, A3) || !BN_add_word(check, 2))
+ goto err;
+
+ j = witness(check, a, A1, A1_odd, k, ctx, mont);
+ if (j == -1)
+ goto err;
+ if (j) {
+ ret = 0;
+ goto err;
+ }
+ if (!BN_GENCB_call(cb, 1, i))
+ goto err;
+ }
+ ret = 1;
+ err:
+ if (ctx != NULL) {
+ BN_CTX_end(ctx);
+ if (ctx_passed == NULL)
+ BN_CTX_free(ctx);
+ }
+ BN_MONT_CTX_free(mont);
+
+ return ret;
+}
+
+static int witness(BIGNUM *w, const BIGNUM *a, const BIGNUM *a1,
+ const BIGNUM *a1_odd, int k, BN_CTX *ctx,
+ BN_MONT_CTX *mont)
+{
+ if (!BN_mod_exp_mont(w, w, a1_odd, a, ctx, mont)) /* w := w^a1_odd mod a */
+ return -1;
+ if (BN_is_one(w))
+ return 0; /* probably prime */
+ if (BN_cmp(w, a1) == 0)
+ return 0; /* w == -1 (mod a), 'a' is probably prime */
+ while (--k) {
+ if (!BN_mod_mul(w, w, w, a, ctx)) /* w := w^2 mod a */
+ return -1;
+ if (BN_is_one(w))
+ return 1; /* 'a' is composite, otherwise a previous 'w'
+ * would have been == -1 (mod 'a') */
+ if (BN_cmp(w, a1) == 0)
+ return 0; /* w == -1 (mod a), 'a' is probably prime */
+ }
+ /*
+ * If we get here, 'w' is the (a-1)/2-th power of the original 'w', and
+ * it is neither -1 nor +1 -- so 'a' cannot be prime
+ */
+ bn_check_top(w);
+ return 1;
+}
+
+static int probable_prime(BIGNUM *rnd, int bits, int safe, prime_t *mods)
+{
+ int i;
+ BN_ULONG delta;
+ BN_ULONG maxdelta = BN_MASK2 - primes[NUMPRIMES - 1];
+
+ again:
+ /* TODO: Not all primes are private */
+ if (!BN_priv_rand(rnd, bits, BN_RAND_TOP_TWO, BN_RAND_BOTTOM_ODD))
+ return 0;
+ if (safe && !BN_set_bit(rnd, 1))
+ return 0;
+ /* we now have a random number 'rnd' to test. */
+ for (i = 1; i < NUMPRIMES; i++) {
+ BN_ULONG mod = BN_mod_word(rnd, (BN_ULONG)primes[i]);
+ if (mod == (BN_ULONG)-1)
+ return 0;
+ mods[i] = (prime_t) mod;
+ }
+ delta = 0;
+ loop:
+ for (i = 1; i < NUMPRIMES; i++) {
+ /*
+ * check that rnd is a prime and also that
+ * gcd(rnd-1,primes) == 1 (except for 2)
+ * do the second check only if we are interested in safe primes
+ * in the case that the candidate prime is a single word then
+ * we check only the primes up to sqrt(rnd)
+ */
+ if (bits <= 31 && delta <= 0x7fffffff
+ && square(primes[i]) > BN_get_word(rnd) + delta)
+ break;
+ if (safe ? (mods[i] + delta) % primes[i] <= 1
+ : (mods[i] + delta) % primes[i] == 0) {
+ delta += safe ? 4 : 2;
+ if (delta > maxdelta)
+ goto again;
+ goto loop;
+ }
+ }
+ if (!BN_add_word(rnd, delta))
+ return 0;
+ if (BN_num_bits(rnd) != bits)
+ goto again;
+ bn_check_top(rnd);
+ return 1;
+}
+
+static int probable_prime_dh(BIGNUM *rnd, int bits, int safe, prime_t *mods,
+ const BIGNUM *add, const BIGNUM *rem,
+ BN_CTX *ctx)
+{
+ int i, ret = 0;
+ BIGNUM *t1;
+ BN_ULONG delta;
+ BN_ULONG maxdelta = BN_MASK2 - primes[NUMPRIMES - 1];
+
+ BN_CTX_start(ctx);
+ if ((t1 = BN_CTX_get(ctx)) == NULL)
+ goto err;
+
+ if (maxdelta > BN_MASK2 - BN_get_word(add))
+ maxdelta = BN_MASK2 - BN_get_word(add);
+
+ again:
+ if (!BN_rand(rnd, bits, BN_RAND_TOP_ONE, BN_RAND_BOTTOM_ODD))
+ goto err;
+
+ /* we need ((rnd-rem) % add) == 0 */
+
+ if (!BN_mod(t1, rnd, add, ctx))
+ goto err;
+ if (!BN_sub(rnd, rnd, t1))
+ goto err;
+ if (rem == NULL) {
+ if (!BN_add_word(rnd, safe ? 3u : 1u))
+ goto err;
+ } else {
+ if (!BN_add(rnd, rnd, rem))
+ goto err;
+ }
+
+ if (BN_num_bits(rnd) < bits
+ || BN_get_word(rnd) < (safe ? 5u : 3u)) {
+ if (!BN_add(rnd, rnd, add))
+ goto err;
+ }
+
+ /* we now have a random number 'rnd' to test. */
+ for (i = 1; i < NUMPRIMES; i++) {
+ BN_ULONG mod = BN_mod_word(rnd, (BN_ULONG)primes[i]);
+ if (mod == (BN_ULONG)-1)
+ goto err;
+ mods[i] = (prime_t) mod;
+ }
+ delta = 0;
+ loop:
+ for (i = 1; i < NUMPRIMES; i++) {
+ /* check that rnd is a prime */
+ if (bits <= 31 && delta <= 0x7fffffff
+ && square(primes[i]) > BN_get_word(rnd) + delta)
+ break;
+ /* rnd mod p == 1 implies q = (rnd-1)/2 is divisible by p */
+ if (safe ? (mods[i] + delta) % primes[i] <= 1
+ : (mods[i] + delta) % primes[i] == 0) {
+ delta += BN_get_word(add);
+ if (delta > maxdelta)
+ goto again;
+ goto loop;
+ }
+ }
+ if (!BN_add_word(rnd, delta))
+ goto err;
+ ret = 1;
+
+ err:
+ BN_CTX_end(ctx);
+ bn_check_top(rnd);
+ return ret;
+}
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_prime.h b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_prime.h
new file mode 100644
index 0000000..8f2d7e9
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_prime.h
@@ -0,0 +1,273 @@
+/*
+ * WARNING: do not edit!
+ * Generated by crypto/bn/bn_prime.pl
+ *
+ * Copyright 1998-2022 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+typedef unsigned short prime_t;
+# define NUMPRIMES 2048
+
+static const prime_t primes[2048] = {
+ 2, 3, 5, 7, 11, 13, 17, 19,
+ 23, 29, 31, 37, 41, 43, 47, 53,
+ 59, 61, 67, 71, 73, 79, 83, 89,
+ 97, 101, 103, 107, 109, 113, 127, 131,
+ 137, 139, 149, 151, 157, 163, 167, 173,
+ 179, 181, 191, 193, 197, 199, 211, 223,
+ 227, 229, 233, 239, 241, 251, 257, 263,
+ 269, 271, 277, 281, 283, 293, 307, 311,
+ 313, 317, 331, 337, 347, 349, 353, 359,
+ 367, 373, 379, 383, 389, 397, 401, 409,
+ 419, 421, 431, 433, 439, 443, 449, 457,
+ 461, 463, 467, 479, 487, 491, 499, 503,
+ 509, 521, 523, 541, 547, 557, 563, 569,
+ 571, 577, 587, 593, 599, 601, 607, 613,
+ 617, 619, 631, 641, 643, 647, 653, 659,
+ 661, 673, 677, 683, 691, 701, 709, 719,
+ 727, 733, 739, 743, 751, 757, 761, 769,
+ 773, 787, 797, 809, 811, 821, 823, 827,
+ 829, 839, 853, 857, 859, 863, 877, 881,
+ 883, 887, 907, 911, 919, 929, 937, 941,
+ 947, 953, 967, 971, 977, 983, 991, 997,
+ 1009, 1013, 1019, 1021, 1031, 1033, 1039, 1049,
+ 1051, 1061, 1063, 1069, 1087, 1091, 1093, 1097,
+ 1103, 1109, 1117, 1123, 1129, 1151, 1153, 1163,
+ 1171, 1181, 1187, 1193, 1201, 1213, 1217, 1223,
+ 1229, 1231, 1237, 1249, 1259, 1277, 1279, 1283,
+ 1289, 1291, 1297, 1301, 1303, 1307, 1319, 1321,
+ 1327, 1361, 1367, 1373, 1381, 1399, 1409, 1423,
+ 1427, 1429, 1433, 1439, 1447, 1451, 1453, 1459,
+ 1471, 1481, 1483, 1487, 1489, 1493, 1499, 1511,
+ 1523, 1531, 1543, 1549, 1553, 1559, 1567, 1571,
+ 1579, 1583, 1597, 1601, 1607, 1609, 1613, 1619,
+ 1621, 1627, 1637, 1657, 1663, 1667, 1669, 1693,
+ 1697, 1699, 1709, 1721, 1723, 1733, 1741, 1747,
+ 1753, 1759, 1777, 1783, 1787, 1789, 1801, 1811,
+ 1823, 1831, 1847, 1861, 1867, 1871, 1873, 1877,
+ 1879, 1889, 1901, 1907, 1913, 1931, 1933, 1949,
+ 1951, 1973, 1979, 1987, 1993, 1997, 1999, 2003,
+ 2011, 2017, 2027, 2029, 2039, 2053, 2063, 2069,
+ 2081, 2083, 2087, 2089, 2099, 2111, 2113, 2129,
+ 2131, 2137, 2141, 2143, 2153, 2161, 2179, 2203,
+ 2207, 2213, 2221, 2237, 2239, 2243, 2251, 2267,
+ 2269, 2273, 2281, 2287, 2293, 2297, 2309, 2311,
+ 2333, 2339, 2341, 2347, 2351, 2357, 2371, 2377,
+ 2381, 2383, 2389, 2393, 2399, 2411, 2417, 2423,
+ 2437, 2441, 2447, 2459, 2467, 2473, 2477, 2503,
+ 2521, 2531, 2539, 2543, 2549, 2551, 2557, 2579,
+ 2591, 2593, 2609, 2617, 2621, 2633, 2647, 2657,
+ 2659, 2663, 2671, 2677, 2683, 2687, 2689, 2693,
+ 2699, 2707, 2711, 2713, 2719, 2729, 2731, 2741,
+ 2749, 2753, 2767, 2777, 2789, 2791, 2797, 2801,
+ 2803, 2819, 2833, 2837, 2843, 2851, 2857, 2861,
+ 2879, 2887, 2897, 2903, 2909, 2917, 2927, 2939,
+ 2953, 2957, 2963, 2969, 2971, 2999, 3001, 3011,
+ 3019, 3023, 3037, 3041, 3049, 3061, 3067, 3079,
+ 3083, 3089, 3109, 3119, 3121, 3137, 3163, 3167,
+ 3169, 3181, 3187, 3191, 3203, 3209, 3217, 3221,
+ 3229, 3251, 3253, 3257, 3259, 3271, 3299, 3301,
+ 3307, 3313, 3319, 3323, 3329, 3331, 3343, 3347,
+ 3359, 3361, 3371, 3373, 3389, 3391, 3407, 3413,
+ 3433, 3449, 3457, 3461, 3463, 3467, 3469, 3491,
+ 3499, 3511, 3517, 3527, 3529, 3533, 3539, 3541,
+ 3547, 3557, 3559, 3571, 3581, 3583, 3593, 3607,
+ 3613, 3617, 3623, 3631, 3637, 3643, 3659, 3671,
+ 3673, 3677, 3691, 3697, 3701, 3709, 3719, 3727,
+ 3733, 3739, 3761, 3767, 3769, 3779, 3793, 3797,
+ 3803, 3821, 3823, 3833, 3847, 3851, 3853, 3863,
+ 3877, 3881, 3889, 3907, 3911, 3917, 3919, 3923,
+ 3929, 3931, 3943, 3947, 3967, 3989, 4001, 4003,
+ 4007, 4013, 4019, 4021, 4027, 4049, 4051, 4057,
+ 4073, 4079, 4091, 4093, 4099, 4111, 4127, 4129,
+ 4133, 4139, 4153, 4157, 4159, 4177, 4201, 4211,
+ 4217, 4219, 4229, 4231, 4241, 4243, 4253, 4259,
+ 4261, 4271, 4273, 4283, 4289, 4297, 4327, 4337,
+ 4339, 4349, 4357, 4363, 4373, 4391, 4397, 4409,
+ 4421, 4423, 4441, 4447, 4451, 4457, 4463, 4481,
+ 4483, 4493, 4507, 4513, 4517, 4519, 4523, 4547,
+ 4549, 4561, 4567, 4583, 4591, 4597, 4603, 4621,
+ 4637, 4639, 4643, 4649, 4651, 4657, 4663, 4673,
+ 4679, 4691, 4703, 4721, 4723, 4729, 4733, 4751,
+ 4759, 4783, 4787, 4789, 4793, 4799, 4801, 4813,
+ 4817, 4831, 4861, 4871, 4877, 4889, 4903, 4909,
+ 4919, 4931, 4933, 4937, 4943, 4951, 4957, 4967,
+ 4969, 4973, 4987, 4993, 4999, 5003, 5009, 5011,
+ 5021, 5023, 5039, 5051, 5059, 5077, 5081, 5087,
+ 5099, 5101, 5107, 5113, 5119, 5147, 5153, 5167,
+ 5171, 5179, 5189, 5197, 5209, 5227, 5231, 5233,
+ 5237, 5261, 5273, 5279, 5281, 5297, 5303, 5309,
+ 5323, 5333, 5347, 5351, 5381, 5387, 5393, 5399,
+ 5407, 5413, 5417, 5419, 5431, 5437, 5441, 5443,
+ 5449, 5471, 5477, 5479, 5483, 5501, 5503, 5507,
+ 5519, 5521, 5527, 5531, 5557, 5563, 5569, 5573,
+ 5581, 5591, 5623, 5639, 5641, 5647, 5651, 5653,
+ 5657, 5659, 5669, 5683, 5689, 5693, 5701, 5711,
+ 5717, 5737, 5741, 5743, 5749, 5779, 5783, 5791,
+ 5801, 5807, 5813, 5821, 5827, 5839, 5843, 5849,
+ 5851, 5857, 5861, 5867, 5869, 5879, 5881, 5897,
+ 5903, 5923, 5927, 5939, 5953, 5981, 5987, 6007,
+ 6011, 6029, 6037, 6043, 6047, 6053, 6067, 6073,
+ 6079, 6089, 6091, 6101, 6113, 6121, 6131, 6133,
+ 6143, 6151, 6163, 6173, 6197, 6199, 6203, 6211,
+ 6217, 6221, 6229, 6247, 6257, 6263, 6269, 6271,
+ 6277, 6287, 6299, 6301, 6311, 6317, 6323, 6329,
+ 6337, 6343, 6353, 6359, 6361, 6367, 6373, 6379,
+ 6389, 6397, 6421, 6427, 6449, 6451, 6469, 6473,
+ 6481, 6491, 6521, 6529, 6547, 6551, 6553, 6563,
+ 6569, 6571, 6577, 6581, 6599, 6607, 6619, 6637,
+ 6653, 6659, 6661, 6673, 6679, 6689, 6691, 6701,
+ 6703, 6709, 6719, 6733, 6737, 6761, 6763, 6779,
+ 6781, 6791, 6793, 6803, 6823, 6827, 6829, 6833,
+ 6841, 6857, 6863, 6869, 6871, 6883, 6899, 6907,
+ 6911, 6917, 6947, 6949, 6959, 6961, 6967, 6971,
+ 6977, 6983, 6991, 6997, 7001, 7013, 7019, 7027,
+ 7039, 7043, 7057, 7069, 7079, 7103, 7109, 7121,
+ 7127, 7129, 7151, 7159, 7177, 7187, 7193, 7207,
+ 7211, 7213, 7219, 7229, 7237, 7243, 7247, 7253,
+ 7283, 7297, 7307, 7309, 7321, 7331, 7333, 7349,
+ 7351, 7369, 7393, 7411, 7417, 7433, 7451, 7457,
+ 7459, 7477, 7481, 7487, 7489, 7499, 7507, 7517,
+ 7523, 7529, 7537, 7541, 7547, 7549, 7559, 7561,
+ 7573, 7577, 7583, 7589, 7591, 7603, 7607, 7621,
+ 7639, 7643, 7649, 7669, 7673, 7681, 7687, 7691,
+ 7699, 7703, 7717, 7723, 7727, 7741, 7753, 7757,
+ 7759, 7789, 7793, 7817, 7823, 7829, 7841, 7853,
+ 7867, 7873, 7877, 7879, 7883, 7901, 7907, 7919,
+ 7927, 7933, 7937, 7949, 7951, 7963, 7993, 8009,
+ 8011, 8017, 8039, 8053, 8059, 8069, 8081, 8087,
+ 8089, 8093, 8101, 8111, 8117, 8123, 8147, 8161,
+ 8167, 8171, 8179, 8191, 8209, 8219, 8221, 8231,
+ 8233, 8237, 8243, 8263, 8269, 8273, 8287, 8291,
+ 8293, 8297, 8311, 8317, 8329, 8353, 8363, 8369,
+ 8377, 8387, 8389, 8419, 8423, 8429, 8431, 8443,
+ 8447, 8461, 8467, 8501, 8513, 8521, 8527, 8537,
+ 8539, 8543, 8563, 8573, 8581, 8597, 8599, 8609,
+ 8623, 8627, 8629, 8641, 8647, 8663, 8669, 8677,
+ 8681, 8689, 8693, 8699, 8707, 8713, 8719, 8731,
+ 8737, 8741, 8747, 8753, 8761, 8779, 8783, 8803,
+ 8807, 8819, 8821, 8831, 8837, 8839, 8849, 8861,
+ 8863, 8867, 8887, 8893, 8923, 8929, 8933, 8941,
+ 8951, 8963, 8969, 8971, 8999, 9001, 9007, 9011,
+ 9013, 9029, 9041, 9043, 9049, 9059, 9067, 9091,
+ 9103, 9109, 9127, 9133, 9137, 9151, 9157, 9161,
+ 9173, 9181, 9187, 9199, 9203, 9209, 9221, 9227,
+ 9239, 9241, 9257, 9277, 9281, 9283, 9293, 9311,
+ 9319, 9323, 9337, 9341, 9343, 9349, 9371, 9377,
+ 9391, 9397, 9403, 9413, 9419, 9421, 9431, 9433,
+ 9437, 9439, 9461, 9463, 9467, 9473, 9479, 9491,
+ 9497, 9511, 9521, 9533, 9539, 9547, 9551, 9587,
+ 9601, 9613, 9619, 9623, 9629, 9631, 9643, 9649,
+ 9661, 9677, 9679, 9689, 9697, 9719, 9721, 9733,
+ 9739, 9743, 9749, 9767, 9769, 9781, 9787, 9791,
+ 9803, 9811, 9817, 9829, 9833, 9839, 9851, 9857,
+ 9859, 9871, 9883, 9887, 9901, 9907, 9923, 9929,
+ 9931, 9941, 9949, 9967, 9973, 10007, 10009, 10037,
+ 10039, 10061, 10067, 10069, 10079, 10091, 10093, 10099,
+ 10103, 10111, 10133, 10139, 10141, 10151, 10159, 10163,
+ 10169, 10177, 10181, 10193, 10211, 10223, 10243, 10247,
+ 10253, 10259, 10267, 10271, 10273, 10289, 10301, 10303,
+ 10313, 10321, 10331, 10333, 10337, 10343, 10357, 10369,
+ 10391, 10399, 10427, 10429, 10433, 10453, 10457, 10459,
+ 10463, 10477, 10487, 10499, 10501, 10513, 10529, 10531,
+ 10559, 10567, 10589, 10597, 10601, 10607, 10613, 10627,
+ 10631, 10639, 10651, 10657, 10663, 10667, 10687, 10691,
+ 10709, 10711, 10723, 10729, 10733, 10739, 10753, 10771,
+ 10781, 10789, 10799, 10831, 10837, 10847, 10853, 10859,
+ 10861, 10867, 10883, 10889, 10891, 10903, 10909, 10937,
+ 10939, 10949, 10957, 10973, 10979, 10987, 10993, 11003,
+ 11027, 11047, 11057, 11059, 11069, 11071, 11083, 11087,
+ 11093, 11113, 11117, 11119, 11131, 11149, 11159, 11161,
+ 11171, 11173, 11177, 11197, 11213, 11239, 11243, 11251,
+ 11257, 11261, 11273, 11279, 11287, 11299, 11311, 11317,
+ 11321, 11329, 11351, 11353, 11369, 11383, 11393, 11399,
+ 11411, 11423, 11437, 11443, 11447, 11467, 11471, 11483,
+ 11489, 11491, 11497, 11503, 11519, 11527, 11549, 11551,
+ 11579, 11587, 11593, 11597, 11617, 11621, 11633, 11657,
+ 11677, 11681, 11689, 11699, 11701, 11717, 11719, 11731,
+ 11743, 11777, 11779, 11783, 11789, 11801, 11807, 11813,
+ 11821, 11827, 11831, 11833, 11839, 11863, 11867, 11887,
+ 11897, 11903, 11909, 11923, 11927, 11933, 11939, 11941,
+ 11953, 11959, 11969, 11971, 11981, 11987, 12007, 12011,
+ 12037, 12041, 12043, 12049, 12071, 12073, 12097, 12101,
+ 12107, 12109, 12113, 12119, 12143, 12149, 12157, 12161,
+ 12163, 12197, 12203, 12211, 12227, 12239, 12241, 12251,
+ 12253, 12263, 12269, 12277, 12281, 12289, 12301, 12323,
+ 12329, 12343, 12347, 12373, 12377, 12379, 12391, 12401,
+ 12409, 12413, 12421, 12433, 12437, 12451, 12457, 12473,
+ 12479, 12487, 12491, 12497, 12503, 12511, 12517, 12527,
+ 12539, 12541, 12547, 12553, 12569, 12577, 12583, 12589,
+ 12601, 12611, 12613, 12619, 12637, 12641, 12647, 12653,
+ 12659, 12671, 12689, 12697, 12703, 12713, 12721, 12739,
+ 12743, 12757, 12763, 12781, 12791, 12799, 12809, 12821,
+ 12823, 12829, 12841, 12853, 12889, 12893, 12899, 12907,
+ 12911, 12917, 12919, 12923, 12941, 12953, 12959, 12967,
+ 12973, 12979, 12983, 13001, 13003, 13007, 13009, 13033,
+ 13037, 13043, 13049, 13063, 13093, 13099, 13103, 13109,
+ 13121, 13127, 13147, 13151, 13159, 13163, 13171, 13177,
+ 13183, 13187, 13217, 13219, 13229, 13241, 13249, 13259,
+ 13267, 13291, 13297, 13309, 13313, 13327, 13331, 13337,
+ 13339, 13367, 13381, 13397, 13399, 13411, 13417, 13421,
+ 13441, 13451, 13457, 13463, 13469, 13477, 13487, 13499,
+ 13513, 13523, 13537, 13553, 13567, 13577, 13591, 13597,
+ 13613, 13619, 13627, 13633, 13649, 13669, 13679, 13681,
+ 13687, 13691, 13693, 13697, 13709, 13711, 13721, 13723,
+ 13729, 13751, 13757, 13759, 13763, 13781, 13789, 13799,
+ 13807, 13829, 13831, 13841, 13859, 13873, 13877, 13879,
+ 13883, 13901, 13903, 13907, 13913, 13921, 13931, 13933,
+ 13963, 13967, 13997, 13999, 14009, 14011, 14029, 14033,
+ 14051, 14057, 14071, 14081, 14083, 14087, 14107, 14143,
+ 14149, 14153, 14159, 14173, 14177, 14197, 14207, 14221,
+ 14243, 14249, 14251, 14281, 14293, 14303, 14321, 14323,
+ 14327, 14341, 14347, 14369, 14387, 14389, 14401, 14407,
+ 14411, 14419, 14423, 14431, 14437, 14447, 14449, 14461,
+ 14479, 14489, 14503, 14519, 14533, 14537, 14543, 14549,
+ 14551, 14557, 14561, 14563, 14591, 14593, 14621, 14627,
+ 14629, 14633, 14639, 14653, 14657, 14669, 14683, 14699,
+ 14713, 14717, 14723, 14731, 14737, 14741, 14747, 14753,
+ 14759, 14767, 14771, 14779, 14783, 14797, 14813, 14821,
+ 14827, 14831, 14843, 14851, 14867, 14869, 14879, 14887,
+ 14891, 14897, 14923, 14929, 14939, 14947, 14951, 14957,
+ 14969, 14983, 15013, 15017, 15031, 15053, 15061, 15073,
+ 15077, 15083, 15091, 15101, 15107, 15121, 15131, 15137,
+ 15139, 15149, 15161, 15173, 15187, 15193, 15199, 15217,
+ 15227, 15233, 15241, 15259, 15263, 15269, 15271, 15277,
+ 15287, 15289, 15299, 15307, 15313, 15319, 15329, 15331,
+ 15349, 15359, 15361, 15373, 15377, 15383, 15391, 15401,
+ 15413, 15427, 15439, 15443, 15451, 15461, 15467, 15473,
+ 15493, 15497, 15511, 15527, 15541, 15551, 15559, 15569,
+ 15581, 15583, 15601, 15607, 15619, 15629, 15641, 15643,
+ 15647, 15649, 15661, 15667, 15671, 15679, 15683, 15727,
+ 15731, 15733, 15737, 15739, 15749, 15761, 15767, 15773,
+ 15787, 15791, 15797, 15803, 15809, 15817, 15823, 15859,
+ 15877, 15881, 15887, 15889, 15901, 15907, 15913, 15919,
+ 15923, 15937, 15959, 15971, 15973, 15991, 16001, 16007,
+ 16033, 16057, 16061, 16063, 16067, 16069, 16073, 16087,
+ 16091, 16097, 16103, 16111, 16127, 16139, 16141, 16183,
+ 16187, 16189, 16193, 16217, 16223, 16229, 16231, 16249,
+ 16253, 16267, 16273, 16301, 16319, 16333, 16339, 16349,
+ 16361, 16363, 16369, 16381, 16411, 16417, 16421, 16427,
+ 16433, 16447, 16451, 16453, 16477, 16481, 16487, 16493,
+ 16519, 16529, 16547, 16553, 16561, 16567, 16573, 16603,
+ 16607, 16619, 16631, 16633, 16649, 16651, 16657, 16661,
+ 16673, 16691, 16693, 16699, 16703, 16729, 16741, 16747,
+ 16759, 16763, 16787, 16811, 16823, 16829, 16831, 16843,
+ 16871, 16879, 16883, 16889, 16901, 16903, 16921, 16927,
+ 16931, 16937, 16943, 16963, 16979, 16981, 16987, 16993,
+ 17011, 17021, 17027, 17029, 17033, 17041, 17047, 17053,
+ 17077, 17093, 17099, 17107, 17117, 17123, 17137, 17159,
+ 17167, 17183, 17189, 17191, 17203, 17207, 17209, 17231,
+ 17239, 17257, 17291, 17293, 17299, 17317, 17321, 17327,
+ 17333, 17341, 17351, 17359, 17377, 17383, 17387, 17389,
+ 17393, 17401, 17417, 17419, 17431, 17443, 17449, 17467,
+ 17471, 17477, 17483, 17489, 17491, 17497, 17509, 17519,
+ 17539, 17551, 17569, 17573, 17579, 17581, 17597, 17599,
+ 17609, 17623, 17627, 17657, 17659, 17669, 17681, 17683,
+ 17707, 17713, 17729, 17737, 17747, 17749, 17761, 17783,
+ 17789, 17791, 17807, 17827, 17837, 17839, 17851, 17863,
+};
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_prime.pl b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_prime.pl
new file mode 100644
index 0000000..b0b1608
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_prime.pl
@@ -0,0 +1,48 @@
+#! /usr/bin/env perl
+# Copyright 1998-2019 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
+
+# Output year depends on the year of the script.
+my $YEAR = [localtime([stat($0)]->[9])]->[5] + 1900;
+print <<"EOF";
+/*
+ * WARNING: do not edit!
+ * Generated by crypto/bn/bn_prime.pl
+ *
+ * Copyright 1998-$YEAR The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+EOF
+
+
+my $num = shift || 2048;
+my @primes = ( 2 );
+my $p = 1;
+loop: while ($#primes < $num-1) {
+ $p += 2;
+ my $s = int(sqrt($p));
+
+ for (my $i = 0; defined($primes[$i]) && $primes[$i] <= $s; $i++) {
+ next loop if ($p % $primes[$i]) == 0;
+ }
+ push(@primes, $p);
+}
+
+print "typedef unsigned short prime_t;\n";
+printf "# define NUMPRIMES %d\n\n", $num;
+
+printf "static const prime_t primes[%d] = {", $num;
+for (my $i = 0; $i <= $#primes; $i++) {
+ printf "\n " if ($i % 8) == 0;
+ printf " %5d,", $primes[$i];
+}
+print "\n};\n";
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_print.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_print.c
new file mode 100644
index 0000000..17ac6e7
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_print.c
@@ -0,0 +1,345 @@
+/*
+ * Copyright 1995-2017 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include <stdio.h>
+#include "crypto/ctype.h"
+#include <limits.h>
+#include "internal/cryptlib.h"
+#include <openssl/buffer.h>
+#include "bn_local.h"
+
+static const char Hex[] = "0123456789ABCDEF";
+
+/* Must 'OPENSSL_free' the returned data */
+char *BN_bn2hex(const BIGNUM *a)
+{
+ int i, j, v, z = 0;
+ char *buf;
+ char *p;
+
+ if (BN_is_zero(a))
+ return OPENSSL_strdup("0");
+ buf = OPENSSL_malloc(a->top * BN_BYTES * 2 + 2);
+ if (buf == NULL) {
+ BNerr(BN_F_BN_BN2HEX, ERR_R_MALLOC_FAILURE);
+ goto err;
+ }
+ p = buf;
+ if (a->neg)
+ *p++ = '-';
+ for (i = a->top - 1; i >= 0; i--) {
+ for (j = BN_BITS2 - 8; j >= 0; j -= 8) {
+ /* strip leading zeros */
+ v = (int)((a->d[i] >> j) & 0xff);
+ if (z || v != 0) {
+ *p++ = Hex[v >> 4];
+ *p++ = Hex[v & 0x0f];
+ z = 1;
+ }
+ }
+ }
+ *p = '\0';
+ err:
+ return buf;
+}
+
+/* Must 'OPENSSL_free' the returned data */
+char *BN_bn2dec(const BIGNUM *a)
+{
+ int i = 0, num, ok = 0, n, tbytes;
+ char *buf = NULL;
+ char *p;
+ BIGNUM *t = NULL;
+ BN_ULONG *bn_data = NULL, *lp;
+ int bn_data_num;
+
+ /*-
+ * get an upper bound for the length of the decimal integer
+ * num <= (BN_num_bits(a) + 1) * log(2)
+ * <= 3 * BN_num_bits(a) * 0.101 + log(2) + 1 (rounding error)
+ * <= 3 * BN_num_bits(a) / 10 + 3 * BN_num_bits / 1000 + 1 + 1
+ */
+ i = BN_num_bits(a) * 3;
+ num = (i / 10 + i / 1000 + 1) + 1;
+ tbytes = num + 3; /* negative and terminator and one spare? */
+ bn_data_num = num / BN_DEC_NUM + 1;
+ bn_data = OPENSSL_malloc(bn_data_num * sizeof(BN_ULONG));
+ buf = OPENSSL_malloc(tbytes);
+ if (buf == NULL || bn_data == NULL) {
+ BNerr(BN_F_BN_BN2DEC, ERR_R_MALLOC_FAILURE);
+ goto err;
+ }
+ if ((t = BN_dup(a)) == NULL)
+ goto err;
+
+ p = buf;
+ lp = bn_data;
+ if (BN_is_zero(t)) {
+ *p++ = '0';
+ *p++ = '\0';
+ } else {
+ if (BN_is_negative(t))
+ *p++ = '-';
+
+ while (!BN_is_zero(t)) {
+ if (lp - bn_data >= bn_data_num)
+ goto err;
+ *lp = BN_div_word(t, BN_DEC_CONV);
+ if (*lp == (BN_ULONG)-1)
+ goto err;
+ lp++;
+ }
+ lp--;
+ /*
+ * We now have a series of blocks, BN_DEC_NUM chars in length, where
+ * the last one needs truncation. The blocks need to be reversed in
+ * order.
+ */
+ n = BIO_snprintf(p, tbytes - (size_t)(p - buf), BN_DEC_FMT1, *lp);
+ if (n < 0)
+ goto err;
+ p += n;
+ while (lp != bn_data) {
+ lp--;
+ n = BIO_snprintf(p, tbytes - (size_t)(p - buf), BN_DEC_FMT2, *lp);
+ if (n < 0)
+ goto err;
+ p += n;
+ }
+ }
+ ok = 1;
+ err:
+ OPENSSL_free(bn_data);
+ BN_free(t);
+ if (ok)
+ return buf;
+ OPENSSL_free(buf);
+ return NULL;
+}
+
+int BN_hex2bn(BIGNUM **bn, const char *a)
+{
+ BIGNUM *ret = NULL;
+ BN_ULONG l = 0;
+ int neg = 0, h, m, i, j, k, c;
+ int num;
+
+ if (a == NULL || *a == '\0')
+ return 0;
+
+ if (*a == '-') {
+ neg = 1;
+ a++;
+ }
+
+ for (i = 0; i <= INT_MAX / 4 && ossl_isxdigit(a[i]); i++)
+ continue;
+
+ if (i == 0 || i > INT_MAX / 4)
+ return 0;
+
+ num = i + neg;
+ if (bn == NULL)
+ return num;
+
+ /* a is the start of the hex digits, and it is 'i' long */
+ if (*bn == NULL) {
+ if ((ret = BN_new()) == NULL)
+ return 0;
+ } else {
+ ret = *bn;
+ BN_zero(ret);
+ }
+
+ /* i is the number of hex digits */
+ if (bn_expand(ret, i * 4) == NULL)
+ goto err;
+
+ j = i; /* least significant 'hex' */
+ m = 0;
+ h = 0;
+ while (j > 0) {
+ m = (BN_BYTES * 2 <= j) ? BN_BYTES * 2 : j;
+ l = 0;
+ for (;;) {
+ c = a[j - m];
+ k = OPENSSL_hexchar2int(c);
+ if (k < 0)
+ k = 0; /* paranoia */
+ l = (l << 4) | k;
+
+ if (--m <= 0) {
+ ret->d[h++] = l;
+ break;
+ }
+ }
+ j -= BN_BYTES * 2;
+ }
+ ret->top = h;
+ bn_correct_top(ret);
+
+ *bn = ret;
+ bn_check_top(ret);
+ /* Don't set the negative flag if it's zero. */
+ if (ret->top != 0)
+ ret->neg = neg;
+ return num;
+ err:
+ if (*bn == NULL)
+ BN_free(ret);
+ return 0;
+}
+
+int BN_dec2bn(BIGNUM **bn, const char *a)
+{
+ BIGNUM *ret = NULL;
+ BN_ULONG l = 0;
+ int neg = 0, i, j;
+ int num;
+
+ if (a == NULL || *a == '\0')
+ return 0;
+ if (*a == '-') {
+ neg = 1;
+ a++;
+ }
+
+ for (i = 0; i <= INT_MAX / 4 && ossl_isdigit(a[i]); i++)
+ continue;
+
+ if (i == 0 || i > INT_MAX / 4)
+ goto err;
+
+ num = i + neg;
+ if (bn == NULL)
+ return num;
+
+ /*
+ * a is the start of the digits, and it is 'i' long. We chop it into
+ * BN_DEC_NUM digits at a time
+ */
+ if (*bn == NULL) {
+ if ((ret = BN_new()) == NULL)
+ return 0;
+ } else {
+ ret = *bn;
+ BN_zero(ret);
+ }
+
+ /* i is the number of digits, a bit of an over expand */
+ if (bn_expand(ret, i * 4) == NULL)
+ goto err;
+
+ j = BN_DEC_NUM - i % BN_DEC_NUM;
+ if (j == BN_DEC_NUM)
+ j = 0;
+ l = 0;
+ while (--i >= 0) {
+ l *= 10;
+ l += *a - '0';
+ a++;
+ if (++j == BN_DEC_NUM) {
+ if (!BN_mul_word(ret, BN_DEC_CONV)
+ || !BN_add_word(ret, l))
+ goto err;
+ l = 0;
+ j = 0;
+ }
+ }
+
+ bn_correct_top(ret);
+ *bn = ret;
+ bn_check_top(ret);
+ /* Don't set the negative flag if it's zero. */
+ if (ret->top != 0)
+ ret->neg = neg;
+ return num;
+ err:
+ if (*bn == NULL)
+ BN_free(ret);
+ return 0;
+}
+
+int BN_asc2bn(BIGNUM **bn, const char *a)
+{
+ const char *p = a;
+
+ if (*p == '-')
+ p++;
+
+ if (p[0] == '0' && (p[1] == 'X' || p[1] == 'x')) {
+ if (!BN_hex2bn(bn, p + 2))
+ return 0;
+ } else {
+ if (!BN_dec2bn(bn, p))
+ return 0;
+ }
+ /* Don't set the negative flag if it's zero. */
+ if (*a == '-' && (*bn)->top != 0)
+ (*bn)->neg = 1;
+ return 1;
+}
+
+# ifndef OPENSSL_NO_STDIO
+int BN_print_fp(FILE *fp, const BIGNUM *a)
+{
+ BIO *b;
+ int ret;
+
+ if ((b = BIO_new(BIO_s_file())) == NULL)
+ return 0;
+ BIO_set_fp(b, fp, BIO_NOCLOSE);
+ ret = BN_print(b, a);
+ BIO_free(b);
+ return ret;
+}
+# endif
+
+int BN_print(BIO *bp, const BIGNUM *a)
+{
+ int i, j, v, z = 0;
+ int ret = 0;
+
+ if ((a->neg) && BIO_write(bp, "-", 1) != 1)
+ goto end;
+ if (BN_is_zero(a) && BIO_write(bp, "0", 1) != 1)
+ goto end;
+ for (i = a->top - 1; i >= 0; i--) {
+ for (j = BN_BITS2 - 4; j >= 0; j -= 4) {
+ /* strip leading zeros */
+ v = (int)((a->d[i] >> j) & 0x0f);
+ if (z || v != 0) {
+ if (BIO_write(bp, &Hex[v], 1) != 1)
+ goto end;
+ z = 1;
+ }
+ }
+ }
+ ret = 1;
+ end:
+ return ret;
+}
+
+char *BN_options(void)
+{
+ static int init = 0;
+ static char data[16];
+
+ if (!init) {
+ init++;
+#ifdef BN_LLONG
+ BIO_snprintf(data, sizeof(data), "bn(%zu,%zu)",
+ sizeof(BN_ULLONG) * 8, sizeof(BN_ULONG) * 8);
+#else
+ BIO_snprintf(data, sizeof(data), "bn(%zu,%zu)",
+ sizeof(BN_ULONG) * 8, sizeof(BN_ULONG) * 8);
+#endif
+ }
+ return data;
+}
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_rand.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_rand.c
new file mode 100644
index 0000000..6b4b50a
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_rand.c
@@ -0,0 +1,265 @@
+/*
+ * Copyright 1995-2019 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include <stdio.h>
+#include <time.h>
+#include "internal/cryptlib.h"
+#include "bn_local.h"
+#include <openssl/rand.h>
+#include <openssl/sha.h>
+
+typedef enum bnrand_flag_e {
+ NORMAL, TESTING, PRIVATE
+} BNRAND_FLAG;
+
+static int bnrand(BNRAND_FLAG flag, BIGNUM *rnd, int bits, int top, int bottom)
+{
+ unsigned char *buf = NULL;
+ int b, ret = 0, bit, bytes, mask;
+
+ if (bits == 0) {
+ if (top != BN_RAND_TOP_ANY || bottom != BN_RAND_BOTTOM_ANY)
+ goto toosmall;
+ BN_zero(rnd);
+ return 1;
+ }
+ if (bits < 0 || (bits == 1 && top > 0))
+ goto toosmall;
+
+ bytes = (bits + 7) / 8;
+ bit = (bits - 1) % 8;
+ mask = 0xff << (bit + 1);
+
+ buf = OPENSSL_malloc(bytes);
+ if (buf == NULL) {
+ BNerr(BN_F_BNRAND, ERR_R_MALLOC_FAILURE);
+ goto err;
+ }
+
+ /* make a random number and set the top and bottom bits */
+ b = flag == NORMAL ? RAND_bytes(buf, bytes) : RAND_priv_bytes(buf, bytes);
+ if (b <= 0)
+ goto err;
+
+ if (flag == TESTING) {
+ /*
+ * generate patterns that are more likely to trigger BN library bugs
+ */
+ int i;
+ unsigned char c;
+
+ for (i = 0; i < bytes; i++) {
+ if (RAND_bytes(&c, 1) <= 0)
+ goto err;
+ if (c >= 128 && i > 0)
+ buf[i] = buf[i - 1];
+ else if (c < 42)
+ buf[i] = 0;
+ else if (c < 84)
+ buf[i] = 255;
+ }
+ }
+
+ if (top >= 0) {
+ if (top) {
+ if (bit == 0) {
+ buf[0] = 1;
+ buf[1] |= 0x80;
+ } else {
+ buf[0] |= (3 << (bit - 1));
+ }
+ } else {
+ buf[0] |= (1 << bit);
+ }
+ }
+ buf[0] &= ~mask;
+ if (bottom) /* set bottom bit if requested */
+ buf[bytes - 1] |= 1;
+ if (!BN_bin2bn(buf, bytes, rnd))
+ goto err;
+ ret = 1;
+ err:
+ OPENSSL_clear_free(buf, bytes);
+ bn_check_top(rnd);
+ return ret;
+
+toosmall:
+ BNerr(BN_F_BNRAND, BN_R_BITS_TOO_SMALL);
+ return 0;
+}
+
+int BN_rand(BIGNUM *rnd, int bits, int top, int bottom)
+{
+ return bnrand(NORMAL, rnd, bits, top, bottom);
+}
+
+int BN_bntest_rand(BIGNUM *rnd, int bits, int top, int bottom)
+{
+ return bnrand(TESTING, rnd, bits, top, bottom);
+}
+
+int BN_priv_rand(BIGNUM *rnd, int bits, int top, int bottom)
+{
+ return bnrand(PRIVATE, rnd, bits, top, bottom);
+}
+
+/* random number r: 0 <= r < range */
+static int bnrand_range(BNRAND_FLAG flag, BIGNUM *r, const BIGNUM *range)
+{
+ int n;
+ int count = 100;
+
+ if (range->neg || BN_is_zero(range)) {
+ BNerr(BN_F_BNRAND_RANGE, BN_R_INVALID_RANGE);
+ return 0;
+ }
+
+ n = BN_num_bits(range); /* n > 0 */
+
+ /* BN_is_bit_set(range, n - 1) always holds */
+
+ if (n == 1)
+ BN_zero(r);
+ else if (!BN_is_bit_set(range, n - 2) && !BN_is_bit_set(range, n - 3)) {
+ /*
+ * range = 100..._2, so 3*range (= 11..._2) is exactly one bit longer
+ * than range
+ */
+ do {
+ if (!bnrand(flag, r, n + 1, BN_RAND_TOP_ANY, BN_RAND_BOTTOM_ANY))
+ return 0;
+
+ /*
+ * If r < 3*range, use r := r MOD range (which is either r, r -
+ * range, or r - 2*range). Otherwise, iterate once more. Since
+ * 3*range = 11..._2, each iteration succeeds with probability >=
+ * .75.
+ */
+ if (BN_cmp(r, range) >= 0) {
+ if (!BN_sub(r, r, range))
+ return 0;
+ if (BN_cmp(r, range) >= 0)
+ if (!BN_sub(r, r, range))
+ return 0;
+ }
+
+ if (!--count) {
+ BNerr(BN_F_BNRAND_RANGE, BN_R_TOO_MANY_ITERATIONS);
+ return 0;
+ }
+
+ }
+ while (BN_cmp(r, range) >= 0);
+ } else {
+ do {
+ /* range = 11..._2 or range = 101..._2 */
+ if (!bnrand(flag, r, n, BN_RAND_TOP_ANY, BN_RAND_BOTTOM_ANY))
+ return 0;
+
+ if (!--count) {
+ BNerr(BN_F_BNRAND_RANGE, BN_R_TOO_MANY_ITERATIONS);
+ return 0;
+ }
+ }
+ while (BN_cmp(r, range) >= 0);
+ }
+
+ bn_check_top(r);
+ return 1;
+}
+
+int BN_rand_range(BIGNUM *r, const BIGNUM *range)
+{
+ return bnrand_range(NORMAL, r, range);
+}
+
+int BN_priv_rand_range(BIGNUM *r, const BIGNUM *range)
+{
+ return bnrand_range(PRIVATE, r, range);
+}
+
+int BN_pseudo_rand(BIGNUM *rnd, int bits, int top, int bottom)
+{
+ return BN_rand(rnd, bits, top, bottom);
+}
+
+int BN_pseudo_rand_range(BIGNUM *r, const BIGNUM *range)
+{
+ return BN_rand_range(r, range);
+}
+
+/*
+ * BN_generate_dsa_nonce generates a random number 0 <= out < range. Unlike
+ * BN_rand_range, it also includes the contents of |priv| and |message| in
+ * the generation so that an RNG failure isn't fatal as long as |priv|
+ * remains secret. This is intended for use in DSA and ECDSA where an RNG
+ * weakness leads directly to private key exposure unless this function is
+ * used.
+ */
+int BN_generate_dsa_nonce(BIGNUM *out, const BIGNUM *range,
+ const BIGNUM *priv, const unsigned char *message,
+ size_t message_len, BN_CTX *ctx)
+{
+ SHA512_CTX sha;
+ /*
+ * We use 512 bits of random data per iteration to ensure that we have at
+ * least |range| bits of randomness.
+ */
+ unsigned char random_bytes[64];
+ unsigned char digest[SHA512_DIGEST_LENGTH];
+ unsigned done, todo;
+ /* We generate |range|+8 bytes of random output. */
+ const unsigned num_k_bytes = BN_num_bytes(range) + 8;
+ unsigned char private_bytes[96];
+ unsigned char *k_bytes;
+ int ret = 0;
+
+ k_bytes = OPENSSL_malloc(num_k_bytes);
+ if (k_bytes == NULL)
+ goto err;
+
+ /* We copy |priv| into a local buffer to avoid exposing its length. */
+ if (BN_bn2binpad(priv, private_bytes, sizeof(private_bytes)) < 0) {
+ /*
+ * No reasonable DSA or ECDSA key should have a private key this
+ * large and we don't handle this case in order to avoid leaking the
+ * length of the private key.
+ */
+ BNerr(BN_F_BN_GENERATE_DSA_NONCE, BN_R_PRIVATE_KEY_TOO_LARGE);
+ goto err;
+ }
+
+ for (done = 0; done < num_k_bytes;) {
+ if (RAND_priv_bytes(random_bytes, sizeof(random_bytes)) != 1)
+ goto err;
+ SHA512_Init(&sha);
+ SHA512_Update(&sha, &done, sizeof(done));
+ SHA512_Update(&sha, private_bytes, sizeof(private_bytes));
+ SHA512_Update(&sha, message, message_len);
+ SHA512_Update(&sha, random_bytes, sizeof(random_bytes));
+ SHA512_Final(digest, &sha);
+
+ todo = num_k_bytes - done;
+ if (todo > SHA512_DIGEST_LENGTH)
+ todo = SHA512_DIGEST_LENGTH;
+ memcpy(k_bytes + done, digest, todo);
+ done += todo;
+ }
+
+ if (!BN_bin2bn(k_bytes, num_k_bytes, out))
+ goto err;
+ if (BN_mod(out, out, range, ctx) != 1)
+ goto err;
+ ret = 1;
+
+ err:
+ OPENSSL_free(k_bytes);
+ OPENSSL_cleanse(private_bytes, sizeof(private_bytes));
+ return ret;
+}
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_recp.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_recp.c
new file mode 100644
index 0000000..e822313
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_recp.c
@@ -0,0 +1,194 @@
+/*
+ * Copyright 1995-2018 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include "internal/cryptlib.h"
+#include "bn_local.h"
+
+void BN_RECP_CTX_init(BN_RECP_CTX *recp)
+{
+ memset(recp, 0, sizeof(*recp));
+ bn_init(&(recp->N));
+ bn_init(&(recp->Nr));
+}
+
+BN_RECP_CTX *BN_RECP_CTX_new(void)
+{
+ BN_RECP_CTX *ret;
+
+ if ((ret = OPENSSL_zalloc(sizeof(*ret))) == NULL) {
+ BNerr(BN_F_BN_RECP_CTX_NEW, ERR_R_MALLOC_FAILURE);
+ return NULL;
+ }
+
+ bn_init(&(ret->N));
+ bn_init(&(ret->Nr));
+ ret->flags = BN_FLG_MALLOCED;
+ return ret;
+}
+
+void BN_RECP_CTX_free(BN_RECP_CTX *recp)
+{
+ if (recp == NULL)
+ return;
+ BN_free(&recp->N);
+ BN_free(&recp->Nr);
+ if (recp->flags & BN_FLG_MALLOCED)
+ OPENSSL_free(recp);
+}
+
+int BN_RECP_CTX_set(BN_RECP_CTX *recp, const BIGNUM *d, BN_CTX *ctx)
+{
+ if (!BN_copy(&(recp->N), d))
+ return 0;
+ BN_zero(&(recp->Nr));
+ recp->num_bits = BN_num_bits(d);
+ recp->shift = 0;
+ return 1;
+}
+
+int BN_mod_mul_reciprocal(BIGNUM *r, const BIGNUM *x, const BIGNUM *y,
+ BN_RECP_CTX *recp, BN_CTX *ctx)
+{
+ int ret = 0;
+ BIGNUM *a;
+ const BIGNUM *ca;
+
+ BN_CTX_start(ctx);
+ if ((a = BN_CTX_get(ctx)) == NULL)
+ goto err;
+ if (y != NULL) {
+ if (x == y) {
+ if (!BN_sqr(a, x, ctx))
+ goto err;
+ } else {
+ if (!BN_mul(a, x, y, ctx))
+ goto err;
+ }
+ ca = a;
+ } else
+ ca = x; /* Just do the mod */
+
+ ret = BN_div_recp(NULL, r, ca, recp, ctx);
+ err:
+ BN_CTX_end(ctx);
+ bn_check_top(r);
+ return ret;
+}
+
+int BN_div_recp(BIGNUM *dv, BIGNUM *rem, const BIGNUM *m,
+ BN_RECP_CTX *recp, BN_CTX *ctx)
+{
+ int i, j, ret = 0;
+ BIGNUM *a, *b, *d, *r;
+
+ BN_CTX_start(ctx);
+ d = (dv != NULL) ? dv : BN_CTX_get(ctx);
+ r = (rem != NULL) ? rem : BN_CTX_get(ctx);
+ a = BN_CTX_get(ctx);
+ b = BN_CTX_get(ctx);
+ if (b == NULL)
+ goto err;
+
+ if (BN_ucmp(m, &(recp->N)) < 0) {
+ BN_zero(d);
+ if (!BN_copy(r, m)) {
+ BN_CTX_end(ctx);
+ return 0;
+ }
+ BN_CTX_end(ctx);
+ return 1;
+ }
+
+ /*
+ * We want the remainder Given input of ABCDEF / ab we need multiply
+ * ABCDEF by 3 digests of the reciprocal of ab
+ */
+
+ /* i := max(BN_num_bits(m), 2*BN_num_bits(N)) */
+ i = BN_num_bits(m);
+ j = recp->num_bits << 1;
+ if (j > i)
+ i = j;
+
+ /* Nr := round(2^i / N) */
+ if (i != recp->shift)
+ recp->shift = BN_reciprocal(&(recp->Nr), &(recp->N), i, ctx);
+ /* BN_reciprocal could have returned -1 for an error */
+ if (recp->shift == -1)
+ goto err;
+
+ /*-
+ * d := |round(round(m / 2^BN_num_bits(N)) * recp->Nr / 2^(i - BN_num_bits(N)))|
+ * = |round(round(m / 2^BN_num_bits(N)) * round(2^i / N) / 2^(i - BN_num_bits(N)))|
+ * <= |(m / 2^BN_num_bits(N)) * (2^i / N) * (2^BN_num_bits(N) / 2^i)|
+ * = |m/N|
+ */
+ if (!BN_rshift(a, m, recp->num_bits))
+ goto err;
+ if (!BN_mul(b, a, &(recp->Nr), ctx))
+ goto err;
+ if (!BN_rshift(d, b, i - recp->num_bits))
+ goto err;
+ d->neg = 0;
+
+ if (!BN_mul(b, &(recp->N), d, ctx))
+ goto err;
+ if (!BN_usub(r, m, b))
+ goto err;
+ r->neg = 0;
+
+ j = 0;
+ while (BN_ucmp(r, &(recp->N)) >= 0) {
+ if (j++ > 2) {
+ BNerr(BN_F_BN_DIV_RECP, BN_R_BAD_RECIPROCAL);
+ goto err;
+ }
+ if (!BN_usub(r, r, &(recp->N)))
+ goto err;
+ if (!BN_add_word(d, 1))
+ goto err;
+ }
+
+ r->neg = BN_is_zero(r) ? 0 : m->neg;
+ d->neg = m->neg ^ recp->N.neg;
+ ret = 1;
+ err:
+ BN_CTX_end(ctx);
+ bn_check_top(dv);
+ bn_check_top(rem);
+ return ret;
+}
+
+/*
+ * len is the expected size of the result We actually calculate with an extra
+ * word of precision, so we can do faster division if the remainder is not
+ * required.
+ */
+/* r := 2^len / m */
+int BN_reciprocal(BIGNUM *r, const BIGNUM *m, int len, BN_CTX *ctx)
+{
+ int ret = -1;
+ BIGNUM *t;
+
+ BN_CTX_start(ctx);
+ if ((t = BN_CTX_get(ctx)) == NULL)
+ goto err;
+
+ if (!BN_set_bit(t, len))
+ goto err;
+
+ if (!BN_div(r, NULL, t, m, ctx))
+ goto err;
+
+ ret = len;
+ err:
+ bn_check_top(r);
+ BN_CTX_end(ctx);
+ return ret;
+}
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_shift.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_shift.c
new file mode 100644
index 0000000..210a83f
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_shift.c
@@ -0,0 +1,216 @@
+/*
+ * Copyright 1995-2018 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include <assert.h>
+#include "internal/cryptlib.h"
+#include "bn_local.h"
+
+int BN_lshift1(BIGNUM *r, const BIGNUM *a)
+{
+ register BN_ULONG *ap, *rp, t, c;
+ int i;
+
+ bn_check_top(r);
+ bn_check_top(a);
+
+ if (r != a) {
+ r->neg = a->neg;
+ if (bn_wexpand(r, a->top + 1) == NULL)
+ return 0;
+ r->top = a->top;
+ } else {
+ if (bn_wexpand(r, a->top + 1) == NULL)
+ return 0;
+ }
+ ap = a->d;
+ rp = r->d;
+ c = 0;
+ for (i = 0; i < a->top; i++) {
+ t = *(ap++);
+ *(rp++) = ((t << 1) | c) & BN_MASK2;
+ c = t >> (BN_BITS2 - 1);
+ }
+ *rp = c;
+ r->top += c;
+ bn_check_top(r);
+ return 1;
+}
+
+int BN_rshift1(BIGNUM *r, const BIGNUM *a)
+{
+ BN_ULONG *ap, *rp, t, c;
+ int i;
+
+ bn_check_top(r);
+ bn_check_top(a);
+
+ if (BN_is_zero(a)) {
+ BN_zero(r);
+ return 1;
+ }
+ i = a->top;
+ ap = a->d;
+ if (a != r) {
+ if (bn_wexpand(r, i) == NULL)
+ return 0;
+ r->neg = a->neg;
+ }
+ rp = r->d;
+ r->top = i;
+ t = ap[--i];
+ rp[i] = t >> 1;
+ c = t << (BN_BITS2 - 1);
+ r->top -= (t == 1);
+ while (i > 0) {
+ t = ap[--i];
+ rp[i] = ((t >> 1) & BN_MASK2) | c;
+ c = t << (BN_BITS2 - 1);
+ }
+ if (!r->top)
+ r->neg = 0; /* don't allow negative zero */
+ bn_check_top(r);
+ return 1;
+}
+
+int BN_lshift(BIGNUM *r, const BIGNUM *a, int n)
+{
+ int ret;
+
+ if (n < 0) {
+ BNerr(BN_F_BN_LSHIFT, BN_R_INVALID_SHIFT);
+ return 0;
+ }
+
+ ret = bn_lshift_fixed_top(r, a, n);
+
+ bn_correct_top(r);
+ bn_check_top(r);
+
+ return ret;
+}
+
+/*
+ * In respect to shift factor the execution time is invariant of
+ * |n % BN_BITS2|, but not |n / BN_BITS2|. Or in other words pre-condition
+ * for constant-time-ness is |n < BN_BITS2| or |n / BN_BITS2| being
+ * non-secret.
+ */
+int bn_lshift_fixed_top(BIGNUM *r, const BIGNUM *a, int n)
+{
+ int i, nw;
+ unsigned int lb, rb;
+ BN_ULONG *t, *f;
+ BN_ULONG l, m, rmask = 0;
+
+ assert(n >= 0);
+
+ bn_check_top(r);
+ bn_check_top(a);
+
+ nw = n / BN_BITS2;
+ if (bn_wexpand(r, a->top + nw + 1) == NULL)
+ return 0;
+
+ if (a->top != 0) {
+ lb = (unsigned int)n % BN_BITS2;
+ rb = BN_BITS2 - lb;
+ rb %= BN_BITS2; /* say no to undefined behaviour */
+ rmask = (BN_ULONG)0 - rb; /* rmask = 0 - (rb != 0) */
+ rmask |= rmask >> 8;
+ f = &(a->d[0]);
+ t = &(r->d[nw]);
+ l = f[a->top - 1];
+ t[a->top] = (l >> rb) & rmask;
+ for (i = a->top - 1; i > 0; i--) {
+ m = l << lb;
+ l = f[i - 1];
+ t[i] = (m | ((l >> rb) & rmask)) & BN_MASK2;
+ }
+ t[0] = (l << lb) & BN_MASK2;
+ } else {
+ /* shouldn't happen, but formally required */
+ r->d[nw] = 0;
+ }
+ if (nw != 0)
+ memset(r->d, 0, sizeof(*t) * nw);
+
+ r->neg = a->neg;
+ r->top = a->top + nw + 1;
+ r->flags |= BN_FLG_FIXED_TOP;
+
+ return 1;
+}
+
+int BN_rshift(BIGNUM *r, const BIGNUM *a, int n)
+{
+ int ret = 0;
+
+ if (n < 0) {
+ BNerr(BN_F_BN_RSHIFT, BN_R_INVALID_SHIFT);
+ return 0;
+ }
+
+ ret = bn_rshift_fixed_top(r, a, n);
+
+ bn_correct_top(r);
+ bn_check_top(r);
+
+ return ret;
+}
+
+/*
+ * In respect to shift factor the execution time is invariant of
+ * |n % BN_BITS2|, but not |n / BN_BITS2|. Or in other words pre-condition
+ * for constant-time-ness for sufficiently[!] zero-padded inputs is
+ * |n < BN_BITS2| or |n / BN_BITS2| being non-secret.
+ */
+int bn_rshift_fixed_top(BIGNUM *r, const BIGNUM *a, int n)
+{
+ int i, top, nw;
+ unsigned int lb, rb;
+ BN_ULONG *t, *f;
+ BN_ULONG l, m, mask;
+
+ bn_check_top(r);
+ bn_check_top(a);
+
+ assert(n >= 0);
+
+ nw = n / BN_BITS2;
+ if (nw >= a->top) {
+ /* shouldn't happen, but formally required */
+ BN_zero(r);
+ return 1;
+ }
+
+ rb = (unsigned int)n % BN_BITS2;
+ lb = BN_BITS2 - rb;
+ lb %= BN_BITS2; /* say no to undefined behaviour */
+ mask = (BN_ULONG)0 - lb; /* mask = 0 - (lb != 0) */
+ mask |= mask >> 8;
+ top = a->top - nw;
+ if (r != a && bn_wexpand(r, top) == NULL)
+ return 0;
+
+ t = &(r->d[0]);
+ f = &(a->d[nw]);
+ l = f[0];
+ for (i = 0; i < top - 1; i++) {
+ m = f[i + 1];
+ t[i] = (l >> rb) | ((m << lb) & mask);
+ l = m;
+ }
+ t[i] = l >> rb;
+
+ r->neg = a->neg;
+ r->top = top;
+ r->flags |= BN_FLG_FIXED_TOP;
+
+ return 1;
+}
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_sqr.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_sqr.c
new file mode 100644
index 0000000..7f3a179
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_sqr.c
@@ -0,0 +1,239 @@
+/*
+ * Copyright 1995-2018 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include "internal/cryptlib.h"
+#include "bn_local.h"
+
+/* r must not be a */
+/*
+ * I've just gone over this and it is now %20 faster on x86 - eay - 27 Jun 96
+ */
+int BN_sqr(BIGNUM *r, const BIGNUM *a, BN_CTX *ctx)
+{
+ int ret = bn_sqr_fixed_top(r, a, ctx);
+
+ bn_correct_top(r);
+ bn_check_top(r);
+
+ return ret;
+}
+
+int bn_sqr_fixed_top(BIGNUM *r, const BIGNUM *a, BN_CTX *ctx)
+{
+ int max, al;
+ int ret = 0;
+ BIGNUM *tmp, *rr;
+
+ bn_check_top(a);
+
+ al = a->top;
+ if (al <= 0) {
+ r->top = 0;
+ r->neg = 0;
+ return 1;
+ }
+
+ BN_CTX_start(ctx);
+ rr = (a != r) ? r : BN_CTX_get(ctx);
+ tmp = BN_CTX_get(ctx);
+ if (rr == NULL || tmp == NULL)
+ goto err;
+
+ max = 2 * al; /* Non-zero (from above) */
+ if (bn_wexpand(rr, max) == NULL)
+ goto err;
+
+ if (al == 4) {
+#ifndef BN_SQR_COMBA
+ BN_ULONG t[8];
+ bn_sqr_normal(rr->d, a->d, 4, t);
+#else
+ bn_sqr_comba4(rr->d, a->d);
+#endif
+ } else if (al == 8) {
+#ifndef BN_SQR_COMBA
+ BN_ULONG t[16];
+ bn_sqr_normal(rr->d, a->d, 8, t);
+#else
+ bn_sqr_comba8(rr->d, a->d);
+#endif
+ } else {
+#if defined(BN_RECURSION)
+ if (al < BN_SQR_RECURSIVE_SIZE_NORMAL) {
+ BN_ULONG t[BN_SQR_RECURSIVE_SIZE_NORMAL * 2];
+ bn_sqr_normal(rr->d, a->d, al, t);
+ } else {
+ int j, k;
+
+ j = BN_num_bits_word((BN_ULONG)al);
+ j = 1 << (j - 1);
+ k = j + j;
+ if (al == j) {
+ if (bn_wexpand(tmp, k * 2) == NULL)
+ goto err;
+ bn_sqr_recursive(rr->d, a->d, al, tmp->d);
+ } else {
+ if (bn_wexpand(tmp, max) == NULL)
+ goto err;
+ bn_sqr_normal(rr->d, a->d, al, tmp->d);
+ }
+ }
+#else
+ if (bn_wexpand(tmp, max) == NULL)
+ goto err;
+ bn_sqr_normal(rr->d, a->d, al, tmp->d);
+#endif
+ }
+
+ rr->neg = 0;
+ rr->top = max;
+ rr->flags |= BN_FLG_FIXED_TOP;
+ if (r != rr && BN_copy(r, rr) == NULL)
+ goto err;
+
+ ret = 1;
+ err:
+ bn_check_top(rr);
+ bn_check_top(tmp);
+ BN_CTX_end(ctx);
+ return ret;
+}
+
+/* tmp must have 2*n words */
+void bn_sqr_normal(BN_ULONG *r, const BN_ULONG *a, int n, BN_ULONG *tmp)
+{
+ int i, j, max;
+ const BN_ULONG *ap;
+ BN_ULONG *rp;
+
+ max = n * 2;
+ ap = a;
+ rp = r;
+ rp[0] = rp[max - 1] = 0;
+ rp++;
+ j = n;
+
+ if (--j > 0) {
+ ap++;
+ rp[j] = bn_mul_words(rp, ap, j, ap[-1]);
+ rp += 2;
+ }
+
+ for (i = n - 2; i > 0; i--) {
+ j--;
+ ap++;
+ rp[j] = bn_mul_add_words(rp, ap, j, ap[-1]);
+ rp += 2;
+ }
+
+ bn_add_words(r, r, r, max);
+
+ /* There will not be a carry */
+
+ bn_sqr_words(tmp, a, n);
+
+ bn_add_words(r, r, tmp, max);
+}
+
+#ifdef BN_RECURSION
+/*-
+ * r is 2*n words in size,
+ * a and b are both n words in size. (There's not actually a 'b' here ...)
+ * n must be a power of 2.
+ * We multiply and return the result.
+ * t must be 2*n words in size
+ * We calculate
+ * a[0]*b[0]
+ * a[0]*b[0]+a[1]*b[1]+(a[0]-a[1])*(b[1]-b[0])
+ * a[1]*b[1]
+ */
+void bn_sqr_recursive(BN_ULONG *r, const BN_ULONG *a, int n2, BN_ULONG *t)
+{
+ int n = n2 / 2;
+ int zero, c1;
+ BN_ULONG ln, lo, *p;
+
+ if (n2 == 4) {
+# ifndef BN_SQR_COMBA
+ bn_sqr_normal(r, a, 4, t);
+# else
+ bn_sqr_comba4(r, a);
+# endif
+ return;
+ } else if (n2 == 8) {
+# ifndef BN_SQR_COMBA
+ bn_sqr_normal(r, a, 8, t);
+# else
+ bn_sqr_comba8(r, a);
+# endif
+ return;
+ }
+ if (n2 < BN_SQR_RECURSIVE_SIZE_NORMAL) {
+ bn_sqr_normal(r, a, n2, t);
+ return;
+ }
+ /* r=(a[0]-a[1])*(a[1]-a[0]) */
+ c1 = bn_cmp_words(a, &(a[n]), n);
+ zero = 0;
+ if (c1 > 0)
+ bn_sub_words(t, a, &(a[n]), n);
+ else if (c1 < 0)
+ bn_sub_words(t, &(a[n]), a, n);
+ else
+ zero = 1;
+
+ /* The result will always be negative unless it is zero */
+ p = &(t[n2 * 2]);
+
+ if (!zero)
+ bn_sqr_recursive(&(t[n2]), t, n, p);
+ else
+ memset(&t[n2], 0, sizeof(*t) * n2);
+ bn_sqr_recursive(r, a, n, p);
+ bn_sqr_recursive(&(r[n2]), &(a[n]), n, p);
+
+ /*-
+ * t[32] holds (a[0]-a[1])*(a[1]-a[0]), it is negative or zero
+ * r[10] holds (a[0]*b[0])
+ * r[32] holds (b[1]*b[1])
+ */
+
+ c1 = (int)(bn_add_words(t, r, &(r[n2]), n2));
+
+ /* t[32] is negative */
+ c1 -= (int)(bn_sub_words(&(t[n2]), t, &(t[n2]), n2));
+
+ /*-
+ * t[32] holds (a[0]-a[1])*(a[1]-a[0])+(a[0]*a[0])+(a[1]*a[1])
+ * r[10] holds (a[0]*a[0])
+ * r[32] holds (a[1]*a[1])
+ * c1 holds the carry bits
+ */
+ c1 += (int)(bn_add_words(&(r[n]), &(r[n]), &(t[n2]), n2));
+ if (c1) {
+ p = &(r[n + n2]);
+ lo = *p;
+ ln = (lo + c1) & BN_MASK2;
+ *p = ln;
+
+ /*
+ * The overflow will stop before we over write words we should not
+ * overwrite
+ */
+ if (ln < (BN_ULONG)c1) {
+ do {
+ p++;
+ lo = *p;
+ ln = (lo + 1) & BN_MASK2;
+ *p = ln;
+ } while (ln == 0);
+ }
+ }
+}
+#endif
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_sqrt.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_sqrt.c
new file mode 100644
index 0000000..6a42ce8
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_sqrt.c
@@ -0,0 +1,365 @@
+/*
+ * Copyright 2000-2022 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include "internal/cryptlib.h"
+#include "bn_local.h"
+
+BIGNUM *BN_mod_sqrt(BIGNUM *in, const BIGNUM *a, const BIGNUM *p, BN_CTX *ctx)
+/*
+ * Returns 'ret' such that ret^2 == a (mod p), using the Tonelli/Shanks
+ * algorithm (cf. Henri Cohen, "A Course in Algebraic Computational Number
+ * Theory", algorithm 1.5.1). 'p' must be prime, otherwise an error or
+ * an incorrect "result" will be returned.
+ */
+{
+ BIGNUM *ret = in;
+ int err = 1;
+ int r;
+ BIGNUM *A, *b, *q, *t, *x, *y;
+ int e, i, j;
+
+ if (!BN_is_odd(p) || BN_abs_is_word(p, 1)) {
+ if (BN_abs_is_word(p, 2)) {
+ if (ret == NULL)
+ ret = BN_new();
+ if (ret == NULL)
+ goto end;
+ if (!BN_set_word(ret, BN_is_bit_set(a, 0))) {
+ if (ret != in)
+ BN_free(ret);
+ return NULL;
+ }
+ bn_check_top(ret);
+ return ret;
+ }
+
+ BNerr(BN_F_BN_MOD_SQRT, BN_R_P_IS_NOT_PRIME);
+ return NULL;
+ }
+
+ if (BN_is_zero(a) || BN_is_one(a)) {
+ if (ret == NULL)
+ ret = BN_new();
+ if (ret == NULL)
+ goto end;
+ if (!BN_set_word(ret, BN_is_one(a))) {
+ if (ret != in)
+ BN_free(ret);
+ return NULL;
+ }
+ bn_check_top(ret);
+ return ret;
+ }
+
+ BN_CTX_start(ctx);
+ A = BN_CTX_get(ctx);
+ b = BN_CTX_get(ctx);
+ q = BN_CTX_get(ctx);
+ t = BN_CTX_get(ctx);
+ x = BN_CTX_get(ctx);
+ y = BN_CTX_get(ctx);
+ if (y == NULL)
+ goto end;
+
+ if (ret == NULL)
+ ret = BN_new();
+ if (ret == NULL)
+ goto end;
+
+ /* A = a mod p */
+ if (!BN_nnmod(A, a, p, ctx))
+ goto end;
+
+ /* now write |p| - 1 as 2^e*q where q is odd */
+ e = 1;
+ while (!BN_is_bit_set(p, e))
+ e++;
+ /* we'll set q later (if needed) */
+
+ if (e == 1) {
+ /*-
+ * The easy case: (|p|-1)/2 is odd, so 2 has an inverse
+ * modulo (|p|-1)/2, and square roots can be computed
+ * directly by modular exponentiation.
+ * We have
+ * 2 * (|p|+1)/4 == 1 (mod (|p|-1)/2),
+ * so we can use exponent (|p|+1)/4, i.e. (|p|-3)/4 + 1.
+ */
+ if (!BN_rshift(q, p, 2))
+ goto end;
+ q->neg = 0;
+ if (!BN_add_word(q, 1))
+ goto end;
+ if (!BN_mod_exp(ret, A, q, p, ctx))
+ goto end;
+ err = 0;
+ goto vrfy;
+ }
+
+ if (e == 2) {
+ /*-
+ * |p| == 5 (mod 8)
+ *
+ * In this case 2 is always a non-square since
+ * Legendre(2,p) = (-1)^((p^2-1)/8) for any odd prime.
+ * So if a really is a square, then 2*a is a non-square.
+ * Thus for
+ * b := (2*a)^((|p|-5)/8),
+ * i := (2*a)*b^2
+ * we have
+ * i^2 = (2*a)^((1 + (|p|-5)/4)*2)
+ * = (2*a)^((p-1)/2)
+ * = -1;
+ * so if we set
+ * x := a*b*(i-1),
+ * then
+ * x^2 = a^2 * b^2 * (i^2 - 2*i + 1)
+ * = a^2 * b^2 * (-2*i)
+ * = a*(-i)*(2*a*b^2)
+ * = a*(-i)*i
+ * = a.
+ *
+ * (This is due to A.O.L. Atkin,
+ * Subject: Square Roots and Cognate Matters modulo p=8n+5.
+ * URL: https://listserv.nodak.edu/cgi-bin/wa.exe?A2=ind9211&L=NMBRTHRY&P=4026
+ * November 1992.)
+ */
+
+ /* t := 2*a */
+ if (!BN_mod_lshift1_quick(t, A, p))
+ goto end;
+
+ /* b := (2*a)^((|p|-5)/8) */
+ if (!BN_rshift(q, p, 3))
+ goto end;
+ q->neg = 0;
+ if (!BN_mod_exp(b, t, q, p, ctx))
+ goto end;
+
+ /* y := b^2 */
+ if (!BN_mod_sqr(y, b, p, ctx))
+ goto end;
+
+ /* t := (2*a)*b^2 - 1 */
+ if (!BN_mod_mul(t, t, y, p, ctx))
+ goto end;
+ if (!BN_sub_word(t, 1))
+ goto end;
+
+ /* x = a*b*t */
+ if (!BN_mod_mul(x, A, b, p, ctx))
+ goto end;
+ if (!BN_mod_mul(x, x, t, p, ctx))
+ goto end;
+
+ if (!BN_copy(ret, x))
+ goto end;
+ err = 0;
+ goto vrfy;
+ }
+
+ /*
+ * e > 2, so we really have to use the Tonelli/Shanks algorithm. First,
+ * find some y that is not a square.
+ */
+ if (!BN_copy(q, p))
+ goto end; /* use 'q' as temp */
+ q->neg = 0;
+ i = 2;
+ do {
+ /*
+ * For efficiency, try small numbers first; if this fails, try random
+ * numbers.
+ */
+ if (i < 22) {
+ if (!BN_set_word(y, i))
+ goto end;
+ } else {
+ if (!BN_priv_rand(y, BN_num_bits(p), 0, 0))
+ goto end;
+ if (BN_ucmp(y, p) >= 0) {
+ if (!(p->neg ? BN_add : BN_sub) (y, y, p))
+ goto end;
+ }
+ /* now 0 <= y < |p| */
+ if (BN_is_zero(y))
+ if (!BN_set_word(y, i))
+ goto end;
+ }
+
+ r = BN_kronecker(y, q, ctx); /* here 'q' is |p| */
+ if (r < -1)
+ goto end;
+ if (r == 0) {
+ /* m divides p */
+ BNerr(BN_F_BN_MOD_SQRT, BN_R_P_IS_NOT_PRIME);
+ goto end;
+ }
+ }
+ while (r == 1 && ++i < 82);
+
+ if (r != -1) {
+ /*
+ * Many rounds and still no non-square -- this is more likely a bug
+ * than just bad luck. Even if p is not prime, we should have found
+ * some y such that r == -1.
+ */
+ BNerr(BN_F_BN_MOD_SQRT, BN_R_TOO_MANY_ITERATIONS);
+ goto end;
+ }
+
+ /* Here's our actual 'q': */
+ if (!BN_rshift(q, q, e))
+ goto end;
+
+ /*
+ * Now that we have some non-square, we can find an element of order 2^e
+ * by computing its q'th power.
+ */
+ if (!BN_mod_exp(y, y, q, p, ctx))
+ goto end;
+ if (BN_is_one(y)) {
+ BNerr(BN_F_BN_MOD_SQRT, BN_R_P_IS_NOT_PRIME);
+ goto end;
+ }
+
+ /*-
+ * Now we know that (if p is indeed prime) there is an integer
+ * k, 0 <= k < 2^e, such that
+ *
+ * a^q * y^k == 1 (mod p).
+ *
+ * As a^q is a square and y is not, k must be even.
+ * q+1 is even, too, so there is an element
+ *
+ * X := a^((q+1)/2) * y^(k/2),
+ *
+ * and it satisfies
+ *
+ * X^2 = a^q * a * y^k
+ * = a,
+ *
+ * so it is the square root that we are looking for.
+ */
+
+ /* t := (q-1)/2 (note that q is odd) */
+ if (!BN_rshift1(t, q))
+ goto end;
+
+ /* x := a^((q-1)/2) */
+ if (BN_is_zero(t)) { /* special case: p = 2^e + 1 */
+ if (!BN_nnmod(t, A, p, ctx))
+ goto end;
+ if (BN_is_zero(t)) {
+ /* special case: a == 0 (mod p) */
+ BN_zero(ret);
+ err = 0;
+ goto end;
+ } else if (!BN_one(x))
+ goto end;
+ } else {
+ if (!BN_mod_exp(x, A, t, p, ctx))
+ goto end;
+ if (BN_is_zero(x)) {
+ /* special case: a == 0 (mod p) */
+ BN_zero(ret);
+ err = 0;
+ goto end;
+ }
+ }
+
+ /* b := a*x^2 (= a^q) */
+ if (!BN_mod_sqr(b, x, p, ctx))
+ goto end;
+ if (!BN_mod_mul(b, b, A, p, ctx))
+ goto end;
+
+ /* x := a*x (= a^((q+1)/2)) */
+ if (!BN_mod_mul(x, x, A, p, ctx))
+ goto end;
+
+ while (1) {
+ /*-
+ * Now b is a^q * y^k for some even k (0 <= k < 2^E
+ * where E refers to the original value of e, which we
+ * don't keep in a variable), and x is a^((q+1)/2) * y^(k/2).
+ *
+ * We have a*b = x^2,
+ * y^2^(e-1) = -1,
+ * b^2^(e-1) = 1.
+ */
+
+ if (BN_is_one(b)) {
+ if (!BN_copy(ret, x))
+ goto end;
+ err = 0;
+ goto vrfy;
+ }
+
+ /* Find the smallest i, 0 < i < e, such that b^(2^i) = 1. */
+ for (i = 1; i < e; i++) {
+ if (i == 1) {
+ if (!BN_mod_sqr(t, b, p, ctx))
+ goto end;
+
+ } else {
+ if (!BN_mod_mul(t, t, t, p, ctx))
+ goto end;
+ }
+ if (BN_is_one(t))
+ break;
+ }
+ /* If not found, a is not a square or p is not prime. */
+ if (i >= e) {
+ BNerr(BN_F_BN_MOD_SQRT, BN_R_NOT_A_SQUARE);
+ goto end;
+ }
+
+ /* t := y^2^(e - i - 1) */
+ if (!BN_copy(t, y))
+ goto end;
+ for (j = e - i - 1; j > 0; j--) {
+ if (!BN_mod_sqr(t, t, p, ctx))
+ goto end;
+ }
+ if (!BN_mod_mul(y, t, t, p, ctx))
+ goto end;
+ if (!BN_mod_mul(x, x, t, p, ctx))
+ goto end;
+ if (!BN_mod_mul(b, b, y, p, ctx))
+ goto end;
+ e = i;
+ }
+
+ vrfy:
+ if (!err) {
+ /*
+ * verify the result -- the input might have been not a square (test
+ * added in 0.9.8)
+ */
+
+ if (!BN_mod_sqr(x, ret, p, ctx))
+ err = 1;
+
+ if (!err && 0 != BN_cmp(x, A)) {
+ BNerr(BN_F_BN_MOD_SQRT, BN_R_NOT_A_SQUARE);
+ err = 1;
+ }
+ }
+
+ end:
+ if (err) {
+ if (ret != in)
+ BN_clear_free(ret);
+ ret = NULL;
+ }
+ BN_CTX_end(ctx);
+ bn_check_top(ret);
+ return ret;
+}
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_srp.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_srp.c
new file mode 100644
index 0000000..820757b
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_srp.c
@@ -0,0 +1,545 @@
+/*
+ * Copyright 2014-2017 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include "bn_local.h"
+#include "internal/nelem.h"
+
+#ifndef OPENSSL_NO_SRP
+
+#include <openssl/srp.h>
+#include "crypto/bn_srp.h"
+
+# if (BN_BYTES == 8)
+# if (defined(_WIN32) || defined(_WIN64)) && !defined(__MINGW32__)
+# define bn_pack4(a1,a2,a3,a4) ((a1##UI64<<48)|(a2##UI64<<32)|(a3##UI64<<16)|a4##UI64)
+# elif defined(__arch64__)
+# define bn_pack4(a1,a2,a3,a4) ((a1##UL<<48)|(a2##UL<<32)|(a3##UL<<16)|a4##UL)
+# else
+# define bn_pack4(a1,a2,a3,a4) ((a1##ULL<<48)|(a2##ULL<<32)|(a3##ULL<<16)|a4##ULL)
+# endif
+# elif (BN_BYTES == 4)
+# define bn_pack4(a1,a2,a3,a4) ((a3##UL<<16)|a4##UL), ((a1##UL<<16)|a2##UL)
+# else
+# error "unsupported BN_BYTES"
+# endif
+
+static const BN_ULONG bn_group_1024_value[] = {
+ bn_pack4(0x9FC6, 0x1D2F, 0xC0EB, 0x06E3),
+ bn_pack4(0xFD51, 0x38FE, 0x8376, 0x435B),
+ bn_pack4(0x2FD4, 0xCBF4, 0x976E, 0xAA9A),
+ bn_pack4(0x68ED, 0xBC3C, 0x0572, 0x6CC0),
+ bn_pack4(0xC529, 0xF566, 0x660E, 0x57EC),
+ bn_pack4(0x8255, 0x9B29, 0x7BCF, 0x1885),
+ bn_pack4(0xCE8E, 0xF4AD, 0x69B1, 0x5D49),
+ bn_pack4(0x5DC7, 0xD7B4, 0x6154, 0xD6B6),
+ bn_pack4(0x8E49, 0x5C1D, 0x6089, 0xDAD1),
+ bn_pack4(0xE0D5, 0xD8E2, 0x50B9, 0x8BE4),
+ bn_pack4(0x383B, 0x4813, 0xD692, 0xC6E0),
+ bn_pack4(0xD674, 0xDF74, 0x96EA, 0x81D3),
+ bn_pack4(0x9EA2, 0x314C, 0x9C25, 0x6576),
+ bn_pack4(0x6072, 0x6187, 0x75FF, 0x3C0B),
+ bn_pack4(0x9C33, 0xF80A, 0xFA8F, 0xC5E8),
+ bn_pack4(0xEEAF, 0x0AB9, 0xADB3, 0x8DD6)
+};
+
+const BIGNUM bn_group_1024 = {
+ (BN_ULONG *)bn_group_1024_value,
+ OSSL_NELEM(bn_group_1024_value),
+ OSSL_NELEM(bn_group_1024_value),
+ 0,
+ BN_FLG_STATIC_DATA
+};
+
+static const BN_ULONG bn_group_1536_value[] = {
+ bn_pack4(0xCF76, 0xE3FE, 0xD135, 0xF9BB),
+ bn_pack4(0x1518, 0x0F93, 0x499A, 0x234D),
+ bn_pack4(0x8CE7, 0xA28C, 0x2442, 0xC6F3),
+ bn_pack4(0x5A02, 0x1FFF, 0x5E91, 0x479E),
+ bn_pack4(0x7F8A, 0x2FE9, 0xB8B5, 0x292E),
+ bn_pack4(0x837C, 0x264A, 0xE3A9, 0xBEB8),
+ bn_pack4(0xE442, 0x734A, 0xF7CC, 0xB7AE),
+ bn_pack4(0x6577, 0x2E43, 0x7D6C, 0x7F8C),
+ bn_pack4(0xDB2F, 0xD53D, 0x24B7, 0xC486),
+ bn_pack4(0x6EDF, 0x0195, 0x3934, 0x9627),
+ bn_pack4(0x158B, 0xFD3E, 0x2B9C, 0x8CF5),
+ bn_pack4(0x764E, 0x3F4B, 0x53DD, 0x9DA1),
+ bn_pack4(0x4754, 0x8381, 0xDBC5, 0xB1FC),
+ bn_pack4(0x9B60, 0x9E0B, 0xE3BA, 0xB63D),
+ bn_pack4(0x8134, 0xB1C8, 0xB979, 0x8914),
+ bn_pack4(0xDF02, 0x8A7C, 0xEC67, 0xF0D0),
+ bn_pack4(0x80B6, 0x55BB, 0x9A22, 0xE8DC),
+ bn_pack4(0x1558, 0x903B, 0xA0D0, 0xF843),
+ bn_pack4(0x51C6, 0xA94B, 0xE460, 0x7A29),
+ bn_pack4(0x5F4F, 0x5F55, 0x6E27, 0xCBDE),
+ bn_pack4(0xBEEE, 0xA961, 0x4B19, 0xCC4D),
+ bn_pack4(0xDBA5, 0x1DF4, 0x99AC, 0x4C80),
+ bn_pack4(0xB1F1, 0x2A86, 0x17A4, 0x7BBB),
+ bn_pack4(0x9DEF, 0x3CAF, 0xB939, 0x277A)
+};
+
+const BIGNUM bn_group_1536 = {
+ (BN_ULONG *)bn_group_1536_value,
+ OSSL_NELEM(bn_group_1536_value),
+ OSSL_NELEM(bn_group_1536_value),
+ 0,
+ BN_FLG_STATIC_DATA
+};
+
+static const BN_ULONG bn_group_2048_value[] = {
+ bn_pack4(0x0FA7, 0x111F, 0x9E4A, 0xFF73),
+ bn_pack4(0x9B65, 0xE372, 0xFCD6, 0x8EF2),
+ bn_pack4(0x35DE, 0x236D, 0x525F, 0x5475),
+ bn_pack4(0x94B5, 0xC803, 0xD89F, 0x7AE4),
+ bn_pack4(0x71AE, 0x35F8, 0xE9DB, 0xFBB6),
+ bn_pack4(0x2A56, 0x98F3, 0xA8D0, 0xC382),
+ bn_pack4(0x9CCC, 0x041C, 0x7BC3, 0x08D8),
+ bn_pack4(0xAF87, 0x4E73, 0x03CE, 0x5329),
+ bn_pack4(0x6160, 0x2790, 0x04E5, 0x7AE6),
+ bn_pack4(0x032C, 0xFBDB, 0xF52F, 0xB378),
+ bn_pack4(0x5EA7, 0x7A27, 0x75D2, 0xECFA),
+ bn_pack4(0x5445, 0x23B5, 0x24B0, 0xD57D),
+ bn_pack4(0x5B9D, 0x32E6, 0x88F8, 0x7748),
+ bn_pack4(0xF1D2, 0xB907, 0x8717, 0x461A),
+ bn_pack4(0x76BD, 0x207A, 0x436C, 0x6481),
+ bn_pack4(0xCA97, 0xB43A, 0x23FB, 0x8016),
+ bn_pack4(0x1D28, 0x1E44, 0x6B14, 0x773B),
+ bn_pack4(0x7359, 0xD041, 0xD5C3, 0x3EA7),
+ bn_pack4(0xA80D, 0x740A, 0xDBF4, 0xFF74),
+ bn_pack4(0x55F9, 0x7993, 0xEC97, 0x5EEA),
+ bn_pack4(0x2918, 0xA996, 0x2F0B, 0x93B8),
+ bn_pack4(0x661A, 0x05FB, 0xD5FA, 0xAAE8),
+ bn_pack4(0xCF60, 0x9517, 0x9A16, 0x3AB3),
+ bn_pack4(0xE808, 0x3969, 0xEDB7, 0x67B0),
+ bn_pack4(0xCD7F, 0x48A9, 0xDA04, 0xFD50),
+ bn_pack4(0xD523, 0x12AB, 0x4B03, 0x310D),
+ bn_pack4(0x8193, 0xE075, 0x7767, 0xA13D),
+ bn_pack4(0xA373, 0x29CB, 0xB4A0, 0x99ED),
+ bn_pack4(0xFC31, 0x9294, 0x3DB5, 0x6050),
+ bn_pack4(0xAF72, 0xB665, 0x1987, 0xEE07),
+ bn_pack4(0xF166, 0xDE5E, 0x1389, 0x582F),
+ bn_pack4(0xAC6B, 0xDB41, 0x324A, 0x9A9B)
+};
+
+const BIGNUM bn_group_2048 = {
+ (BN_ULONG *)bn_group_2048_value,
+ OSSL_NELEM(bn_group_2048_value),
+ OSSL_NELEM(bn_group_2048_value),
+ 0,
+ BN_FLG_STATIC_DATA
+};
+
+static const BN_ULONG bn_group_3072_value[] = {
+ bn_pack4(0xFFFF, 0xFFFF, 0xFFFF, 0xFFFF),
+ bn_pack4(0x4B82, 0xD120, 0xA93A, 0xD2CA),
+ bn_pack4(0x43DB, 0x5BFC, 0xE0FD, 0x108E),
+ bn_pack4(0x08E2, 0x4FA0, 0x74E5, 0xAB31),
+ bn_pack4(0x7709, 0x88C0, 0xBAD9, 0x46E2),
+ bn_pack4(0xBBE1, 0x1757, 0x7A61, 0x5D6C),
+ bn_pack4(0x521F, 0x2B18, 0x177B, 0x200C),
+ bn_pack4(0xD876, 0x0273, 0x3EC8, 0x6A64),
+ bn_pack4(0xF12F, 0xFA06, 0xD98A, 0x0864),
+ bn_pack4(0xCEE3, 0xD226, 0x1AD2, 0xEE6B),
+ bn_pack4(0x1E8C, 0x94E0, 0x4A25, 0x619D),
+ bn_pack4(0xABF5, 0xAE8C, 0xDB09, 0x33D7),
+ bn_pack4(0xB397, 0x0F85, 0xA6E1, 0xE4C7),
+ bn_pack4(0x8AEA, 0x7157, 0x5D06, 0x0C7D),
+ bn_pack4(0xECFB, 0x8504, 0x58DB, 0xEF0A),
+ bn_pack4(0xA855, 0x21AB, 0xDF1C, 0xBA64),
+ bn_pack4(0xAD33, 0x170D, 0x0450, 0x7A33),
+ bn_pack4(0x1572, 0x8E5A, 0x8AAA, 0xC42D),
+ bn_pack4(0x15D2, 0x2618, 0x98FA, 0x0510),
+ bn_pack4(0x3995, 0x497C, 0xEA95, 0x6AE5),
+ bn_pack4(0xDE2B, 0xCBF6, 0x9558, 0x1718),
+ bn_pack4(0xB5C5, 0x5DF0, 0x6F4C, 0x52C9),
+ bn_pack4(0x9B27, 0x83A2, 0xEC07, 0xA28F),
+ bn_pack4(0xE39E, 0x772C, 0x180E, 0x8603),
+ bn_pack4(0x3290, 0x5E46, 0x2E36, 0xCE3B),
+ bn_pack4(0xF174, 0x6C08, 0xCA18, 0x217C),
+ bn_pack4(0x670C, 0x354E, 0x4ABC, 0x9804),
+ bn_pack4(0x9ED5, 0x2907, 0x7096, 0x966D),
+ bn_pack4(0x1C62, 0xF356, 0x2085, 0x52BB),
+ bn_pack4(0x8365, 0x5D23, 0xDCA3, 0xAD96),
+ bn_pack4(0x6916, 0x3FA8, 0xFD24, 0xCF5F),
+ bn_pack4(0x98DA, 0x4836, 0x1C55, 0xD39A),
+ bn_pack4(0xC200, 0x7CB8, 0xA163, 0xBF05),
+ bn_pack4(0x4928, 0x6651, 0xECE4, 0x5B3D),
+ bn_pack4(0xAE9F, 0x2411, 0x7C4B, 0x1FE6),
+ bn_pack4(0xEE38, 0x6BFB, 0x5A89, 0x9FA5),
+ bn_pack4(0x0BFF, 0x5CB6, 0xF406, 0xB7ED),
+ bn_pack4(0xF44C, 0x42E9, 0xA637, 0xED6B),
+ bn_pack4(0xE485, 0xB576, 0x625E, 0x7EC6),
+ bn_pack4(0x4FE1, 0x356D, 0x6D51, 0xC245),
+ bn_pack4(0x302B, 0x0A6D, 0xF25F, 0x1437),
+ bn_pack4(0xEF95, 0x19B3, 0xCD3A, 0x431B),
+ bn_pack4(0x514A, 0x0879, 0x8E34, 0x04DD),
+ bn_pack4(0x020B, 0xBEA6, 0x3B13, 0x9B22),
+ bn_pack4(0x2902, 0x4E08, 0x8A67, 0xCC74),
+ bn_pack4(0xC4C6, 0x628B, 0x80DC, 0x1CD1),
+ bn_pack4(0xC90F, 0xDAA2, 0x2168, 0xC234),
+ bn_pack4(0xFFFF, 0xFFFF, 0xFFFF, 0xFFFF)
+};
+
+const BIGNUM bn_group_3072 = {
+ (BN_ULONG *)bn_group_3072_value,
+ OSSL_NELEM(bn_group_3072_value),
+ OSSL_NELEM(bn_group_3072_value),
+ 0,
+ BN_FLG_STATIC_DATA
+};
+
+static const BN_ULONG bn_group_4096_value[] = {
+ bn_pack4(0xFFFF, 0xFFFF, 0xFFFF, 0xFFFF),
+ bn_pack4(0x4DF4, 0x35C9, 0x3406, 0x3199),
+ bn_pack4(0x86FF, 0xB7DC, 0x90A6, 0xC08F),
+ bn_pack4(0x93B4, 0xEA98, 0x8D8F, 0xDDC1),
+ bn_pack4(0xD006, 0x9127, 0xD5B0, 0x5AA9),
+ bn_pack4(0xB81B, 0xDD76, 0x2170, 0x481C),
+ bn_pack4(0x1F61, 0x2970, 0xCEE2, 0xD7AF),
+ bn_pack4(0x233B, 0xA186, 0x515B, 0xE7ED),
+ bn_pack4(0x99B2, 0x964F, 0xA090, 0xC3A2),
+ bn_pack4(0x287C, 0x5947, 0x4E6B, 0xC05D),
+ bn_pack4(0x2E8E, 0xFC14, 0x1FBE, 0xCAA6),
+ bn_pack4(0xDBBB, 0xC2DB, 0x04DE, 0x8EF9),
+ bn_pack4(0x2583, 0xE9CA, 0x2AD4, 0x4CE8),
+ bn_pack4(0x1A94, 0x6834, 0xB615, 0x0BDA),
+ bn_pack4(0x99C3, 0x2718, 0x6AF4, 0xE23C),
+ bn_pack4(0x8871, 0x9A10, 0xBDBA, 0x5B26),
+ bn_pack4(0x1A72, 0x3C12, 0xA787, 0xE6D7),
+ bn_pack4(0x4B82, 0xD120, 0xA921, 0x0801),
+ bn_pack4(0x43DB, 0x5BFC, 0xE0FD, 0x108E),
+ bn_pack4(0x08E2, 0x4FA0, 0x74E5, 0xAB31),
+ bn_pack4(0x7709, 0x88C0, 0xBAD9, 0x46E2),
+ bn_pack4(0xBBE1, 0x1757, 0x7A61, 0x5D6C),
+ bn_pack4(0x521F, 0x2B18, 0x177B, 0x200C),
+ bn_pack4(0xD876, 0x0273, 0x3EC8, 0x6A64),
+ bn_pack4(0xF12F, 0xFA06, 0xD98A, 0x0864),
+ bn_pack4(0xCEE3, 0xD226, 0x1AD2, 0xEE6B),
+ bn_pack4(0x1E8C, 0x94E0, 0x4A25, 0x619D),
+ bn_pack4(0xABF5, 0xAE8C, 0xDB09, 0x33D7),
+ bn_pack4(0xB397, 0x0F85, 0xA6E1, 0xE4C7),
+ bn_pack4(0x8AEA, 0x7157, 0x5D06, 0x0C7D),
+ bn_pack4(0xECFB, 0x8504, 0x58DB, 0xEF0A),
+ bn_pack4(0xA855, 0x21AB, 0xDF1C, 0xBA64),
+ bn_pack4(0xAD33, 0x170D, 0x0450, 0x7A33),
+ bn_pack4(0x1572, 0x8E5A, 0x8AAA, 0xC42D),
+ bn_pack4(0x15D2, 0x2618, 0x98FA, 0x0510),
+ bn_pack4(0x3995, 0x497C, 0xEA95, 0x6AE5),
+ bn_pack4(0xDE2B, 0xCBF6, 0x9558, 0x1718),
+ bn_pack4(0xB5C5, 0x5DF0, 0x6F4C, 0x52C9),
+ bn_pack4(0x9B27, 0x83A2, 0xEC07, 0xA28F),
+ bn_pack4(0xE39E, 0x772C, 0x180E, 0x8603),
+ bn_pack4(0x3290, 0x5E46, 0x2E36, 0xCE3B),
+ bn_pack4(0xF174, 0x6C08, 0xCA18, 0x217C),
+ bn_pack4(0x670C, 0x354E, 0x4ABC, 0x9804),
+ bn_pack4(0x9ED5, 0x2907, 0x7096, 0x966D),
+ bn_pack4(0x1C62, 0xF356, 0x2085, 0x52BB),
+ bn_pack4(0x8365, 0x5D23, 0xDCA3, 0xAD96),
+ bn_pack4(0x6916, 0x3FA8, 0xFD24, 0xCF5F),
+ bn_pack4(0x98DA, 0x4836, 0x1C55, 0xD39A),
+ bn_pack4(0xC200, 0x7CB8, 0xA163, 0xBF05),
+ bn_pack4(0x4928, 0x6651, 0xECE4, 0x5B3D),
+ bn_pack4(0xAE9F, 0x2411, 0x7C4B, 0x1FE6),
+ bn_pack4(0xEE38, 0x6BFB, 0x5A89, 0x9FA5),
+ bn_pack4(0x0BFF, 0x5CB6, 0xF406, 0xB7ED),
+ bn_pack4(0xF44C, 0x42E9, 0xA637, 0xED6B),
+ bn_pack4(0xE485, 0xB576, 0x625E, 0x7EC6),
+ bn_pack4(0x4FE1, 0x356D, 0x6D51, 0xC245),
+ bn_pack4(0x302B, 0x0A6D, 0xF25F, 0x1437),
+ bn_pack4(0xEF95, 0x19B3, 0xCD3A, 0x431B),
+ bn_pack4(0x514A, 0x0879, 0x8E34, 0x04DD),
+ bn_pack4(0x020B, 0xBEA6, 0x3B13, 0x9B22),
+ bn_pack4(0x2902, 0x4E08, 0x8A67, 0xCC74),
+ bn_pack4(0xC4C6, 0x628B, 0x80DC, 0x1CD1),
+ bn_pack4(0xC90F, 0xDAA2, 0x2168, 0xC234),
+ bn_pack4(0xFFFF, 0xFFFF, 0xFFFF, 0xFFFF)
+};
+
+const BIGNUM bn_group_4096 = {
+ (BN_ULONG *)bn_group_4096_value,
+ OSSL_NELEM(bn_group_4096_value),
+ OSSL_NELEM(bn_group_4096_value),
+ 0,
+ BN_FLG_STATIC_DATA
+};
+
+static const BN_ULONG bn_group_6144_value[] = {
+ bn_pack4(0xFFFF, 0xFFFF, 0xFFFF, 0xFFFF),
+ bn_pack4(0xE694, 0xF91E, 0x6DCC, 0x4024),
+ bn_pack4(0x12BF, 0x2D5B, 0x0B74, 0x74D6),
+ bn_pack4(0x043E, 0x8F66, 0x3F48, 0x60EE),
+ bn_pack4(0x387F, 0xE8D7, 0x6E3C, 0x0468),
+ bn_pack4(0xDA56, 0xC9EC, 0x2EF2, 0x9632),
+ bn_pack4(0xEB19, 0xCCB1, 0xA313, 0xD55C),
+ bn_pack4(0xF550, 0xAA3D, 0x8A1F, 0xBFF0),
+ bn_pack4(0x06A1, 0xD58B, 0xB7C5, 0xDA76),
+ bn_pack4(0xA797, 0x15EE, 0xF29B, 0xE328),
+ bn_pack4(0x14CC, 0x5ED2, 0x0F80, 0x37E0),
+ bn_pack4(0xCC8F, 0x6D7E, 0xBF48, 0xE1D8),
+ bn_pack4(0x4BD4, 0x07B2, 0x2B41, 0x54AA),
+ bn_pack4(0x0F1D, 0x45B7, 0xFF58, 0x5AC5),
+ bn_pack4(0x23A9, 0x7A7E, 0x36CC, 0x88BE),
+ bn_pack4(0x59E7, 0xC97F, 0xBEC7, 0xE8F3),
+ bn_pack4(0xB5A8, 0x4031, 0x900B, 0x1C9E),
+ bn_pack4(0xD55E, 0x702F, 0x4698, 0x0C82),
+ bn_pack4(0xF482, 0xD7CE, 0x6E74, 0xFEF6),
+ bn_pack4(0xF032, 0xEA15, 0xD172, 0x1D03),
+ bn_pack4(0x5983, 0xCA01, 0xC64B, 0x92EC),
+ bn_pack4(0x6FB8, 0xF401, 0x378C, 0xD2BF),
+ bn_pack4(0x3320, 0x5151, 0x2BD7, 0xAF42),
+ bn_pack4(0xDB7F, 0x1447, 0xE6CC, 0x254B),
+ bn_pack4(0x44CE, 0x6CBA, 0xCED4, 0xBB1B),
+ bn_pack4(0xDA3E, 0xDBEB, 0xCF9B, 0x14ED),
+ bn_pack4(0x1797, 0x27B0, 0x865A, 0x8918),
+ bn_pack4(0xB06A, 0x53ED, 0x9027, 0xD831),
+ bn_pack4(0xE5DB, 0x382F, 0x4130, 0x01AE),
+ bn_pack4(0xF8FF, 0x9406, 0xAD9E, 0x530E),
+ bn_pack4(0xC975, 0x1E76, 0x3DBA, 0x37BD),
+ bn_pack4(0xC1D4, 0xDCB2, 0x6026, 0x46DE),
+ bn_pack4(0x36C3, 0xFAB4, 0xD27C, 0x7026),
+ bn_pack4(0x4DF4, 0x35C9, 0x3402, 0x8492),
+ bn_pack4(0x86FF, 0xB7DC, 0x90A6, 0xC08F),
+ bn_pack4(0x93B4, 0xEA98, 0x8D8F, 0xDDC1),
+ bn_pack4(0xD006, 0x9127, 0xD5B0, 0x5AA9),
+ bn_pack4(0xB81B, 0xDD76, 0x2170, 0x481C),
+ bn_pack4(0x1F61, 0x2970, 0xCEE2, 0xD7AF),
+ bn_pack4(0x233B, 0xA186, 0x515B, 0xE7ED),
+ bn_pack4(0x99B2, 0x964F, 0xA090, 0xC3A2),
+ bn_pack4(0x287C, 0x5947, 0x4E6B, 0xC05D),
+ bn_pack4(0x2E8E, 0xFC14, 0x1FBE, 0xCAA6),
+ bn_pack4(0xDBBB, 0xC2DB, 0x04DE, 0x8EF9),
+ bn_pack4(0x2583, 0xE9CA, 0x2AD4, 0x4CE8),
+ bn_pack4(0x1A94, 0x6834, 0xB615, 0x0BDA),
+ bn_pack4(0x99C3, 0x2718, 0x6AF4, 0xE23C),
+ bn_pack4(0x8871, 0x9A10, 0xBDBA, 0x5B26),
+ bn_pack4(0x1A72, 0x3C12, 0xA787, 0xE6D7),
+ bn_pack4(0x4B82, 0xD120, 0xA921, 0x0801),
+ bn_pack4(0x43DB, 0x5BFC, 0xE0FD, 0x108E),
+ bn_pack4(0x08E2, 0x4FA0, 0x74E5, 0xAB31),
+ bn_pack4(0x7709, 0x88C0, 0xBAD9, 0x46E2),
+ bn_pack4(0xBBE1, 0x1757, 0x7A61, 0x5D6C),
+ bn_pack4(0x521F, 0x2B18, 0x177B, 0x200C),
+ bn_pack4(0xD876, 0x0273, 0x3EC8, 0x6A64),
+ bn_pack4(0xF12F, 0xFA06, 0xD98A, 0x0864),
+ bn_pack4(0xCEE3, 0xD226, 0x1AD2, 0xEE6B),
+ bn_pack4(0x1E8C, 0x94E0, 0x4A25, 0x619D),
+ bn_pack4(0xABF5, 0xAE8C, 0xDB09, 0x33D7),
+ bn_pack4(0xB397, 0x0F85, 0xA6E1, 0xE4C7),
+ bn_pack4(0x8AEA, 0x7157, 0x5D06, 0x0C7D),
+ bn_pack4(0xECFB, 0x8504, 0x58DB, 0xEF0A),
+ bn_pack4(0xA855, 0x21AB, 0xDF1C, 0xBA64),
+ bn_pack4(0xAD33, 0x170D, 0x0450, 0x7A33),
+ bn_pack4(0x1572, 0x8E5A, 0x8AAA, 0xC42D),
+ bn_pack4(0x15D2, 0x2618, 0x98FA, 0x0510),
+ bn_pack4(0x3995, 0x497C, 0xEA95, 0x6AE5),
+ bn_pack4(0xDE2B, 0xCBF6, 0x9558, 0x1718),
+ bn_pack4(0xB5C5, 0x5DF0, 0x6F4C, 0x52C9),
+ bn_pack4(0x9B27, 0x83A2, 0xEC07, 0xA28F),
+ bn_pack4(0xE39E, 0x772C, 0x180E, 0x8603),
+ bn_pack4(0x3290, 0x5E46, 0x2E36, 0xCE3B),
+ bn_pack4(0xF174, 0x6C08, 0xCA18, 0x217C),
+ bn_pack4(0x670C, 0x354E, 0x4ABC, 0x9804),
+ bn_pack4(0x9ED5, 0x2907, 0x7096, 0x966D),
+ bn_pack4(0x1C62, 0xF356, 0x2085, 0x52BB),
+ bn_pack4(0x8365, 0x5D23, 0xDCA3, 0xAD96),
+ bn_pack4(0x6916, 0x3FA8, 0xFD24, 0xCF5F),
+ bn_pack4(0x98DA, 0x4836, 0x1C55, 0xD39A),
+ bn_pack4(0xC200, 0x7CB8, 0xA163, 0xBF05),
+ bn_pack4(0x4928, 0x6651, 0xECE4, 0x5B3D),
+ bn_pack4(0xAE9F, 0x2411, 0x7C4B, 0x1FE6),
+ bn_pack4(0xEE38, 0x6BFB, 0x5A89, 0x9FA5),
+ bn_pack4(0x0BFF, 0x5CB6, 0xF406, 0xB7ED),
+ bn_pack4(0xF44C, 0x42E9, 0xA637, 0xED6B),
+ bn_pack4(0xE485, 0xB576, 0x625E, 0x7EC6),
+ bn_pack4(0x4FE1, 0x356D, 0x6D51, 0xC245),
+ bn_pack4(0x302B, 0x0A6D, 0xF25F, 0x1437),
+ bn_pack4(0xEF95, 0x19B3, 0xCD3A, 0x431B),
+ bn_pack4(0x514A, 0x0879, 0x8E34, 0x04DD),
+ bn_pack4(0x020B, 0xBEA6, 0x3B13, 0x9B22),
+ bn_pack4(0x2902, 0x4E08, 0x8A67, 0xCC74),
+ bn_pack4(0xC4C6, 0x628B, 0x80DC, 0x1CD1),
+ bn_pack4(0xC90F, 0xDAA2, 0x2168, 0xC234),
+ bn_pack4(0xFFFF, 0xFFFF, 0xFFFF, 0xFFFF)
+};
+
+const BIGNUM bn_group_6144 = {
+ (BN_ULONG *)bn_group_6144_value,
+ OSSL_NELEM(bn_group_6144_value),
+ OSSL_NELEM(bn_group_6144_value),
+ 0,
+ BN_FLG_STATIC_DATA
+};
+
+static const BN_ULONG bn_group_8192_value[] = {
+ bn_pack4(0xFFFF, 0xFFFF, 0xFFFF, 0xFFFF),
+ bn_pack4(0x60C9, 0x80DD, 0x98ED, 0xD3DF),
+ bn_pack4(0xC81F, 0x56E8, 0x80B9, 0x6E71),
+ bn_pack4(0x9E30, 0x50E2, 0x7656, 0x94DF),
+ bn_pack4(0x9558, 0xE447, 0x5677, 0xE9AA),
+ bn_pack4(0xC919, 0x0DA6, 0xFC02, 0x6E47),
+ bn_pack4(0x889A, 0x002E, 0xD5EE, 0x382B),
+ bn_pack4(0x4009, 0x438B, 0x481C, 0x6CD7),
+ bn_pack4(0x3590, 0x46F4, 0xEB87, 0x9F92),
+ bn_pack4(0xFAF3, 0x6BC3, 0x1ECF, 0xA268),
+ bn_pack4(0xB1D5, 0x10BD, 0x7EE7, 0x4D73),
+ bn_pack4(0xF9AB, 0x4819, 0x5DED, 0x7EA1),
+ bn_pack4(0x64F3, 0x1CC5, 0x0846, 0x851D),
+ bn_pack4(0x4597, 0xE899, 0xA025, 0x5DC1),
+ bn_pack4(0xDF31, 0x0EE0, 0x74AB, 0x6A36),
+ bn_pack4(0x6D2A, 0x13F8, 0x3F44, 0xF82D),
+ bn_pack4(0x062B, 0x3CF5, 0xB3A2, 0x78A6),
+ bn_pack4(0x7968, 0x3303, 0xED5B, 0xDD3A),
+ bn_pack4(0xFA9D, 0x4B7F, 0xA2C0, 0x87E8),
+ bn_pack4(0x4BCB, 0xC886, 0x2F83, 0x85DD),
+ bn_pack4(0x3473, 0xFC64, 0x6CEA, 0x306B),
+ bn_pack4(0x13EB, 0x57A8, 0x1A23, 0xF0C7),
+ bn_pack4(0x2222, 0x2E04, 0xA403, 0x7C07),
+ bn_pack4(0xE3FD, 0xB8BE, 0xFC84, 0x8AD9),
+ bn_pack4(0x238F, 0x16CB, 0xE39D, 0x652D),
+ bn_pack4(0x3423, 0xB474, 0x2BF1, 0xC978),
+ bn_pack4(0x3AAB, 0x639C, 0x5AE4, 0xF568),
+ bn_pack4(0x2576, 0xF693, 0x6BA4, 0x2466),
+ bn_pack4(0x741F, 0xA7BF, 0x8AFC, 0x47ED),
+ bn_pack4(0x3BC8, 0x32B6, 0x8D9D, 0xD300),
+ bn_pack4(0xD8BE, 0xC4D0, 0x73B9, 0x31BA),
+ bn_pack4(0x3877, 0x7CB6, 0xA932, 0xDF8C),
+ bn_pack4(0x74A3, 0x926F, 0x12FE, 0xE5E4),
+ bn_pack4(0xE694, 0xF91E, 0x6DBE, 0x1159),
+ bn_pack4(0x12BF, 0x2D5B, 0x0B74, 0x74D6),
+ bn_pack4(0x043E, 0x8F66, 0x3F48, 0x60EE),
+ bn_pack4(0x387F, 0xE8D7, 0x6E3C, 0x0468),
+ bn_pack4(0xDA56, 0xC9EC, 0x2EF2, 0x9632),
+ bn_pack4(0xEB19, 0xCCB1, 0xA313, 0xD55C),
+ bn_pack4(0xF550, 0xAA3D, 0x8A1F, 0xBFF0),
+ bn_pack4(0x06A1, 0xD58B, 0xB7C5, 0xDA76),
+ bn_pack4(0xA797, 0x15EE, 0xF29B, 0xE328),
+ bn_pack4(0x14CC, 0x5ED2, 0x0F80, 0x37E0),
+ bn_pack4(0xCC8F, 0x6D7E, 0xBF48, 0xE1D8),
+ bn_pack4(0x4BD4, 0x07B2, 0x2B41, 0x54AA),
+ bn_pack4(0x0F1D, 0x45B7, 0xFF58, 0x5AC5),
+ bn_pack4(0x23A9, 0x7A7E, 0x36CC, 0x88BE),
+ bn_pack4(0x59E7, 0xC97F, 0xBEC7, 0xE8F3),
+ bn_pack4(0xB5A8, 0x4031, 0x900B, 0x1C9E),
+ bn_pack4(0xD55E, 0x702F, 0x4698, 0x0C82),
+ bn_pack4(0xF482, 0xD7CE, 0x6E74, 0xFEF6),
+ bn_pack4(0xF032, 0xEA15, 0xD172, 0x1D03),
+ bn_pack4(0x5983, 0xCA01, 0xC64B, 0x92EC),
+ bn_pack4(0x6FB8, 0xF401, 0x378C, 0xD2BF),
+ bn_pack4(0x3320, 0x5151, 0x2BD7, 0xAF42),
+ bn_pack4(0xDB7F, 0x1447, 0xE6CC, 0x254B),
+ bn_pack4(0x44CE, 0x6CBA, 0xCED4, 0xBB1B),
+ bn_pack4(0xDA3E, 0xDBEB, 0xCF9B, 0x14ED),
+ bn_pack4(0x1797, 0x27B0, 0x865A, 0x8918),
+ bn_pack4(0xB06A, 0x53ED, 0x9027, 0xD831),
+ bn_pack4(0xE5DB, 0x382F, 0x4130, 0x01AE),
+ bn_pack4(0xF8FF, 0x9406, 0xAD9E, 0x530E),
+ bn_pack4(0xC975, 0x1E76, 0x3DBA, 0x37BD),
+ bn_pack4(0xC1D4, 0xDCB2, 0x6026, 0x46DE),
+ bn_pack4(0x36C3, 0xFAB4, 0xD27C, 0x7026),
+ bn_pack4(0x4DF4, 0x35C9, 0x3402, 0x8492),
+ bn_pack4(0x86FF, 0xB7DC, 0x90A6, 0xC08F),
+ bn_pack4(0x93B4, 0xEA98, 0x8D8F, 0xDDC1),
+ bn_pack4(0xD006, 0x9127, 0xD5B0, 0x5AA9),
+ bn_pack4(0xB81B, 0xDD76, 0x2170, 0x481C),
+ bn_pack4(0x1F61, 0x2970, 0xCEE2, 0xD7AF),
+ bn_pack4(0x233B, 0xA186, 0x515B, 0xE7ED),
+ bn_pack4(0x99B2, 0x964F, 0xA090, 0xC3A2),
+ bn_pack4(0x287C, 0x5947, 0x4E6B, 0xC05D),
+ bn_pack4(0x2E8E, 0xFC14, 0x1FBE, 0xCAA6),
+ bn_pack4(0xDBBB, 0xC2DB, 0x04DE, 0x8EF9),
+ bn_pack4(0x2583, 0xE9CA, 0x2AD4, 0x4CE8),
+ bn_pack4(0x1A94, 0x6834, 0xB615, 0x0BDA),
+ bn_pack4(0x99C3, 0x2718, 0x6AF4, 0xE23C),
+ bn_pack4(0x8871, 0x9A10, 0xBDBA, 0x5B26),
+ bn_pack4(0x1A72, 0x3C12, 0xA787, 0xE6D7),
+ bn_pack4(0x4B82, 0xD120, 0xA921, 0x0801),
+ bn_pack4(0x43DB, 0x5BFC, 0xE0FD, 0x108E),
+ bn_pack4(0x08E2, 0x4FA0, 0x74E5, 0xAB31),
+ bn_pack4(0x7709, 0x88C0, 0xBAD9, 0x46E2),
+ bn_pack4(0xBBE1, 0x1757, 0x7A61, 0x5D6C),
+ bn_pack4(0x521F, 0x2B18, 0x177B, 0x200C),
+ bn_pack4(0xD876, 0x0273, 0x3EC8, 0x6A64),
+ bn_pack4(0xF12F, 0xFA06, 0xD98A, 0x0864),
+ bn_pack4(0xCEE3, 0xD226, 0x1AD2, 0xEE6B),
+ bn_pack4(0x1E8C, 0x94E0, 0x4A25, 0x619D),
+ bn_pack4(0xABF5, 0xAE8C, 0xDB09, 0x33D7),
+ bn_pack4(0xB397, 0x0F85, 0xA6E1, 0xE4C7),
+ bn_pack4(0x8AEA, 0x7157, 0x5D06, 0x0C7D),
+ bn_pack4(0xECFB, 0x8504, 0x58DB, 0xEF0A),
+ bn_pack4(0xA855, 0x21AB, 0xDF1C, 0xBA64),
+ bn_pack4(0xAD33, 0x170D, 0x0450, 0x7A33),
+ bn_pack4(0x1572, 0x8E5A, 0x8AAA, 0xC42D),
+ bn_pack4(0x15D2, 0x2618, 0x98FA, 0x0510),
+ bn_pack4(0x3995, 0x497C, 0xEA95, 0x6AE5),
+ bn_pack4(0xDE2B, 0xCBF6, 0x9558, 0x1718),
+ bn_pack4(0xB5C5, 0x5DF0, 0x6F4C, 0x52C9),
+ bn_pack4(0x9B27, 0x83A2, 0xEC07, 0xA28F),
+ bn_pack4(0xE39E, 0x772C, 0x180E, 0x8603),
+ bn_pack4(0x3290, 0x5E46, 0x2E36, 0xCE3B),
+ bn_pack4(0xF174, 0x6C08, 0xCA18, 0x217C),
+ bn_pack4(0x670C, 0x354E, 0x4ABC, 0x9804),
+ bn_pack4(0x9ED5, 0x2907, 0x7096, 0x966D),
+ bn_pack4(0x1C62, 0xF356, 0x2085, 0x52BB),
+ bn_pack4(0x8365, 0x5D23, 0xDCA3, 0xAD96),
+ bn_pack4(0x6916, 0x3FA8, 0xFD24, 0xCF5F),
+ bn_pack4(0x98DA, 0x4836, 0x1C55, 0xD39A),
+ bn_pack4(0xC200, 0x7CB8, 0xA163, 0xBF05),
+ bn_pack4(0x4928, 0x6651, 0xECE4, 0x5B3D),
+ bn_pack4(0xAE9F, 0x2411, 0x7C4B, 0x1FE6),
+ bn_pack4(0xEE38, 0x6BFB, 0x5A89, 0x9FA5),
+ bn_pack4(0x0BFF, 0x5CB6, 0xF406, 0xB7ED),
+ bn_pack4(0xF44C, 0x42E9, 0xA637, 0xED6B),
+ bn_pack4(0xE485, 0xB576, 0x625E, 0x7EC6),
+ bn_pack4(0x4FE1, 0x356D, 0x6D51, 0xC245),
+ bn_pack4(0x302B, 0x0A6D, 0xF25F, 0x1437),
+ bn_pack4(0xEF95, 0x19B3, 0xCD3A, 0x431B),
+ bn_pack4(0x514A, 0x0879, 0x8E34, 0x04DD),
+ bn_pack4(0x020B, 0xBEA6, 0x3B13, 0x9B22),
+ bn_pack4(0x2902, 0x4E08, 0x8A67, 0xCC74),
+ bn_pack4(0xC4C6, 0x628B, 0x80DC, 0x1CD1),
+ bn_pack4(0xC90F, 0xDAA2, 0x2168, 0xC234),
+ bn_pack4(0xFFFF, 0xFFFF, 0xFFFF, 0xFFFF)
+};
+
+const BIGNUM bn_group_8192 = {
+ (BN_ULONG *)bn_group_8192_value,
+ OSSL_NELEM(bn_group_8192_value),
+ OSSL_NELEM(bn_group_8192_value),
+ 0,
+ BN_FLG_STATIC_DATA
+};
+
+static const BN_ULONG bn_generator_19_value[] = { 19 };
+
+const BIGNUM bn_generator_19 = {
+ (BN_ULONG *)bn_generator_19_value,
+ 1,
+ 1,
+ 0,
+ BN_FLG_STATIC_DATA
+};
+static const BN_ULONG bn_generator_5_value[] = { 5 };
+
+const BIGNUM bn_generator_5 = {
+ (BN_ULONG *)bn_generator_5_value,
+ 1,
+ 1,
+ 0,
+ BN_FLG_STATIC_DATA
+};
+static const BN_ULONG bn_generator_2_value[] = { 2 };
+
+const BIGNUM bn_generator_2 = {
+ (BN_ULONG *)bn_generator_2_value,
+ 1,
+ 1,
+ 0,
+ BN_FLG_STATIC_DATA
+};
+
+#endif
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_word.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_word.c
new file mode 100644
index 0000000..18fb303
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_word.c
@@ -0,0 +1,201 @@
+/*
+ * Copyright 1995-2016 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include "internal/cryptlib.h"
+#include "bn_local.h"
+
+BN_ULONG BN_mod_word(const BIGNUM *a, BN_ULONG w)
+{
+#ifndef BN_LLONG
+ BN_ULONG ret = 0;
+#else
+ BN_ULLONG ret = 0;
+#endif
+ int i;
+
+ if (w == 0)
+ return (BN_ULONG)-1;
+
+#ifndef BN_LLONG
+ /*
+ * If |w| is too long and we don't have BN_ULLONG then we need to fall
+ * back to using BN_div_word
+ */
+ if (w > ((BN_ULONG)1 << BN_BITS4)) {
+ BIGNUM *tmp = BN_dup(a);
+ if (tmp == NULL)
+ return (BN_ULONG)-1;
+
+ ret = BN_div_word(tmp, w);
+ BN_free(tmp);
+
+ return ret;
+ }
+#endif
+
+ bn_check_top(a);
+ w &= BN_MASK2;
+ for (i = a->top - 1; i >= 0; i--) {
+#ifndef BN_LLONG
+ /*
+ * We can assume here that | w <= ((BN_ULONG)1 << BN_BITS4) | and so
+ * | ret < ((BN_ULONG)1 << BN_BITS4) | and therefore the shifts here are
+ * safe and will not overflow
+ */
+ ret = ((ret << BN_BITS4) | ((a->d[i] >> BN_BITS4) & BN_MASK2l)) % w;
+ ret = ((ret << BN_BITS4) | (a->d[i] & BN_MASK2l)) % w;
+#else
+ ret = (BN_ULLONG) (((ret << (BN_ULLONG) BN_BITS2) | a->d[i]) %
+ (BN_ULLONG) w);
+#endif
+ }
+ return (BN_ULONG)ret;
+}
+
+BN_ULONG BN_div_word(BIGNUM *a, BN_ULONG w)
+{
+ BN_ULONG ret = 0;
+ int i, j;
+
+ bn_check_top(a);
+ w &= BN_MASK2;
+
+ if (!w)
+ /* actually this an error (division by zero) */
+ return (BN_ULONG)-1;
+ if (a->top == 0)
+ return 0;
+
+ /* normalize input (so bn_div_words doesn't complain) */
+ j = BN_BITS2 - BN_num_bits_word(w);
+ w <<= j;
+ if (!BN_lshift(a, a, j))
+ return (BN_ULONG)-1;
+
+ for (i = a->top - 1; i >= 0; i--) {
+ BN_ULONG l, d;
+
+ l = a->d[i];
+ d = bn_div_words(ret, l, w);
+ ret = (l - ((d * w) & BN_MASK2)) & BN_MASK2;
+ a->d[i] = d;
+ }
+ if ((a->top > 0) && (a->d[a->top - 1] == 0))
+ a->top--;
+ ret >>= j;
+ if (!a->top)
+ a->neg = 0; /* don't allow negative zero */
+ bn_check_top(a);
+ return ret;
+}
+
+int BN_add_word(BIGNUM *a, BN_ULONG w)
+{
+ BN_ULONG l;
+ int i;
+
+ bn_check_top(a);
+ w &= BN_MASK2;
+
+ /* degenerate case: w is zero */
+ if (!w)
+ return 1;
+ /* degenerate case: a is zero */
+ if (BN_is_zero(a))
+ return BN_set_word(a, w);
+ /* handle 'a' when negative */
+ if (a->neg) {
+ a->neg = 0;
+ i = BN_sub_word(a, w);
+ if (!BN_is_zero(a))
+ a->neg = !(a->neg);
+ return i;
+ }
+ for (i = 0; w != 0 && i < a->top; i++) {
+ a->d[i] = l = (a->d[i] + w) & BN_MASK2;
+ w = (w > l) ? 1 : 0;
+ }
+ if (w && i == a->top) {
+ if (bn_wexpand(a, a->top + 1) == NULL)
+ return 0;
+ a->top++;
+ a->d[i] = w;
+ }
+ bn_check_top(a);
+ return 1;
+}
+
+int BN_sub_word(BIGNUM *a, BN_ULONG w)
+{
+ int i;
+
+ bn_check_top(a);
+ w &= BN_MASK2;
+
+ /* degenerate case: w is zero */
+ if (!w)
+ return 1;
+ /* degenerate case: a is zero */
+ if (BN_is_zero(a)) {
+ i = BN_set_word(a, w);
+ if (i != 0)
+ BN_set_negative(a, 1);
+ return i;
+ }
+ /* handle 'a' when negative */
+ if (a->neg) {
+ a->neg = 0;
+ i = BN_add_word(a, w);
+ a->neg = 1;
+ return i;
+ }
+
+ if ((a->top == 1) && (a->d[0] < w)) {
+ a->d[0] = w - a->d[0];
+ a->neg = 1;
+ return 1;
+ }
+ i = 0;
+ for (;;) {
+ if (a->d[i] >= w) {
+ a->d[i] -= w;
+ break;
+ } else {
+ a->d[i] = (a->d[i] - w) & BN_MASK2;
+ i++;
+ w = 1;
+ }
+ }
+ if ((a->d[i] == 0) && (i == (a->top - 1)))
+ a->top--;
+ bn_check_top(a);
+ return 1;
+}
+
+int BN_mul_word(BIGNUM *a, BN_ULONG w)
+{
+ BN_ULONG ll;
+
+ bn_check_top(a);
+ w &= BN_MASK2;
+ if (a->top) {
+ if (w == 0)
+ BN_zero(a);
+ else {
+ ll = bn_mul_words(a->d, a->d, a->top, w);
+ if (ll) {
+ if (bn_wexpand(a, a->top + 1) == NULL)
+ return 0;
+ a->d[a->top++] = ll;
+ }
+ }
+ }
+ bn_check_top(a);
+ return 1;
+}
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_x931p.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_x931p.c
new file mode 100644
index 0000000..0099502
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/bn_x931p.c
@@ -0,0 +1,244 @@
+/*
+ * Copyright 2011-2018 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include <stdio.h>
+#include <openssl/bn.h>
+#include "bn_local.h"
+
+/* X9.31 routines for prime derivation */
+
+/*
+ * X9.31 prime derivation. This is used to generate the primes pi (p1, p2,
+ * q1, q2) from a parameter Xpi by checking successive odd integers.
+ */
+
+static int bn_x931_derive_pi(BIGNUM *pi, const BIGNUM *Xpi, BN_CTX *ctx,
+ BN_GENCB *cb)
+{
+ int i = 0, is_prime;
+ if (!BN_copy(pi, Xpi))
+ return 0;
+ if (!BN_is_odd(pi) && !BN_add_word(pi, 1))
+ return 0;
+ for (;;) {
+ i++;
+ BN_GENCB_call(cb, 0, i);
+ /* NB 27 MR is specified in X9.31 */
+ is_prime = BN_is_prime_fasttest_ex(pi, 27, ctx, 1, cb);
+ if (is_prime < 0)
+ return 0;
+ if (is_prime)
+ break;
+ if (!BN_add_word(pi, 2))
+ return 0;
+ }
+ BN_GENCB_call(cb, 2, i);
+ return 1;
+}
+
+/*
+ * This is the main X9.31 prime derivation function. From parameters Xp1, Xp2
+ * and Xp derive the prime p. If the parameters p1 or p2 are not NULL they
+ * will be returned too: this is needed for testing.
+ */
+
+int BN_X931_derive_prime_ex(BIGNUM *p, BIGNUM *p1, BIGNUM *p2,
+ const BIGNUM *Xp, const BIGNUM *Xp1,
+ const BIGNUM *Xp2, const BIGNUM *e, BN_CTX *ctx,
+ BN_GENCB *cb)
+{
+ int ret = 0;
+
+ BIGNUM *t, *p1p2, *pm1;
+
+ /* Only even e supported */
+ if (!BN_is_odd(e))
+ return 0;
+
+ BN_CTX_start(ctx);
+ if (p1 == NULL)
+ p1 = BN_CTX_get(ctx);
+
+ if (p2 == NULL)
+ p2 = BN_CTX_get(ctx);
+
+ t = BN_CTX_get(ctx);
+
+ p1p2 = BN_CTX_get(ctx);
+
+ pm1 = BN_CTX_get(ctx);
+
+ if (pm1 == NULL)
+ goto err;
+
+ if (!bn_x931_derive_pi(p1, Xp1, ctx, cb))
+ goto err;
+
+ if (!bn_x931_derive_pi(p2, Xp2, ctx, cb))
+ goto err;
+
+ if (!BN_mul(p1p2, p1, p2, ctx))
+ goto err;
+
+ /* First set p to value of Rp */
+
+ if (!BN_mod_inverse(p, p2, p1, ctx))
+ goto err;
+
+ if (!BN_mul(p, p, p2, ctx))
+ goto err;
+
+ if (!BN_mod_inverse(t, p1, p2, ctx))
+ goto err;
+
+ if (!BN_mul(t, t, p1, ctx))
+ goto err;
+
+ if (!BN_sub(p, p, t))
+ goto err;
+
+ if (p->neg && !BN_add(p, p, p1p2))
+ goto err;
+
+ /* p now equals Rp */
+
+ if (!BN_mod_sub(p, p, Xp, p1p2, ctx))
+ goto err;
+
+ if (!BN_add(p, p, Xp))
+ goto err;
+
+ /* p now equals Yp0 */
+
+ for (;;) {
+ int i = 1;
+ BN_GENCB_call(cb, 0, i++);
+ if (!BN_copy(pm1, p))
+ goto err;
+ if (!BN_sub_word(pm1, 1))
+ goto err;
+ if (!BN_gcd(t, pm1, e, ctx))
+ goto err;
+ if (BN_is_one(t)) {
+ /*
+ * X9.31 specifies 8 MR and 1 Lucas test or any prime test
+ * offering similar or better guarantees 50 MR is considerably
+ * better.
+ */
+ int r = BN_is_prime_fasttest_ex(p, 50, ctx, 1, cb);
+ if (r < 0)
+ goto err;
+ if (r)
+ break;
+ }
+ if (!BN_add(p, p, p1p2))
+ goto err;
+ }
+
+ BN_GENCB_call(cb, 3, 0);
+
+ ret = 1;
+
+ err:
+
+ BN_CTX_end(ctx);
+
+ return ret;
+}
+
+/*
+ * Generate pair of parameters Xp, Xq for X9.31 prime generation. Note: nbits
+ * parameter is sum of number of bits in both.
+ */
+
+int BN_X931_generate_Xpq(BIGNUM *Xp, BIGNUM *Xq, int nbits, BN_CTX *ctx)
+{
+ BIGNUM *t;
+ int i;
+ /*
+ * Number of bits for each prime is of the form 512+128s for s = 0, 1,
+ * ...
+ */
+ if ((nbits < 1024) || (nbits & 0xff))
+ return 0;
+ nbits >>= 1;
+ /*
+ * The random value Xp must be between sqrt(2) * 2^(nbits-1) and 2^nbits
+ * - 1. By setting the top two bits we ensure that the lower bound is
+ * exceeded.
+ */
+ if (!BN_priv_rand(Xp, nbits, BN_RAND_TOP_TWO, BN_RAND_BOTTOM_ANY))
+ goto err;
+
+ BN_CTX_start(ctx);
+ t = BN_CTX_get(ctx);
+ if (t == NULL)
+ goto err;
+
+ for (i = 0; i < 1000; i++) {
+ if (!BN_priv_rand(Xq, nbits, BN_RAND_TOP_TWO, BN_RAND_BOTTOM_ANY))
+ goto err;
+
+ /* Check that |Xp - Xq| > 2^(nbits - 100) */
+ if (!BN_sub(t, Xp, Xq))
+ goto err;
+ if (BN_num_bits(t) > (nbits - 100))
+ break;
+ }
+
+ BN_CTX_end(ctx);
+
+ if (i < 1000)
+ return 1;
+
+ return 0;
+
+ err:
+ BN_CTX_end(ctx);
+ return 0;
+}
+
+/*
+ * Generate primes using X9.31 algorithm. Of the values p, p1, p2, Xp1 and
+ * Xp2 only 'p' needs to be non-NULL. If any of the others are not NULL the
+ * relevant parameter will be stored in it. Due to the fact that |Xp - Xq| >
+ * 2^(nbits - 100) must be satisfied Xp and Xq are generated using the
+ * previous function and supplied as input.
+ */
+
+int BN_X931_generate_prime_ex(BIGNUM *p, BIGNUM *p1, BIGNUM *p2,
+ BIGNUM *Xp1, BIGNUM *Xp2,
+ const BIGNUM *Xp,
+ const BIGNUM *e, BN_CTX *ctx, BN_GENCB *cb)
+{
+ int ret = 0;
+
+ BN_CTX_start(ctx);
+ if (Xp1 == NULL)
+ Xp1 = BN_CTX_get(ctx);
+ if (Xp2 == NULL)
+ Xp2 = BN_CTX_get(ctx);
+ if (Xp1 == NULL || Xp2 == NULL)
+ goto error;
+
+ if (!BN_priv_rand(Xp1, 101, BN_RAND_TOP_ONE, BN_RAND_BOTTOM_ANY))
+ goto error;
+ if (!BN_priv_rand(Xp2, 101, BN_RAND_TOP_ONE, BN_RAND_BOTTOM_ANY))
+ goto error;
+ if (!BN_X931_derive_prime_ex(p, p1, p2, Xp, Xp1, Xp2, e, ctx, cb))
+ goto error;
+
+ ret = 1;
+
+ error:
+ BN_CTX_end(ctx);
+
+ return ret;
+
+}
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/build.info b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/build.info
new file mode 100644
index 0000000..b9ed532
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/build.info
@@ -0,0 +1,66 @@
+LIBS=../../libcrypto
+SOURCE[../../libcrypto]=\
+ bn_add.c bn_div.c bn_exp.c bn_lib.c bn_ctx.c bn_mul.c bn_mod.c \
+ bn_print.c bn_rand.c bn_shift.c bn_word.c bn_blind.c \
+ bn_kron.c bn_sqrt.c bn_gcd.c bn_prime.c bn_err.c bn_sqr.c \
+ {- $target{bn_asm_src} -} \
+ bn_recp.c bn_mont.c bn_mpi.c bn_exp2.c bn_gf2m.c bn_nist.c \
+ bn_depr.c bn_const.c bn_x931p.c bn_intern.c bn_dh.c bn_srp.c
+
+INCLUDE[bn_exp.o]=..
+
+GENERATE[bn-586.s]=asm/bn-586.pl \
+ $(PERLASM_SCHEME) $(LIB_CFLAGS) $(LIB_CPPFLAGS) $(PROCESSOR)
+DEPEND[bn-586.s]=../perlasm/x86asm.pl
+GENERATE[co-586.s]=asm/co-586.pl \
+ $(PERLASM_SCHEME) $(LIB_CFLAGS) $(LIB_CPPFLAGS) $(PROCESSOR)
+DEPEND[co-586.s]=../perlasm/x86asm.pl
+GENERATE[x86-mont.s]=asm/x86-mont.pl \
+ $(PERLASM_SCHEME) $(LIB_CFLAGS) $(LIB_CPPFLAGS) $(PROCESSOR)
+DEPEND[x86-mont.s]=../perlasm/x86asm.pl
+GENERATE[x86-gf2m.s]=asm/x86-gf2m.pl \
+ $(PERLASM_SCHEME) $(LIB_CFLAGS) $(LIB_CPPFLAGS) $(PROCESSOR)
+DEPEND[x86-gf2m.s]=../perlasm/x86asm.pl
+
+GENERATE[sparcv9a-mont.S]=asm/sparcv9a-mont.pl $(PERLASM_SCHEME)
+INCLUDE[sparcv9a-mont.o]=..
+GENERATE[sparcv9-mont.S]=asm/sparcv9-mont.pl $(PERLASM_SCHEME)
+INCLUDE[sparcv9-mont.o]=..
+GENERATE[vis3-mont.S]=asm/vis3-mont.pl $(PERLASM_SCHEME)
+INCLUDE[vis3-mont.o]=..
+GENERATE[sparct4-mont.S]=asm/sparct4-mont.pl $(PERLASM_SCHEME)
+INCLUDE[sparct4-mont.o]=..
+GENERATE[sparcv9-gf2m.S]=asm/sparcv9-gf2m.pl $(PERLASM_SCHEME)
+INCLUDE[sparcv9-gf2m.o]=..
+
+GENERATE[bn-mips.S]=asm/mips.pl $(PERLASM_SCHEME)
+INCLUDE[bn-mips.o]=..
+GENERATE[mips-mont.S]=asm/mips-mont.pl $(PERLASM_SCHEME)
+INCLUDE[mips-mont.o]=..
+
+GENERATE[s390x-mont.S]=asm/s390x-mont.pl $(PERLASM_SCHEME)
+GENERATE[s390x-gf2m.s]=asm/s390x-gf2m.pl $(PERLASM_SCHEME)
+
+GENERATE[x86_64-mont.s]=asm/x86_64-mont.pl $(PERLASM_SCHEME)
+GENERATE[x86_64-mont5.s]=asm/x86_64-mont5.pl $(PERLASM_SCHEME)
+GENERATE[x86_64-gf2m.s]=asm/x86_64-gf2m.pl $(PERLASM_SCHEME)
+GENERATE[rsaz-x86_64.s]=asm/rsaz-x86_64.pl $(PERLASM_SCHEME)
+GENERATE[rsaz-avx2.s]=asm/rsaz-avx2.pl $(PERLASM_SCHEME)
+
+GENERATE[bn-ia64.s]=asm/ia64.S
+GENERATE[ia64-mont.s]=asm/ia64-mont.pl $(LIB_CFLAGS) $(LIB_CPPFLAGS)
+
+GENERATE[parisc-mont.s]=asm/parisc-mont.pl $(PERLASM_SCHEME)
+
+# ppc - AIX, Linux, MacOS X...
+GENERATE[bn-ppc.s]=asm/ppc.pl $(PERLASM_SCHEME)
+GENERATE[ppc-mont.s]=asm/ppc-mont.pl $(PERLASM_SCHEME)
+GENERATE[ppc64-mont.s]=asm/ppc64-mont.pl $(PERLASM_SCHEME)
+
+GENERATE[alpha-mont.S]=asm/alpha-mont.pl $(PERLASM_SCHEME)
+
+GENERATE[armv4-mont.S]=asm/armv4-mont.pl $(PERLASM_SCHEME)
+INCLUDE[armv4-mont.o]=..
+GENERATE[armv4-gf2m.S]=asm/armv4-gf2m.pl $(PERLASM_SCHEME)
+INCLUDE[armv4-gf2m.o]=..
+GENERATE[armv8-mont.S]=asm/armv8-mont.pl $(PERLASM_SCHEME)
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/rsaz_exp.c b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/rsaz_exp.c
new file mode 100644
index 0000000..a2ab58b
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/rsaz_exp.c
@@ -0,0 +1,323 @@
+/*
+ * Copyright 2013-2022 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright (c) 2012, Intel Corporation. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ *
+ * Originally written by Shay Gueron (1, 2), and Vlad Krasnov (1)
+ * (1) Intel Corporation, Israel Development Center, Haifa, Israel
+ * (2) University of Haifa, Israel
+ */
+
+#include <openssl/opensslconf.h>
+#include "rsaz_exp.h"
+
+#ifndef RSAZ_ENABLED
+NON_EMPTY_TRANSLATION_UNIT
+#else
+
+/*
+ * See crypto/bn/asm/rsaz-avx2.pl for further details.
+ */
+void rsaz_1024_norm2red_avx2(void *red, const void *norm);
+void rsaz_1024_mul_avx2(void *ret, const void *a, const void *b,
+ const void *n, BN_ULONG k);
+void rsaz_1024_sqr_avx2(void *ret, const void *a, const void *n, BN_ULONG k,
+ int cnt);
+void rsaz_1024_scatter5_avx2(void *tbl, const void *val, int i);
+void rsaz_1024_gather5_avx2(void *val, const void *tbl, int i);
+void rsaz_1024_red2norm_avx2(void *norm, const void *red);
+
+#if defined(__GNUC__)
+# define ALIGN64 __attribute__((aligned(64)))
+#elif defined(_MSC_VER)
+# define ALIGN64 __declspec(align(64))
+#elif defined(__SUNPRO_C)
+# define ALIGN64
+# pragma align 64(one,two80)
+#else
+/* not fatal, might hurt performance a little */
+# define ALIGN64
+#endif
+
+ALIGN64 static const BN_ULONG one[40] = {
+ 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
+};
+
+ALIGN64 static const BN_ULONG two80[40] = {
+ 0, 0, 1 << 22, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
+};
+
+void RSAZ_1024_mod_exp_avx2(BN_ULONG result_norm[16],
+ const BN_ULONG base_norm[16],
+ const BN_ULONG exponent[16],
+ const BN_ULONG m_norm[16], const BN_ULONG RR[16],
+ BN_ULONG k0)
+{
+ unsigned char storage[320 * 3 + 32 * 9 * 16 + 64]; /* 5.5KB */
+ unsigned char *p_str = storage + (64 - ((size_t)storage % 64));
+ unsigned char *a_inv, *m, *result;
+ unsigned char *table_s = p_str + 320 * 3;
+ unsigned char *R2 = table_s; /* borrow */
+ int index;
+ int wvalue;
+ BN_ULONG tmp[16];
+
+ if ((((size_t)p_str & 4095) + 320) >> 12) {
+ result = p_str;
+ a_inv = p_str + 320;
+ m = p_str + 320 * 2; /* should not cross page */
+ } else {
+ m = p_str; /* should not cross page */
+ result = p_str + 320;
+ a_inv = p_str + 320 * 2;
+ }
+
+ rsaz_1024_norm2red_avx2(m, m_norm);
+ rsaz_1024_norm2red_avx2(a_inv, base_norm);
+ rsaz_1024_norm2red_avx2(R2, RR);
+
+ rsaz_1024_mul_avx2(R2, R2, R2, m, k0);
+ rsaz_1024_mul_avx2(R2, R2, two80, m, k0);
+
+ /* table[0] = 1 */
+ rsaz_1024_mul_avx2(result, R2, one, m, k0);
+ /* table[1] = a_inv^1 */
+ rsaz_1024_mul_avx2(a_inv, a_inv, R2, m, k0);
+
+ rsaz_1024_scatter5_avx2(table_s, result, 0);
+ rsaz_1024_scatter5_avx2(table_s, a_inv, 1);
+
+ /* table[2] = a_inv^2 */
+ rsaz_1024_sqr_avx2(result, a_inv, m, k0, 1);
+ rsaz_1024_scatter5_avx2(table_s, result, 2);
+#if 0
+ /* this is almost 2x smaller and less than 1% slower */
+ for (index = 3; index < 32; index++) {
+ rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
+ rsaz_1024_scatter5_avx2(table_s, result, index);
+ }
+#else
+ /* table[4] = a_inv^4 */
+ rsaz_1024_sqr_avx2(result, result, m, k0, 1);
+ rsaz_1024_scatter5_avx2(table_s, result, 4);
+ /* table[8] = a_inv^8 */
+ rsaz_1024_sqr_avx2(result, result, m, k0, 1);
+ rsaz_1024_scatter5_avx2(table_s, result, 8);
+ /* table[16] = a_inv^16 */
+ rsaz_1024_sqr_avx2(result, result, m, k0, 1);
+ rsaz_1024_scatter5_avx2(table_s, result, 16);
+ /* table[17] = a_inv^17 */
+ rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
+ rsaz_1024_scatter5_avx2(table_s, result, 17);
+
+ /* table[3] */
+ rsaz_1024_gather5_avx2(result, table_s, 2);
+ rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
+ rsaz_1024_scatter5_avx2(table_s, result, 3);
+ /* table[6] */
+ rsaz_1024_sqr_avx2(result, result, m, k0, 1);
+ rsaz_1024_scatter5_avx2(table_s, result, 6);
+ /* table[12] */
+ rsaz_1024_sqr_avx2(result, result, m, k0, 1);
+ rsaz_1024_scatter5_avx2(table_s, result, 12);
+ /* table[24] */
+ rsaz_1024_sqr_avx2(result, result, m, k0, 1);
+ rsaz_1024_scatter5_avx2(table_s, result, 24);
+ /* table[25] */
+ rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
+ rsaz_1024_scatter5_avx2(table_s, result, 25);
+
+ /* table[5] */
+ rsaz_1024_gather5_avx2(result, table_s, 4);
+ rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
+ rsaz_1024_scatter5_avx2(table_s, result, 5);
+ /* table[10] */
+ rsaz_1024_sqr_avx2(result, result, m, k0, 1);
+ rsaz_1024_scatter5_avx2(table_s, result, 10);
+ /* table[20] */
+ rsaz_1024_sqr_avx2(result, result, m, k0, 1);
+ rsaz_1024_scatter5_avx2(table_s, result, 20);
+ /* table[21] */
+ rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
+ rsaz_1024_scatter5_avx2(table_s, result, 21);
+
+ /* table[7] */
+ rsaz_1024_gather5_avx2(result, table_s, 6);
+ rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
+ rsaz_1024_scatter5_avx2(table_s, result, 7);
+ /* table[14] */
+ rsaz_1024_sqr_avx2(result, result, m, k0, 1);
+ rsaz_1024_scatter5_avx2(table_s, result, 14);
+ /* table[28] */
+ rsaz_1024_sqr_avx2(result, result, m, k0, 1);
+ rsaz_1024_scatter5_avx2(table_s, result, 28);
+ /* table[29] */
+ rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
+ rsaz_1024_scatter5_avx2(table_s, result, 29);
+
+ /* table[9] */
+ rsaz_1024_gather5_avx2(result, table_s, 8);
+ rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
+ rsaz_1024_scatter5_avx2(table_s, result, 9);
+ /* table[18] */
+ rsaz_1024_sqr_avx2(result, result, m, k0, 1);
+ rsaz_1024_scatter5_avx2(table_s, result, 18);
+ /* table[19] */
+ rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
+ rsaz_1024_scatter5_avx2(table_s, result, 19);
+
+ /* table[11] */
+ rsaz_1024_gather5_avx2(result, table_s, 10);
+ rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
+ rsaz_1024_scatter5_avx2(table_s, result, 11);
+ /* table[22] */
+ rsaz_1024_sqr_avx2(result, result, m, k0, 1);
+ rsaz_1024_scatter5_avx2(table_s, result, 22);
+ /* table[23] */
+ rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
+ rsaz_1024_scatter5_avx2(table_s, result, 23);
+
+ /* table[13] */
+ rsaz_1024_gather5_avx2(result, table_s, 12);
+ rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
+ rsaz_1024_scatter5_avx2(table_s, result, 13);
+ /* table[26] */
+ rsaz_1024_sqr_avx2(result, result, m, k0, 1);
+ rsaz_1024_scatter5_avx2(table_s, result, 26);
+ /* table[27] */
+ rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
+ rsaz_1024_scatter5_avx2(table_s, result, 27);
+
+ /* table[15] */
+ rsaz_1024_gather5_avx2(result, table_s, 14);
+ rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
+ rsaz_1024_scatter5_avx2(table_s, result, 15);
+ /* table[30] */
+ rsaz_1024_sqr_avx2(result, result, m, k0, 1);
+ rsaz_1024_scatter5_avx2(table_s, result, 30);
+ /* table[31] */
+ rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
+ rsaz_1024_scatter5_avx2(table_s, result, 31);
+#endif
+
+ /* load first window */
+ p_str = (unsigned char *)exponent;
+ wvalue = p_str[127] >> 3;
+ rsaz_1024_gather5_avx2(result, table_s, wvalue);
+
+ index = 1014;
+
+ while (index > -1) { /* loop for the remaining 127 windows */
+
+ rsaz_1024_sqr_avx2(result, result, m, k0, 5);
+
+ wvalue = (p_str[(index / 8) + 1] << 8) | p_str[index / 8];
+ wvalue = (wvalue >> (index % 8)) & 31;
+ index -= 5;
+
+ rsaz_1024_gather5_avx2(a_inv, table_s, wvalue); /* borrow a_inv */
+ rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
+ }
+
+ /* square four times */
+ rsaz_1024_sqr_avx2(result, result, m, k0, 4);
+
+ wvalue = p_str[0] & 15;
+
+ rsaz_1024_gather5_avx2(a_inv, table_s, wvalue); /* borrow a_inv */
+ rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
+
+ /* from Montgomery */
+ rsaz_1024_mul_avx2(result, result, one, m, k0);
+
+ rsaz_1024_red2norm_avx2(result_norm, result);
+
+ bn_reduce_once_in_place(result_norm, /*carry=*/0, m_norm, tmp, 16);
+
+ OPENSSL_cleanse(storage, sizeof(storage));
+ OPENSSL_cleanse(tmp, sizeof(tmp));
+}
+
+/*
+ * See crypto/bn/rsaz-x86_64.pl for further details.
+ */
+void rsaz_512_mul(void *ret, const void *a, const void *b, const void *n,
+ BN_ULONG k);
+void rsaz_512_mul_scatter4(void *ret, const void *a, const void *n,
+ BN_ULONG k, const void *tbl, unsigned int power);
+void rsaz_512_mul_gather4(void *ret, const void *a, const void *tbl,
+ const void *n, BN_ULONG k, unsigned int power);
+void rsaz_512_mul_by_one(void *ret, const void *a, const void *n, BN_ULONG k);
+void rsaz_512_sqr(void *ret, const void *a, const void *n, BN_ULONG k,
+ int cnt);
+void rsaz_512_scatter4(void *tbl, const BN_ULONG *val, int power);
+void rsaz_512_gather4(BN_ULONG *val, const void *tbl, int power);
+
+void RSAZ_512_mod_exp(BN_ULONG result[8],
+ const BN_ULONG base[8], const BN_ULONG exponent[8],
+ const BN_ULONG m[8], BN_ULONG k0, const BN_ULONG RR[8])
+{
+ unsigned char storage[16 * 8 * 8 + 64 * 2 + 64]; /* 1.2KB */
+ unsigned char *table = storage + (64 - ((size_t)storage % 64));
+ BN_ULONG *a_inv = (BN_ULONG *)(table + 16 * 8 * 8);
+ BN_ULONG *temp = (BN_ULONG *)(table + 16 * 8 * 8 + 8 * 8);
+ unsigned char *p_str = (unsigned char *)exponent;
+ int index;
+ unsigned int wvalue;
+ BN_ULONG tmp[8];
+
+ /* table[0] = 1_inv */
+ temp[0] = 0 - m[0];
+ temp[1] = ~m[1];
+ temp[2] = ~m[2];
+ temp[3] = ~m[3];
+ temp[4] = ~m[4];
+ temp[5] = ~m[5];
+ temp[6] = ~m[6];
+ temp[7] = ~m[7];
+ rsaz_512_scatter4(table, temp, 0);
+
+ /* table [1] = a_inv^1 */
+ rsaz_512_mul(a_inv, base, RR, m, k0);
+ rsaz_512_scatter4(table, a_inv, 1);
+
+ /* table [2] = a_inv^2 */
+ rsaz_512_sqr(temp, a_inv, m, k0, 1);
+ rsaz_512_scatter4(table, temp, 2);
+
+ for (index = 3; index < 16; index++)
+ rsaz_512_mul_scatter4(temp, a_inv, m, k0, table, index);
+
+ /* load first window */
+ wvalue = p_str[63];
+
+ rsaz_512_gather4(temp, table, wvalue >> 4);
+ rsaz_512_sqr(temp, temp, m, k0, 4);
+ rsaz_512_mul_gather4(temp, temp, table, m, k0, wvalue & 0xf);
+
+ for (index = 62; index >= 0; index--) {
+ wvalue = p_str[index];
+
+ rsaz_512_sqr(temp, temp, m, k0, 4);
+ rsaz_512_mul_gather4(temp, temp, table, m, k0, wvalue >> 4);
+
+ rsaz_512_sqr(temp, temp, m, k0, 4);
+ rsaz_512_mul_gather4(temp, temp, table, m, k0, wvalue & 0x0f);
+ }
+
+ /* from Montgomery */
+ rsaz_512_mul_by_one(result, temp, m, k0);
+
+ bn_reduce_once_in_place(result, /*carry=*/0, m, tmp, 8);
+
+ OPENSSL_cleanse(storage, sizeof(storage));
+ OPENSSL_cleanse(tmp, sizeof(tmp));
+}
+
+#endif
diff --git a/ap/lib/libssl/openssl-1.1.1o/crypto/bn/rsaz_exp.h b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/rsaz_exp.h
new file mode 100644
index 0000000..1532a7e
--- /dev/null
+++ b/ap/lib/libssl/openssl-1.1.1o/crypto/bn/rsaz_exp.h
@@ -0,0 +1,63 @@
+/*
+ * Copyright 2013-2022 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright (c) 2012, Intel Corporation. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License"). You may not use
+ * this file except in compliance with the License. You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ *
+ * Originally written by Shay Gueron (1, 2), and Vlad Krasnov (1)
+ * (1) Intel Corporation, Israel Development Center, Haifa, Israel
+ * (2) University of Haifa, Israel
+ */
+
+#ifndef OSSL_CRYPTO_BN_RSAZ_EXP_H
+# define OSSL_CRYPTO_BN_RSAZ_EXP_H
+
+# undef RSAZ_ENABLED
+# if defined(OPENSSL_BN_ASM_MONT) && \
+ (defined(__x86_64) || defined(__x86_64__) || \
+ defined(_M_AMD64) || defined(_M_X64))
+# define RSAZ_ENABLED
+
+# include <openssl/bn.h>
+# include "internal/constant_time.h"
+# include "bn_local.h"
+
+void RSAZ_1024_mod_exp_avx2(BN_ULONG result[16],
+ const BN_ULONG base_norm[16],
+ const BN_ULONG exponent[16],
+ const BN_ULONG m_norm[16], const BN_ULONG RR[16],
+ BN_ULONG k0);
+int rsaz_avx2_eligible(void);
+
+void RSAZ_512_mod_exp(BN_ULONG result[8],
+ const BN_ULONG base_norm[8], const BN_ULONG exponent[8],
+ const BN_ULONG m_norm[8], BN_ULONG k0,
+ const BN_ULONG RR[8]);
+
+static ossl_inline void bn_select_words(BN_ULONG *r, BN_ULONG mask,
+ const BN_ULONG *a,
+ const BN_ULONG *b, size_t num)
+{
+ size_t i;
+
+ for (i = 0; i < num; i++) {
+ r[i] = constant_time_select_64(mask, a[i], b[i]);
+ }
+}
+
+static ossl_inline BN_ULONG bn_reduce_once_in_place(BN_ULONG *r,
+ BN_ULONG carry,
+ const BN_ULONG *m,
+ BN_ULONG *tmp, size_t num)
+{
+ carry -= bn_sub_words(tmp, r, m, num);
+ bn_select_words(r, carry, r /* tmp < 0 */, tmp /* tmp >= 0 */, num);
+ return carry;
+}
+
+# endif
+
+#endif