This commit was manufactured by cvs2svn to create branch 'BRANCH_VMS_64BIT'.

18 files changed, 4256 insertions, 0 deletions

diff --git a/VMS/VMSify-conf.pl b/VMS/VMSify-conf.pl
new file mode 100644
index 0000000000..d3be6a29e7
--- /dev/null
+++ b/VMS/VMSify-conf.pl
@@ -0,0 +1,34 @@
+#! /usr/bin/perl
+
+use strict;
+use warnings;
+
+my @directory_vars = ( "dir", "certs", "crl_dir", "new_certs_dir" );
+my @file_vars = ( "database", "certificate", "serial", "crlnumber",
+		  "crl", "private_key", "RANDFILE" );
+while(<STDIN>) {
+    chomp;
+    foreach my $d (@directory_vars) {
+	if (/^(\s*\#?\s*${d}\s*=\s*)\.\/([^\s\#]*)([\s\#].*)$/) {
+	    $_ = "$1sys\\\$disk:\[.$2$3";
+	} elsif (/^(\s*\#?\s*${d}\s*=\s*)(\w[^\s\#]*)([\s\#].*)$/) {
+	    $_ = "$1sys\\\$disk:\[.$2$3";
+	}
+	s/^(\s*\#?\s*${d}\s*=\s*\$\w+)\/([^\s\#]*)([\s\#].*)$/$1.$2\]$3/;
+	while(/^(\s*\#?\s*${d}\s*=\s*(\$\w+\.|sys\\\$disk:\[\.)[\w\.]+)\/([^\]]*)\](.*)$/) {
+	    $_ = "$1.$3]$4";
+	}
+    }
+    foreach my $f (@file_vars) {
+	s/^(\s*\#?\s*${f}\s*=\s*)\.\/(.*)$/$1sys\\\$disk:\[\/$2/;
+	while(/^(\s*\#?\s*${f}\s*=\s*(\$\w+|sys\\\$disk:\[)[^\/]*)\/(\w+\/[^\s\#]*)([\s\#].*)$/) {
+	    $_ = "$1.$3$4";
+	}
+	if (/^(\s*\#?\s*${f}\s*=\s*(\$\w+|sys\\\$disk:\[)[^\/]*)\/(\w+)([\s\#].*)$/) {
+	    $_ = "$1]$3.$4";
+	} elsif  (/^(\s*\#?\s*${f}\s*=\s*(\$\w+|sys\\\$disk:\[)[^\/]*)\/([^\s\#]*)([\s\#].*)$/) {
+	    $_ = "$1]$3$4";
+	}
+   }
+    print $_,"\n";
+}
diff --git a/certs/eng1.pem b/certs/eng1.pem
new file mode 100644
index 0000000000..7ed8b1b5e6
--- /dev/null
+++ b/certs/eng1.pem
@@ -0,0 +1,23 @@
+-----BEGIN CERTIFICATE-----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+-----END CERTIFICATE-----
diff --git a/certs/eng2.pem b/certs/eng2.pem
new file mode 100644
index 0000000000..73066309b1
--- /dev/null
+++ b/certs/eng2.pem
@@ -0,0 +1,23 @@
+-----BEGIN CERTIFICATE-----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+-----END CERTIFICATE-----
diff --git a/certs/eng3.pem b/certs/eng3.pem
new file mode 100644
index 0000000000..28bcce2dfd
--- /dev/null
+++ b/certs/eng3.pem
@@ -0,0 +1,34 @@
+-----BEGIN CERTIFICATE-----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+-----END CERTIFICATE-----
diff --git a/certs/eng4.pem b/certs/eng4.pem
new file mode 100644
index 0000000000..9a7b156226
--- /dev/null
+++ b/certs/eng4.pem
@@ -0,0 +1,23 @@
+-----BEGIN CERTIFICATE-----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+-----END CERTIFICATE-----
diff --git a/certs/eng5.pem b/certs/eng5.pem
new file mode 100644
index 0000000000..3416ccad24
--- /dev/null
+++ b/certs/eng5.pem
@@ -0,0 +1,23 @@
+-----BEGIN CERTIFICATE-----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-----END CERTIFICATE-----
diff --git a/crypto/bn/asm/ppc.pl b/crypto/bn/asm/ppc.pl
new file mode 100644
index 0000000000..307c7ccb35
--- /dev/null
+++ b/crypto/bn/asm/ppc.pl
@@ -0,0 +1,2081 @@
+#!/usr/bin/env perl
+#
+# 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 bechmark 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%
+#
+#	If you have comments or suggestions to improve code send
+#	me a note at schari@us.ibm.com
+#
+
+$opf = shift;
+
+if ($opf =~ /32\.s/) {
+	$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
+	$COUNTZ="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
+	$ROTL=	"rotlwi";	# rotate left by immediate
+} elsif ($opf =~ /64\.s/) {
+	$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
+	$COUNTZ="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
+} else { die "nonsense $opf"; }
+
+( defined shift || open STDOUT,">$opf" ) || die "can't open $opf: $!";
+
+# function entry points from the AIX code
+#
+# There are other, more elegant, ways to handle this. We (IBM) chose
+# this approach as it plays well with scripts we run to 'namespace'
+# OpenSSL .i.e. we add a prefix to all the public symbols so we can
+# co-exist in the same process with other implementations of OpenSSL.
+# 'cleverer' ways of doing these substitutions tend to hide data we
+# need to be obvious.
+#
+my @items = ("bn_sqr_comba4",
+	     "bn_sqr_comba8",
+	     "bn_mul_comba4",
+	     "bn_mul_comba8",
+	     "bn_sub_words",
+	     "bn_add_words",
+	     "bn_div_words",
+	     "bn_sqr_words",
+	     "bn_mul_words",
+	     "bn_mul_add_words");
+
+if    ($opf =~ /linux/)	{  do_linux();	}
+elsif ($opf =~ /aix/)	{  do_aix();	}
+elsif ($opf =~ /osx/)	{  do_osx();	}
+else			{  do_bsd();	}
+
+sub do_linux {
+    $d=&data();
+
+    if ($BITS==64) {
+      foreach $t (@items) {
+        $d =~ s/\.$t:/\
+\t.section\t".opd","aw"\
+\t.align\t3\
+\t.globl\t$t\
+$t:\
+\t.quad\t.$t,.TOC.\@tocbase,0\
+\t.size\t$t,24\
+\t.previous\n\
+\t.type\t.$t,\@function\
+\t.globl\t.$t\
+.$t:/g;
+      }
+    }
+    else {
+      foreach $t (@items) {
+        $d=~s/\.$t/$t/g;
+      }
+    }
+    # hide internal labels to avoid pollution of name table...
+    $d=~s/Lppcasm_/.Lppcasm_/gm;
+    print $d;
+}
+
+sub do_aix {
+    # AIX assembler is smart enough to please the linker without
+    # making us do something special...
+    print &data();
+}
+
+# MacOSX 32 bit
+sub do_osx {
+    $d=&data();
+    # Change the bn symbol prefix from '.' to '_'
+    foreach $t (@items) {
+      $d=~s/\.$t/_$t/g;
+    }
+    # Change .machine to something OS X asm will accept
+    $d=~s/\.machine.*/.text/g;
+    $d=~s/\#/;/g; # change comment from '#' to ';'
+    print $d;
+}
+
+# BSD (Untested)
+sub do_bsd {
+    $d=&data();
+    foreach $t (@items) {
+      $d=~s/\.$t/_$t/g;
+    }
+    print $d;
+}
+
+sub data {
+	local($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...
+
+.set BO_IF_NOT,4
+.set BO_IF,12
+.set BO_dCTR_NZERO,16
+.set BO_dCTR_ZERO,18
+.set BO_ALWAYS,20
+.set CR0_LT,0;
+.set CR0_GT,1;
+.set CR0_EQ,2
+.set CR1_FX,4;
+.set CR1_FEX,5;
+.set CR1_VX,6
+.set LR,8
+
+#	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	$ISA
+
+#
+#	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
+	bclr	BO_ALWAYS,CR0_LT
+	.long	0x00000000
+
+#
+#	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
+# doesnt 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;
+
+
+	bclr	BO_ALWAYS,CR0_LT
+
+	.long	0x00000000
+
+#
+#	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`(r4)
+	$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
+	bclr	BO_ALWAYS,CR0_LT
+	.long	0x00000000
+
+#
+#	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 didnt 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;
+	bclr	BO_ALWAYS,CR0_LT
+	.long	0x00000000
+
+#
+#	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.
+	bc	BO_IF,CR0_EQ,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)
+	bc	BO_dCTR_NZERO,CR0_EQ,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.
+	bclr	BO_ALWAYS,CR0_LT
+	.long	0x00000000
+
+
+#
+#	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.
+	bc	BO_IF,CR0_EQ,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)
+	bc	BO_dCTR_NZERO,CR0_EQ,Lppcasm_add_mainloop
+Lppcasm_add_adios:	
+	addze	r3,r0			#return carry bit.
+	bclr	BO_ALWAYS,CR0_LT
+	.long	0x00000000
+
+#
+#	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
+	bc	BO_IF_NOT,CR0_EQ,Lppcasm_div1	# proceed if d!=0
+	li	r3,-1			# d=0 return -1
+	bclr	BO_ALWAYS,CR0_LT	
+Lppcasm_div1:
+	xor	r0,r0,r0		#r0=0
+	$COUNTZ	r7,r5			#r7 = num leading 0s in d.
+	subfic	r8,r7,$BITS		#r8 = BN_num_bits_word(d)
+	cmpi	0,0,r8,$BITS		#
+	bc	BO_IF,CR0_EQ,Lppcasm_div2	#proceed if (r8==$BITS)	
+	li	r9,1			# r9=1
+	$SHL	r10,r9,r8		# r9<<=r8
+	$UCMP	0,r3,r10		#	
+	bc	BO_IF,CR0_GT,Lppcasm_div2	#or if (h > (1<<r8))
+	$UDIV	r3,r3,r0		#if not assert(0) divide by 0!
+					#that's how we signal overflow
+	bclr	BO_ALWAYS,CR0_LT	#return. NEVER REACHED.
+Lppcasm_div2:
+	$UCMP	0,r3,r5			#h>=d?
+	bc	BO_IF,CR0_LT,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)?
+	bc	BO_IF,CR0_EQ,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
+	bc	BO_IF_NOT,CR0_EQ,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	1,r6,r7			# compare (tl <= r7)
+	bc	BO_IF_NOT,CR0_EQ,Lppcasm_divinnerexit
+	bc	BO_IF_NOT,CR1_FEX,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	1,r4,r11		# compare l and tl
+	add	r12,r12,r10		# th+=t
+	bc	BO_IF_NOT,CR1_FX,Lppcasm_div7  # if (l>=tl) goto Lppcasm_div7
+	addi	r12,r12,1		# th++
+Lppcasm_div7:
+	subf	r11,r11,r4		#r11=l-tl
+	$UCMP	1,r3,r12		#compare h and th
+	bc	BO_IF_NOT,CR1_FX,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
+	bc	BO_dCTR_ZERO,CR0_EQ,Lppcasm_div9#if (count==0) break ;
+	$SHLI	r0,r8,`$BITS/2`		#ret =q<<BN_BITS4
+	b	Lppcasm_divouterloop
+Lppcasm_div9:
+	or	r3,r8,r0
+	bclr	BO_ALWAYS,CR0_LT
+	.long	0x00000000
+
+#
+#	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.
+	bc	BO_IF,CR0_EQ,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)
+	bc	BO_dCTR_NZERO,CR0_EQ,Lppcasm_sqr_mainloop
+Lppcasm_sqr_adios:	
+	bclr	BO_ALWAYS,CR0_LT
+	.long	0x00000000
+
+
+#
+#	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
+	bc	BO_IF,CR0_EQ,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`
+	bc	BO_dCTR_NZERO,CR0_EQ,Lppcasm_mw_LOOP
+
+Lppcasm_mw_REM:
+	andi.	r5,r5,0x3
+	bc	BO_IF,CR0_EQ,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
+	bc	BO_IF,CR0_EQ,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
+	bc	BO_IF,CR0_EQ,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
+	bclr	BO_ALWAYS,CR0_LT
+	.long	0x00000000
+
+#
+#	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
+	bc	BO_IF,CR0_EQ,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`
+	bc	BO_dCTR_NZERO,CR0_EQ,Lppcasm_maw_mainloop
+	
+Lppcasm_maw_leftover:
+	andi.	r5,r5,0x3
+	bc	BO_IF,CR0_EQ,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)
+	
+	bc	BO_dCTR_ZERO,CR0_EQ,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)
+	
+	bc	BO_dCTR_ZERO,CR0_EQ,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
+	bclr	BO_ALWAYS,CR0_LT
+	.long	0x00000000
+	.align	4
+EOF
+	$data =~ s/\`([^\`]*)\`/eval $1/gem;
+
+	# if some assembler chokes on some simplified mnemonic,
+	# this is the spot to fix it up, e.g.:
+	# GNU as doesn't seem to accept cmplw, 32-bit unsigned compare
+	$data =~ s/^(\s*)cmplw(\s+)([^,]+),(.*)/$1cmpl$2$3,0,$4/gm;
+	# assembler X doesn't accept li, load immediate value
+	#$data =~ s/^(\s*)li(\s+)([^,]+),(.*)/$1addi$2$3,0,$4/gm;
+	return($data);
+}
diff --git a/crypto/evp/e_old.c b/crypto/evp/e_old.c
new file mode 100644
index 0000000000..66f3bf4276
--- /dev/null
+++ b/crypto/evp/e_old.c
@@ -0,0 +1,114 @@
+/* crypto/evp/e_old.c -*- mode:C; c-file-style: "eay" -*- */
+/* Written by Richard Levitte (richard@levitte.org) for the OpenSSL
+ * project 2004.
+ */
+/* ====================================================================
+ * Copyright (c) 2004 The OpenSSL Project.  All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer. 
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in
+ *    the documentation and/or other materials provided with the
+ *    distribution.
+ *
+ * 3. All advertising materials mentioning features or use of this
+ *    software must display the following acknowledgment:
+ *    "This product includes software developed by the OpenSSL Project
+ *    for use in the OpenSSL Toolkit. (http://www.openssl.org/)"
+ *
+ * 4. The names "OpenSSL Toolkit" and "OpenSSL Project" must not be used to
+ *    endorse or promote products derived from this software without
+ *    prior written permission. For written permission, please contact
+ *    openssl-core@openssl.org.
+ *
+ * 5. Products derived from this software may not be called "OpenSSL"
+ *    nor may "OpenSSL" appear in their names without prior written
+ *    permission of the OpenSSL Project.
+ *
+ * 6. Redistributions of any form whatsoever must retain the following
+ *    acknowledgment:
+ *    "This product includes software developed by the OpenSSL Project
+ *    for use in the OpenSSL Toolkit (http://www.openssl.org/)"
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE OpenSSL PROJECT ``AS IS'' AND ANY
+ * EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+ * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE OpenSSL PROJECT OR
+ * ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
+ * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
+ * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
+ * OF THE POSSIBILITY OF SUCH DAMAGE.
+ * ====================================================================
+ *
+ * This product includes cryptographic software written by Eric Young
+ * (eay@cryptsoft.com).  This product includes software written by Tim
+ * Hudson (tjh@cryptsoft.com).
+ *
+ */
+
+#ifdef OPENSSL_NO_DEPRECATED
+static void *dummy = &dummy;
+#else
+
+#include <openssl/evp.h>
+
+/* Define some deprecated functions, so older programs
+   don't crash and burn too quickly.  On Windows and VMS,
+   these will never be used, since functions and variables
+   in shared libraries are selected by entry point location,
+   not by name.  */
+
+#ifndef OPENSSL_NO_BF
+#undef EVP_bf_cfb
+const EVP_CIPHER *EVP_bf_cfb(void) { return EVP_bf_cfb64(); }
+#endif
+
+#ifndef OPENSSL_NO_DES
+#undef EVP_des_cfb
+const EVP_CIPHER *EVP_des_cfb(void) { return EVP_des_cfb64(); }
+#undef EVP_des_ede3_cfb
+const EVP_CIPHER *EVP_des_ede3_cfb(void) { return EVP_des_ede3_cfb64(); }
+#undef EVP_des_ede_cfb
+const EVP_CIPHER *EVP_des_ede_cfb(void) { return EVP_des_ede_cfb64(); }
+#endif
+
+#ifndef OPENSSL_NO_IDEA
+#undef EVP_idea_cfb
+const EVP_CIPHER *EVP_idea_cfb(void) { return EVP_idea_cfb64(); }
+#endif
+
+#ifndef OPENSSL_NO_RC2
+#undef EVP_rc2_cfb
+const EVP_CIPHER *EVP_rc2_cfb(void) { return EVP_rc2_cfb64(); }
+#endif
+
+#ifndef OPENSSL_NO_CAST5
+#undef EVP_cast5_cfb
+const EVP_CIPHER *EVP_cast5_cfb(void) { return EVP_cast5_cfb64(); }
+#endif
+
+#ifndef OPENSSL_NO_RC5
+#undef EVP_rc5_32_12_16_cfb
+const EVP_CIPHER *EVP_rc5_32_12_16_cfb(void) { return EVP_rc5_32_12_16_cfb64(); }
+#endif
+
+#ifndef OPENSSL_NO_AES
+#undef EVP_aes_128_cfb
+const EVP_CIPHER *EVP_aes_128_cfb(void) { return EVP_aes_128_cfb128(); }
+#undef EVP_aes_192_cfb
+const EVP_CIPHER *EVP_aes_192_cfb(void) { return EVP_aes_192_cfb128(); }
+#undef EVP_aes_256_cfb
+const EVP_CIPHER *EVP_aes_256_cfb(void) { return EVP_aes_256_cfb128(); }
+#endif
+
+#endif
diff --git a/crypto/sha/asm/sha512-sse2.pl b/crypto/sha/asm/sha512-sse2.pl
new file mode 100644
index 0000000000..797aedacd7
--- /dev/null
+++ b/crypto/sha/asm/sha512-sse2.pl
@@ -0,0 +1,391 @@
+#!/usr/bin/env perl
+#
+# ====================================================================
+# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
+# project. Rights for redistribution and usage in source and binary
+# forms are granted according to the OpenSSL license.
+# ====================================================================
+#
+# SHA512_Transform_SSE2.
+#
+# As the name suggests, this is an IA-32 SSE2 implementation of
+# SHA512_Transform. Motivating factor for the undertaken effort was that
+# SHA512 was observed to *consistently* perform *significantly* poorer
+# than SHA256 [2x and slower is common] on 32-bit platforms. On 64-bit
+# platforms on the other hand SHA512 tend to outperform SHA256 [~50%
+# seem to be common improvement factor]. All this is perfectly natural,
+# as SHA512 is a 64-bit algorithm. But isn't IA-32 SSE2 essentially
+# a 64-bit instruction set? Is it rich enough to implement SHA512?
+# If answer was "no," then you wouldn't have been reading this...
+#
+# Throughput performance in MBps (larger is better):
+#
+#		2.4GHz P4	1.4GHz AMD32	1.4GHz AMD64(*)
+# SHA256/gcc(*)	54		43		59
+# SHA512/gcc	17		23		92
+# SHA512/sse2	54(**)		55(**)
+# SHA512/icc	26		28
+# SHA256/icc(*)	65		54
+#
+# (*)	AMD64 and SHA256 numbers are presented mostly for amusement or
+#	reference purposes.
+# (**)	I.e. it gives ~2-3x speed-up if compared with compiler generated
+#	code. One can argue that hand-coded *non*-SSE2 implementation
+#	would perform better than compiler generated one as well, and
+#	that comparison is therefore not exactly fair. Well, as SHA512
+#	puts enormous pressure on IA-32 GP register bank, I reckon that
+#	hand-coded version wouldn't perform significantly better than
+#	one compiled with icc, ~20% perhaps... So that this code would
+#	still outperform it with distinguishing marginal. But feel free
+#	to prove me wrong:-)
+#						<appro@fy.chalmers.se>
+push(@INC,"perlasm","../../perlasm");
+require "x86asm.pl";
+
+&asm_init($ARGV[0],"sha512-sse2.pl",$ARGV[$#ARGV] eq "386");
+
+$K512="esi";	# K512[80] table, found at the end...
+#$W512="esp";	# $W512 is not just W512[16]: it comprises *two* copies
+		# of W512[16] and a copy of A-H variables...
+$W512_SZ=8*(16+16+8);	# see above...
+#$Kidx="ebx";	# index in K512 table, advances from 0 to 80...
+$Widx="edx";	# index in W512, wraps around at 16...
+$data="edi";	# 16 qwords of input data...
+$A="mm0";	# B-D and
+$E="mm1";	# F-H are allocated dynamically...
+$Aoff=256+0;	# A-H offsets relative to $W512...
+$Boff=256+8;
+$Coff=256+16;
+$Doff=256+24;
+$Eoff=256+32;
+$Foff=256+40;
+$Goff=256+48;
+$Hoff=256+56;
+
+sub SHA2_ROUND()
+{ local ($kidx,$widx)=@_;
+
+	# One can argue that one could reorder instructions for better
+	# performance. Well, I tried and it doesn't seem to make any
+	# noticeable difference. Modern out-of-order execution cores
+	# reorder instructions to their liking in either case and they
+	# apparently do decent job. So we can keep the code more
+	# readable/regular/comprehensible:-)
+
+	# I adhere to 64-bit %mmX registers in order to avoid/not care
+	# about #GP exceptions on misaligned 128-bit access, most
+	# notably in paddq with memory operand. Not to mention that
+	# SSE2 intructions operating on %mmX can be scheduled every
+	# cycle [and not every second one if operating on %xmmN].
+
+	&movq	("mm4",&QWP($Foff,$W512));	# load f
+	&movq	("mm5",&QWP($Goff,$W512));	# load g
+	&movq	("mm6",&QWP($Hoff,$W512));	# load h
+	&movq	(&QWP($Foff,$W512),$E);		# f = e
+	&movq	(&QWP($Goff,$W512),"mm4");	# g = f
+	&movq	(&QWP($Hoff,$W512),"mm5");	# h = g
+
+	&movq	("mm2",$E);			# %mm2 is sliding right
+	&movq	("mm3",$E);			# %mm3 is sliding left
+	&psrlq	("mm2",14);
+	&psllq	("mm3",23);
+	&movq	("mm7","mm2");			# %mm7 is T1
+	&pxor	("mm7","mm3");
+	&psrlq	("mm2",4);
+	&psllq	("mm3",23);
+	&pxor	("mm7","mm2");
+	&pxor	("mm7","mm3");
+	&psrlq	("mm2",23);
+	&psllq	("mm3",4);
+	&pxor	("mm7","mm2");
+	&pxor	("mm7","mm3");			# T1=Sigma1_512(e)
+
+	&pxor	("mm4","mm5");			# f^=g
+	&pand	("mm4",$E);			# f&=e
+	&pxor	("mm4","mm5");			# f^=g
+	&paddq	("mm7","mm4");			# T1+=Ch(e,f,g)
+
+	&movq	("mm2",&QWP($Boff,$W512));	# load b
+	&movq	("mm3",&QWP($Coff,$W512));	# load c
+	&movq	($E,&QWP($Doff,$W512));		# e = d
+	&movq	(&QWP($Boff,$W512),$A);		# b = a
+	&movq	(&QWP($Coff,$W512),"mm2");	# c = b
+	&movq	(&QWP($Doff,$W512),"mm3");	# d = c
+
+	&paddq	("mm7","mm6");			# T1+=h
+	&paddq	("mm7",&QWP(0,$K512,$kidx,8));	# T1+=K512[i]
+	&paddq	("mm7",&QWP(0,$W512,$widx,8));	# T1+=W512[i]
+	&paddq	($E,"mm7");			# e += T1
+
+	&movq	("mm4",$A);			# %mm4 is sliding right
+	&movq	("mm5",$A);			# %mm5 is sliding left
+	&psrlq	("mm4",28);
+	&psllq	("mm5",25);
+	&movq	("mm6","mm4");			# %mm6 is T2
+	&pxor	("mm6","mm5");
+	&psrlq	("mm4",6);
+	&psllq	("mm5",5);
+	&pxor	("mm6","mm4");
+	&pxor	("mm6","mm5");
+	&psrlq	("mm4",5);
+	&psllq	("mm5",6);
+	&pxor	("mm6","mm4");
+	&pxor	("mm6","mm5");			# T2=Sigma0_512(a)
+
+	&movq	("mm4","mm2");			# %mm4=b
+	&pand	("mm2",$A);			# b&=a
+	&pand	("mm4","mm3");			# %mm4&=c
+	&pand	("mm3",$A);			# c&=a
+	&pxor	("mm4","mm2");			# %mm4^=b&a
+	&pxor	("mm4","mm3");			# %mm4^=c&a
+	&paddq	("mm6","mm4");			# T2+=Maj(a,b,c)
+
+	&movq	($A,"mm7");			# a=T1
+	&paddq	($A,"mm6");			# a+=T2
+}
+
+$func="sha512_block_sse2";
+
+&function_begin_B($func);
+	if (0) {# Caller is expected to check if it's appropriate to
+		# call this routine. Below 3 lines are retained for
+		# debugging purposes...
+		&picmeup("eax","OPENSSL_ia32cap");
+		&bt	(&DWP(0,"eax"),26);
+		&jnc	("SHA512_Transform");
+	}
+
+	&push	("ebp");
+	&mov	("ebp","esp");
+	&push	("ebx");
+	&push	("esi");
+	&push	("edi");
+
+	&mov	($Widx,&DWP(8,"ebp"));		# A-H state, 1st arg
+	&mov	($data,&DWP(12,"ebp"));		# input data, 2nd arg
+	&call	(&label("pic_point"));		# make it PIC!
+&set_label("pic_point");
+	&blindpop($K512);
+	&lea	($K512,&DWP(&label("K512")."-".&label("pic_point"),$K512));
+
+	$W512 = "esp";			# start using %esp as W512
+	&sub	($W512,$W512_SZ);
+	&and	($W512,-16);		# ensure 128-bit alignment
+
+	# make private copy of A-H
+	#     v assume the worst and stick to unaligned load
+	&movdqu	("xmm0",&QWP(0,$Widx));
+	&movdqu	("xmm1",&QWP(16,$Widx));
+	&movdqu	("xmm2",&QWP(32,$Widx));
+	&movdqu	("xmm3",&QWP(48,$Widx));
+
+&align(8);
+&set_label("_chunk_loop");
+
+	&movdqa	(&QWP($Aoff,$W512),"xmm0");	# a,b
+	&movdqa	(&QWP($Coff,$W512),"xmm1");	# c,d
+	&movdqa	(&QWP($Eoff,$W512),"xmm2");	# e,f
+	&movdqa	(&QWP($Goff,$W512),"xmm3");	# g,h
+
+	&xor	($Widx,$Widx);
+
+	&movdq2q($A,"xmm0");			# load a
+	&movdq2q($E,"xmm2");			# load e
+
+	# Why aren't loops unrolled? It makes sense to unroll if
+	# execution time for loop body is comparable with branch
+	# penalties and/or if whole data-set resides in register bank.
+	# Neither is case here... Well, it would be possible to
+	# eliminate few store operations, but it would hardly affect
+	# so to say stop-watch performance, as there is a lot of
+	# available memory slots to fill. It will only relieve some
+	# pressure off memory bus...
+
+&align(8);
+&set_label("_1st_loop");		# 0-15
+	# flip input stream byte order...
+	&mov	("eax",&DWP(0,$data,$Widx,8));
+	&mov	("ebx",&DWP(4,$data,$Widx,8));
+	&bswap	("eax");
+	&bswap	("ebx");
+	&mov	(&DWP(0,$W512,$Widx,8),"ebx");		# W512[i]
+	&mov	(&DWP(4,$W512,$Widx,8),"eax");
+	&mov	(&DWP(128+0,$W512,$Widx,8),"ebx");	# copy of W512[i]
+	&mov	(&DWP(128+4,$W512,$Widx,8),"eax");
+
+	&SHA2_ROUND($Widx,$Widx); &inc($Widx);
+
+&cmp	($Widx,16)
+&jl	(&label("_1st_loop"));
+
+	$Kidx = "ebx";			# start using %ebx as Kidx
+	&mov	($Kidx,$Widx);
+
+&align(8);
+&set_label("_2nd_loop");		# 16-79
+	&and($Widx,0xf);
+
+	# 128-bit fragment! I update W512[i] and W512[i+1] in
+	# parallel:-) Note that I refer to W512[(i&0xf)+N] and not to
+	# W512[(i+N)&0xf]! This is exactly what I maintain the second
+	# copy of W512[16] for...
+	&movdqu	("xmm0",&QWP(8*1,$W512,$Widx,8));	# s0=W512[i+1]
+	&movdqa	("xmm2","xmm0");		# %xmm2 is sliding right
+	&movdqa	("xmm3","xmm0");		# %xmm3 is sliding left
+	&psrlq	("xmm2",1);
+	&psllq	("xmm3",56);
+	&movdqa	("xmm0","xmm2");
+	&pxor	("xmm0","xmm3");
+	&psrlq	("xmm2",6);
+	&psllq	("xmm3",7);
+	&pxor	("xmm0","xmm2");
+	&pxor	("xmm0","xmm3");
+	&psrlq	("xmm2",1);
+	&pxor	("xmm0","xmm2");		# s0 = sigma0_512(s0);
+
+	&movdqa	("xmm1",&QWP(8*14,$W512,$Widx,8));	# s1=W512[i+14]
+	&movdqa	("xmm4","xmm1");		# %xmm4 is sliding right
+	&movdqa	("xmm5","xmm1");		# %xmm5 is sliding left
+	&psrlq	("xmm4",6);
+	&psllq	("xmm5",3);
+	&movdqa	("xmm1","xmm4");
+	&pxor	("xmm1","xmm5");
+	&psrlq	("xmm4",13);
+	&psllq	("xmm5",42);
+	&pxor	("xmm1","xmm4");
+	&pxor	("xmm1","xmm5");
+	&psrlq	("xmm4",42);
+	&pxor	("xmm1","xmm4");		# s1 = sigma1_512(s1);
+
+	#     + have to explictly load W512[i+9] as it's not 128-bit
+	#     v	aligned and paddq would throw an exception...
+	&movdqu	("xmm6",&QWP(8*9,$W512,$Widx,8));
+	&paddq	("xmm0","xmm1");		# s0 += s1
+	&paddq	("xmm0","xmm6");		# s0 += W512[i+9]
+	&paddq	("xmm0",&QWP(0,$W512,$Widx,8));	# s0 += W512[i]
+
+	&movdqa	(&QWP(0,$W512,$Widx,8),"xmm0");		# W512[i] = s0
+	&movdqa	(&QWP(16*8,$W512,$Widx,8),"xmm0");	# copy of W512[i]
+
+	# as the above fragment was 128-bit, we "owe" 2 rounds...
+	&SHA2_ROUND($Kidx,$Widx); &inc($Kidx); &inc($Widx);
+	&SHA2_ROUND($Kidx,$Widx); &inc($Kidx); &inc($Widx);
+
+&cmp	($Kidx,80);
+&jl	(&label("_2nd_loop"));
+
+	# update A-H state
+	&mov	($Widx,&DWP(8,"ebp"));		# A-H state, 1st arg
+	&movq	(&QWP($Aoff,$W512),$A);		# write out a
+	&movq	(&QWP($Eoff,$W512),$E);		# write out e
+	&movdqu	("xmm0",&QWP(0,$Widx));
+	&movdqu	("xmm1",&QWP(16,$Widx));
+	&movdqu	("xmm2",&QWP(32,$Widx));
+	&movdqu	("xmm3",&QWP(48,$Widx));
+	&paddq	("xmm0",&QWP($Aoff,$W512));	# 128-bit additions...
+	&paddq	("xmm1",&QWP($Coff,$W512));
+	&paddq	("xmm2",&QWP($Eoff,$W512));
+	&paddq	("xmm3",&QWP($Goff,$W512));
+	&movdqu	(&QWP(0,$Widx),"xmm0");
+	&movdqu	(&QWP(16,$Widx),"xmm1");
+	&movdqu	(&QWP(32,$Widx),"xmm2");
+	&movdqu	(&QWP(48,$Widx),"xmm3");
+
+&add	($data,16*8);				# advance input data pointer
+&dec	(&DWP(16,"ebp"));			# decrement 3rd arg
+&jnz	(&label("_chunk_loop"));
+
+	# epilogue
+	&emms	();	# required for at least ELF and Win32 ABIs
+	&mov	("edi",&DWP(-12,"ebp"));
+	&mov	("esi",&DWP(-8,"ebp"));
+	&mov	("ebx",&DWP(-4,"ebp"));
+	&leave	();
+&ret	();
+
+&align(16);
+&set_label("K512");	# Yes! I keep it in the code segment!
+	&data_word(0xd728ae22,0x428a2f98);	# u64
+	&data_word(0x23ef65cd,0x71374491);	# u64
+	&data_word(0xec4d3b2f,0xb5c0fbcf);	# u64
+	&data_word(0x8189dbbc,0xe9b5dba5);	# u64
+	&data_word(0xf348b538,0x3956c25b);	# u64
+	&data_word(0xb605d019,0x59f111f1);	# u64
+	&data_word(0xaf194f9b,0x923f82a4);	# u64
+	&data_word(0xda6d8118,0xab1c5ed5);	# u64
+	&data_word(0xa3030242,0xd807aa98);	# u64
+	&data_word(0x45706fbe,0x12835b01);	# u64
+	&data_word(0x4ee4b28c,0x243185be);	# u64
+	&data_word(0xd5ffb4e2,0x550c7dc3);	# u64
+	&data_word(0xf27b896f,0x72be5d74);	# u64
+	&data_word(0x3b1696b1,0x80deb1fe);	# u64
+	&data_word(0x25c71235,0x9bdc06a7);	# u64
+	&data_word(0xcf692694,0xc19bf174);	# u64
+	&data_word(0x9ef14ad2,0xe49b69c1);	# u64
+	&data_word(0x384f25e3,0xefbe4786);	# u64
+	&data_word(0x8b8cd5b5,0x0fc19dc6);	# u64
+	&data_word(0x77ac9c65,0x240ca1cc);	# u64
+	&data_word(0x592b0275,0x2de92c6f);	# u64
+	&data_word(0x6ea6e483,0x4a7484aa);	# u64
+	&data_word(0xbd41fbd4,0x5cb0a9dc);	# u64
+	&data_word(0x831153b5,0x76f988da);	# u64
+	&data_word(0xee66dfab,0x983e5152);	# u64
+	&data_word(0x2db43210,0xa831c66d);	# u64
+	&data_word(0x98fb213f,0xb00327c8);	# u64
+	&data_word(0xbeef0ee4,0xbf597fc7);	# u64
+	&data_word(0x3da88fc2,0xc6e00bf3);	# u64
+	&data_word(0x930aa725,0xd5a79147);	# u64
+	&data_word(0xe003826f,0x06ca6351);	# u64
+	&data_word(0x0a0e6e70,0x14292967);	# u64
+	&data_word(0x46d22ffc,0x27b70a85);	# u64
+	&data_word(0x5c26c926,0x2e1b2138);	# u64
+	&data_word(0x5ac42aed,0x4d2c6dfc);	# u64
+	&data_word(0x9d95b3df,0x53380d13);	# u64
+	&data_word(0x8baf63de,0x650a7354);	# u64
+	&data_word(0x3c77b2a8,0x766a0abb);	# u64
+	&data_word(0x47edaee6,0x81c2c92e);	# u64
+	&data_word(0x1482353b,0x92722c85);	# u64
+	&data_word(0x4cf10364,0xa2bfe8a1);	# u64
+	&data_word(0xbc423001,0xa81a664b);	# u64
+	&data_word(0xd0f89791,0xc24b8b70);	# u64
+	&data_word(0x0654be30,0xc76c51a3);	# u64
+	&data_word(0xd6ef5218,0xd192e819);	# u64
+	&data_word(0x5565a910,0xd6990624);	# u64
+	&data_word(0x5771202a,0xf40e3585);	# u64
+	&data_word(0x32bbd1b8,0x106aa070);	# u64
+	&data_word(0xb8d2d0c8,0x19a4c116);	# u64
+	&data_word(0x5141ab53,0x1e376c08);	# u64
+	&data_word(0xdf8eeb99,0x2748774c);	# u64
+	&data_word(0xe19b48a8,0x34b0bcb5);	# u64
+	&data_word(0xc5c95a63,0x391c0cb3);	# u64
+	&data_word(0xe3418acb,0x4ed8aa4a);	# u64
+	&data_word(0x7763e373,0x5b9cca4f);	# u64
+	&data_word(0xd6b2b8a3,0x682e6ff3);	# u64
+	&data_word(0x5defb2fc,0x748f82ee);	# u64
+	&data_word(0x43172f60,0x78a5636f);	# u64
+	&data_word(0xa1f0ab72,0x84c87814);	# u64
+	&data_word(0x1a6439ec,0x8cc70208);	# u64
+	&data_word(0x23631e28,0x90befffa);	# u64
+	&data_word(0xde82bde9,0xa4506ceb);	# u64
+	&data_word(0xb2c67915,0xbef9a3f7);	# u64
+	&data_word(0xe372532b,0xc67178f2);	# u64
+	&data_word(0xea26619c,0xca273ece);	# u64
+	&data_word(0x21c0c207,0xd186b8c7);	# u64
+	&data_word(0xcde0eb1e,0xeada7dd6);	# u64
+	&data_word(0xee6ed178,0xf57d4f7f);	# u64
+	&data_word(0x72176fba,0x06f067aa);	# u64
+	&data_word(0xa2c898a6,0x0a637dc5);	# u64
+	&data_word(0xbef90dae,0x113f9804);	# u64
+	&data_word(0x131c471b,0x1b710b35);	# u64
+	&data_word(0x23047d84,0x28db77f5);	# u64
+	&data_word(0x40c72493,0x32caab7b);	# u64
+	&data_word(0x15c9bebc,0x3c9ebe0a);	# u64
+	&data_word(0x9c100d4c,0x431d67c4);	# u64
+	&data_word(0xcb3e42b6,0x4cc5d4be);	# u64
+	&data_word(0xfc657e2a,0x597f299c);	# u64
+	&data_word(0x3ad6faec,0x5fcb6fab);	# u64
+	&data_word(0x4a475817,0x6c44198c);	# u64
+
+&function_end_B($func);
+
+&asm_finish();
diff --git a/crypto/sha/sha256.c b/crypto/sha/sha256.c
new file mode 100644
index 0000000000..2f4078f444
--- /dev/null
+++ b/crypto/sha/sha256.c
@@ -0,0 +1,309 @@
+/* crypto/sha/sha256.c */
+/* ====================================================================
+ * Copyright (c) 2004 The OpenSSL Project.  All rights reserved
+ * according to the OpenSSL license [found in ../../LICENSE].
+ * ====================================================================
+ */
+#include <stdlib.h>
+#include <string.h>
+
+#include <openssl/opensslconf.h>
+#include <openssl/crypto.h>
+#include <openssl/sha.h>
+#include <openssl/opensslv.h>
+
+const char *SHA256_version="SHA-256" OPENSSL_VERSION_PTEXT;
+
+int SHA224_Init (SHA256_CTX *c)
+	{
+	c->h[0]=0xc1059ed8UL;	c->h[1]=0x367cd507UL;
+	c->h[2]=0x3070dd17UL;	c->h[3]=0xf70e5939UL;
+	c->h[4]=0xffc00b31UL;	c->h[5]=0x68581511UL;
+	c->h[6]=0x64f98fa7UL;	c->h[7]=0xbefa4fa4UL;
+	c->Nl=0;	c->Nh=0;
+	c->num=0;	c->md_len=SHA224_DIGEST_LENGTH;
+	return 1;
+	}
+
+int SHA256_Init (SHA256_CTX *c)
+	{
+	c->h[0]=0x6a09e667UL;	c->h[1]=0xbb67ae85UL;
+	c->h[2]=0x3c6ef372UL;	c->h[3]=0xa54ff53aUL;
+	c->h[4]=0x510e527fUL;	c->h[5]=0x9b05688cUL;
+	c->h[6]=0x1f83d9abUL;	c->h[7]=0x5be0cd19UL;
+	c->Nl=0;	c->Nh=0;
+	c->num=0;	c->md_len=SHA256_DIGEST_LENGTH;
+	return 1;
+	}
+
+unsigned char *SHA224(const unsigned char *d, size_t n, unsigned char *md)
+	{
+	SHA256_CTX c;
+	static unsigned char m[SHA224_DIGEST_LENGTH];
+
+	if (md == NULL) md=m;
+	SHA224_Init(&c);
+	SHA256_Update(&c,d,n);
+	SHA256_Final(md,&c);
+	OPENSSL_cleanse(&c,sizeof(c));
+	return(md);
+	}
+
+unsigned char *SHA256(const unsigned char *d, size_t n, unsigned char *md)
+	{
+	SHA256_CTX c;
+	static unsigned char m[SHA256_DIGEST_LENGTH];
+
+	if (md == NULL) md=m;
+	SHA256_Init(&c);
+	SHA256_Update(&c,d,n);
+	SHA256_Final(md,&c);
+	OPENSSL_cleanse(&c,sizeof(c));
+	return(md);
+	}
+
+int SHA224_Update(SHA256_CTX *c, const void *data, size_t len)
+{   return SHA256_Update (c,data,len);   }
+int SHA224_Final (unsigned char *md, SHA256_CTX *c)
+{   return SHA256_Final (md,c);   }
+
+#ifndef	SHA_LONG_LOG2
+#define	SHA_LONG_LOG2	2	/* default to 32 bits */
+#endif
+
+#define	DATA_ORDER_IS_BIG_ENDIAN
+
+#define	HASH_LONG		SHA_LONG
+#define	HASH_LONG_LOG2		SHA_LONG_LOG2
+#define	HASH_CTX		SHA256_CTX
+#define	HASH_CBLOCK		SHA_CBLOCK
+#define	HASH_LBLOCK		SHA_LBLOCK
+/*
+ * Note that FIPS180-2 discusses "Truncation of the Hash Function Output."
+ * default: case below covers for it. It's not clear however if it's
+ * permitted to truncate to amount of bytes not divisible by 4. I bet not,
+ * but if it is, then default: case shall be extended. For reference.
+ * Idea behind separate cases for pre-defined lenghts is to let the
+ * compiler decide if it's appropriate to unroll small loops.
+ */
+#define	HASH_MAKE_STRING(c,s)	do {	\
+	unsigned long ll;		\
+	unsigned int  n;		\
+	switch ((c)->md_len)		\
+	{   case SHA224_DIGEST_LENGTH:	\
+		for (n=0;n<SHA224_DIGEST_LENGTH/4;n++)	\
+		{   ll=(c)->h[n]; HOST_l2c(ll,(s));   }	\
+		break;			\
+	    case SHA256_DIGEST_LENGTH:	\
+		for (n=0;n<SHA256_DIGEST_LENGTH/4;n++)	\
+		{   ll=(c)->h[n]; HOST_l2c(ll,(s));   }	\
+		break;			\
+	    default:			\
+		if ((c)->md_len > SHA256_DIGEST_LENGTH)	\
+		    return 0;				\
+		for (n=0;n<(c)->md_len/4;n++)		\
+		{   ll=(c)->h[n]; HOST_l2c(ll,(s));   }	\
+		break;			\
+	}				\
+	} while (0)
+
+#define	HASH_UPDATE		SHA256_Update
+#define	HASH_TRANSFORM		SHA256_Transform
+#define	HASH_FINAL		SHA256_Final
+#define	HASH_BLOCK_HOST_ORDER	sha256_block_host_order
+#define	HASH_BLOCK_DATA_ORDER	sha256_block_data_order
+void sha256_block_host_order (SHA256_CTX *ctx, const void *in, size_t num);
+void sha256_block_data_order (SHA256_CTX *ctx, const void *in, size_t num);
+
+#include "md32_common.h"
+
+static const SHA_LONG K256[64] = {
+	0x428a2f98UL,0x71374491UL,0xb5c0fbcfUL,0xe9b5dba5UL,
+	0x3956c25bUL,0x59f111f1UL,0x923f82a4UL,0xab1c5ed5UL,
+	0xd807aa98UL,0x12835b01UL,0x243185beUL,0x550c7dc3UL,
+	0x72be5d74UL,0x80deb1feUL,0x9bdc06a7UL,0xc19bf174UL,
+	0xe49b69c1UL,0xefbe4786UL,0x0fc19dc6UL,0x240ca1ccUL,
+	0x2de92c6fUL,0x4a7484aaUL,0x5cb0a9dcUL,0x76f988daUL,
+	0x983e5152UL,0xa831c66dUL,0xb00327c8UL,0xbf597fc7UL,
+	0xc6e00bf3UL,0xd5a79147UL,0x06ca6351UL,0x14292967UL,
+	0x27b70a85UL,0x2e1b2138UL,0x4d2c6dfcUL,0x53380d13UL,
+	0x650a7354UL,0x766a0abbUL,0x81c2c92eUL,0x92722c85UL,
+	0xa2bfe8a1UL,0xa81a664bUL,0xc24b8b70UL,0xc76c51a3UL,
+	0xd192e819UL,0xd6990624UL,0xf40e3585UL,0x106aa070UL,
+	0x19a4c116UL,0x1e376c08UL,0x2748774cUL,0x34b0bcb5UL,
+	0x391c0cb3UL,0x4ed8aa4aUL,0x5b9cca4fUL,0x682e6ff3UL,
+	0x748f82eeUL,0x78a5636fUL,0x84c87814UL,0x8cc70208UL,
+	0x90befffaUL,0xa4506cebUL,0xbef9a3f7UL,0xc67178f2UL };
+
+/*
+ * FIPS specification refers to right rotations, while our ROTATE macro
+ * is left one. This is why you might notice that rotation coefficients
+ * differ from those observed in FIPS document by 32-N...
+ */
+#define Sigma0(x)	(ROTATE((x),30) ^ ROTATE((x),19) ^ ROTATE((x),10))
+#define Sigma1(x)	(ROTATE((x),26) ^ ROTATE((x),21) ^ ROTATE((x),7))
+#define sigma0(x)	(ROTATE((x),25) ^ ROTATE((x),14) ^ ((x)>>3))
+#define sigma1(x)	(ROTATE((x),15) ^ ROTATE((x),13) ^ ((x)>>10))
+
+#define Ch(x,y,z)	(((x) & (y)) ^ ((~(x)) & (z)))
+#define Maj(x,y,z)	(((x) & (y)) ^ ((x) & (z)) ^ ((y) & (z)))
+
+#ifdef OPENSSL_SMALL_FOOTPRINT
+
+static void sha256_block (SHA256_CTX *ctx, const void *in, size_t num, int host)
+	{
+	unsigned MD32_REG_T a,b,c,d,e,f,g,h,s0,s1,T1,T2;
+	SHA_LONG	X[16];
+	int i;
+	const unsigned char *data=in;
+
+			while (num--) {
+
+	a = ctx->h[0];	b = ctx->h[1];	c = ctx->h[2];	d = ctx->h[3];
+	e = ctx->h[4];	f = ctx->h[5];	g = ctx->h[6];	h = ctx->h[7];
+
+	if (host)
+		{
+		const SHA_LONG *W=(const SHA_LONG *)data;
+
+		for (i=0;i<16;i++)
+			{
+			T1 = X[i] = W[i];
+			T1 += h + Sigma1(e) + Ch(e,f,g) + K256[i];
+			T2 = Sigma0(a) + Maj(a,b,c);
+			h = g;	g = f;	f = e;	e = d + T1;
+			d = c;	c = b;	b = a;	a = T1 + T2;
+			}
+		}
+	else
+		{
+		SHA_LONG l;
+
+		for (i=0;i<16;i++)
+			{
+			HOST_c2l(data,l); T1 = X[i] = l;
+			T1 += h + Sigma1(e) + Ch(e,f,g) + K256[i];
+			T2 = Sigma0(a) + Maj(a,b,c);
+			h = g;	g = f;	f = e;	e = d + T1;
+			d = c;	c = b;	b = a;	a = T1 + T2;
+			}
+		}
+
+	for (;i<64;i++)
+		{
+		s0 = X[(i+1)&0x0f];	s0 = sigma0(s0);
+		s1 = X[(i+14)&0x0f];	s1 = sigma1(s1);
+
+		T1 = X[i&0xf] += s0 + s1 + X[(i+9)&0xf];
+		T1 += h + Sigma1(e) + Ch(e,f,g) + K256[i];
+		T2 = Sigma0(a) + Maj(a,b,c);
+		h = g;	g = f;	f = e;	e = d + T1;
+		d = c;	c = b;	b = a;	a = T1 + T2;
+		}
+
+	ctx->h[0] += a;	ctx->h[1] += b;	ctx->h[2] += c;	ctx->h[3] += d;
+	ctx->h[4] += e;	ctx->h[5] += f;	ctx->h[6] += g;	ctx->h[7] += h;
+
+			data += SHA256_CBLOCK;
+			}
+}
+
+#else
+
+#define	ROUND_00_15(i,a,b,c,d,e,f,g,h)		do {	\
+	T1 += h + Sigma1(e) + Ch(e,f,g) + K256[i];	\
+	h = Sigma0(a) + Maj(a,b,c);			\
+	d += T1;	h += T1;		} while (0)
+
+#define	ROUND_16_63(i,a,b,c,d,e,f,g,h,X)	do {	\
+	s0 = X[(i+1)&0x0f];	s0 = sigma0(s0);	\
+	s1 = X[(i+14)&0x0f];	s1 = sigma1(s1);	\
+	T1 = X[(i)&0x0f] += s0 + s1 + X[(i+9)&0x0f];	\
+	ROUND_00_15(i,a,b,c,d,e,f,g,h);		} while (0)
+
+static void sha256_block (SHA256_CTX *ctx, const void *in, size_t num, int host)
+	{
+	unsigned MD32_REG_T a,b,c,d,e,f,g,h,s0,s1,T1;
+	SHA_LONG	X[16];
+	int i;
+	const unsigned char *data=in;
+
+			while (num--) {
+
+	a = ctx->h[0];	b = ctx->h[1];	c = ctx->h[2];	d = ctx->h[3];
+	e = ctx->h[4];	f = ctx->h[5];	g = ctx->h[6];	h = ctx->h[7];
+
+	if (host)
+		{
+		const SHA_LONG *W=(const SHA_LONG *)data;
+
+		T1 = X[0] = W[0];	ROUND_00_15(0,a,b,c,d,e,f,g,h);
+		T1 = X[1] = W[1];	ROUND_00_15(1,h,a,b,c,d,e,f,g);
+		T1 = X[2] = W[2];	ROUND_00_15(2,g,h,a,b,c,d,e,f);
+		T1 = X[3] = W[3];	ROUND_00_15(3,f,g,h,a,b,c,d,e);
+		T1 = X[4] = W[4];	ROUND_00_15(4,e,f,g,h,a,b,c,d);
+		T1 = X[5] = W[5];	ROUND_00_15(5,d,e,f,g,h,a,b,c);
+		T1 = X[6] = W[6];	ROUND_00_15(6,c,d,e,f,g,h,a,b);
+		T1 = X[7] = W[7];	ROUND_00_15(7,b,c,d,e,f,g,h,a);
+		T1 = X[8] = W[8];	ROUND_00_15(8,a,b,c,d,e,f,g,h);
+		T1 = X[9] = W[9];	ROUND_00_15(9,h,a,b,c,d,e,f,g);
+		T1 = X[10] = W[10];	ROUND_00_15(10,g,h,a,b,c,d,e,f);
+		T1 = X[11] = W[11];	ROUND_00_15(11,f,g,h,a,b,c,d,e);
+		T1 = X[12] = W[12];	ROUND_00_15(12,e,f,g,h,a,b,c,d);
+		T1 = X[13] = W[13];	ROUND_00_15(13,d,e,f,g,h,a,b,c);
+		T1 = X[14] = W[14];	ROUND_00_15(14,c,d,e,f,g,h,a,b);
+		T1 = X[15] = W[15];	ROUND_00_15(15,b,c,d,e,f,g,h,a);
+		}
+	else
+		{
+		SHA_LONG l;
+
+		HOST_c2l(data,l); T1 = X[0] = l;  ROUND_00_15(0,a,b,c,d,e,f,g,h);
+		HOST_c2l(data,l); T1 = X[1] = l;  ROUND_00_15(1,h,a,b,c,d,e,f,g);
+		HOST_c2l(data,l); T1 = X[2] = l;  ROUND_00_15(2,g,h,a,b,c,d,e,f);
+		HOST_c2l(data,l); T1 = X[3] = l;  ROUND_00_15(3,f,g,h,a,b,c,d,e);
+		HOST_c2l(data,l); T1 = X[4] = l;  ROUND_00_15(4,e,f,g,h,a,b,c,d);
+		HOST_c2l(data,l); T1 = X[5] = l;  ROUND_00_15(5,d,e,f,g,h,a,b,c);
+		HOST_c2l(data,l); T1 = X[6] = l;  ROUND_00_15(6,c,d,e,f,g,h,a,b);
+		HOST_c2l(data,l); T1 = X[7] = l;  ROUND_00_15(7,b,c,d,e,f,g,h,a);
+		HOST_c2l(data,l); T1 = X[8] = l;  ROUND_00_15(8,a,b,c,d,e,f,g,h);
+		HOST_c2l(data,l); T1 = X[9] = l;  ROUND_00_15(9,h,a,b,c,d,e,f,g);
+		HOST_c2l(data,l); T1 = X[10] = l; ROUND_00_15(10,g,h,a,b,c,d,e,f);
+		HOST_c2l(data,l); T1 = X[11] = l; ROUND_00_15(11,f,g,h,a,b,c,d,e);
+		HOST_c2l(data,l); T1 = X[12] = l; ROUND_00_15(12,e,f,g,h,a,b,c,d);
+		HOST_c2l(data,l); T1 = X[13] = l; ROUND_00_15(13,d,e,f,g,h,a,b,c);
+		HOST_c2l(data,l); T1 = X[14] = l; ROUND_00_15(14,c,d,e,f,g,h,a,b);
+		HOST_c2l(data,l); T1 = X[15] = l; ROUND_00_15(15,b,c,d,e,f,g,h,a);
+		}
+
+	for (i=16;i<64;i+=8)
+		{
+		ROUND_16_63(i+0,a,b,c,d,e,f,g,h,X);
+		ROUND_16_63(i+1,h,a,b,c,d,e,f,g,X);
+		ROUND_16_63(i+2,g,h,a,b,c,d,e,f,X);
+		ROUND_16_63(i+3,f,g,h,a,b,c,d,e,X);
+		ROUND_16_63(i+4,e,f,g,h,a,b,c,d,X);
+		ROUND_16_63(i+5,d,e,f,g,h,a,b,c,X);
+		ROUND_16_63(i+6,c,d,e,f,g,h,a,b,X);
+		ROUND_16_63(i+7,b,c,d,e,f,g,h,a,X);
+		}
+
+	ctx->h[0] += a;	ctx->h[1] += b;	ctx->h[2] += c;	ctx->h[3] += d;
+	ctx->h[4] += e;	ctx->h[5] += f;	ctx->h[6] += g;	ctx->h[7] += h;
+
+			data += SHA256_CBLOCK;
+			}
+	}
+
+#endif
+
+/*
+ * Idea is to trade couple of cycles for some space. On IA-32 we save
+ * about 4K in "big footprint" case. In "small footprint" case any gain
+ * is appreciated:-)
+ */
+void HASH_BLOCK_HOST_ORDER (SHA256_CTX *ctx, const void *in, size_t num)
+{   sha256_block (ctx,in,num,1);   }
+
+void HASH_BLOCK_DATA_ORDER (SHA256_CTX *ctx, const void *in, size_t num)
+{   sha256_block (ctx,in,num,0);   }
diff --git a/crypto/sha/sha256t.c b/crypto/sha/sha256t.c
new file mode 100644
index 0000000000..e211c9c24f
--- /dev/null
+++ b/crypto/sha/sha256t.c
@@ -0,0 +1,130 @@
+/* crypto/sha/sha256t.c */
+/* ====================================================================
+ * Copyright (c) 2004 The OpenSSL Project.  All rights reserved.
+ * ====================================================================
+ */
+#include <stdio.h>
+#include <string.h>
+#include <stdlib.h>
+
+#include <openssl/sha.h>
+
+unsigned char app_b1[SHA256_DIGEST_LENGTH] = {
+	0xba,0x78,0x16,0xbf,0x8f,0x01,0xcf,0xea,
+	0x41,0x41,0x40,0xde,0x5d,0xae,0x22,0x23,
+	0xb0,0x03,0x61,0xa3,0x96,0x17,0x7a,0x9c,
+	0xb4,0x10,0xff,0x61,0xf2,0x00,0x15,0xad	};
+
+unsigned char app_b2[SHA256_DIGEST_LENGTH] = {
+	0x24,0x8d,0x6a,0x61,0xd2,0x06,0x38,0xb8,
+	0xe5,0xc0,0x26,0x93,0x0c,0x3e,0x60,0x39,
+	0xa3,0x3c,0xe4,0x59,0x64,0xff,0x21,0x67,
+	0xf6,0xec,0xed,0xd4,0x19,0xdb,0x06,0xc1	};
+
+unsigned char app_b3[SHA256_DIGEST_LENGTH] = {
+	0xcd,0xc7,0x6e,0x5c,0x99,0x14,0xfb,0x92,
+	0x81,0xa1,0xc7,0xe2,0x84,0xd7,0x3e,0x67,
+	0xf1,0x80,0x9a,0x48,0xa4,0x97,0x20,0x0e,
+	0x04,0x6d,0x39,0xcc,0xc7,0x11,0x2c,0xd0	};
+
+unsigned char addenum_1[SHA224_DIGEST_LENGTH] = {
+	0x23,0x09,0x7d,0x22,0x34,0x05,0xd8,0x22,
+	0x86,0x42,0xa4,0x77,0xbd,0xa2,0x55,0xb3,
+	0x2a,0xad,0xbc,0xe4,0xbd,0xa0,0xb3,0xf7,
+	0xe3,0x6c,0x9d,0xa7 };
+
+unsigned char addenum_2[SHA224_DIGEST_LENGTH] = {
+	0x75,0x38,0x8b,0x16,0x51,0x27,0x76,0xcc,
+	0x5d,0xba,0x5d,0xa1,0xfd,0x89,0x01,0x50,
+	0xb0,0xc6,0x45,0x5c,0xb4,0xf5,0x8b,0x19,
+	0x52,0x52,0x25,0x25 };
+
+unsigned char addenum_3[SHA224_DIGEST_LENGTH] = {
+	0x20,0x79,0x46,0x55,0x98,0x0c,0x91,0xd8,
+	0xbb,0xb4,0xc1,0xea,0x97,0x61,0x8a,0x4b,
+	0xf0,0x3f,0x42,0x58,0x19,0x48,0xb2,0xee,
+	0x4e,0xe7,0xad,0x67 };
+
+int main ()
+{ unsigned char md[SHA256_DIGEST_LENGTH];
+  int		i;
+  SHA256_CTX	ctx;
+
+    fprintf(stdout,"Testing SHA-256 ");
+
+    SHA256((unsigned char *)"abc",3,md);
+    if (memcmp(md,app_b1,sizeof(app_b1)))
+    {	fflush(stdout);
+	fprintf(stderr,"\nTEST 1 of 3 failed.\n");
+	return 1;
+    }
+    else
+	fprintf(stdout,"."); fflush(stdout);
+
+    SHA256((unsigned char *)"abcdbcde""cdefdefg""efghfghi""ghijhijk"
+	   "ijkljklm""klmnlmno""mnopnopq",56,md);
+    if (memcmp(md,app_b2,sizeof(app_b2)))
+    {	fflush(stdout);
+	fprintf(stderr,"\nTEST 2 of 3 failed.\n");
+	return 1;
+    }
+    else
+	fprintf(stdout,"."); fflush(stdout);
+
+    SHA256_Init(&ctx);
+    for (i=0;i<1000000;i+=64)
+	SHA256_Update(&ctx, "aaaaaaaa""aaaaaaaa""aaaaaaaa""aaaaaaaa"
+			    "aaaaaaaa""aaaaaaaa""aaaaaaaa""aaaaaaaa",
+			    (1000000-i)<64?1000000-i:64);
+    SHA256_Final(md,&ctx);
+
+    if (memcmp(md,app_b3,sizeof(app_b3)))
+    {	fflush(stdout);
+	fprintf(stderr,"\nTEST 3 of 3 failed.\n");
+	return 1;
+    }
+    else
+	fprintf(stdout,"."); fflush(stdout);
+
+    fprintf(stdout," passed.\n"); fflush(stdout);
+
+    fprintf(stdout,"Testing SHA-224 ");
+
+    SHA224((unsigned char *)"abc",3,md);
+    if (memcmp(md,addenum_1,sizeof(addenum_1)))
+    {	fflush(stdout);
+	fprintf(stderr,"\nTEST 1 of 3 failed.\n");
+	return 1;
+    }
+    else
+	fprintf(stdout,"."); fflush(stdout);
+
+    SHA224((unsigned char *)"abcdbcde""cdefdefg""efghfghi""ghijhijk"
+	   "ijkljklm""klmnlmno""mnopnopq",56,md);
+    if (memcmp(md,addenum_2,sizeof(addenum_2)))
+    {	fflush(stdout);
+	fprintf(stderr,"\nTEST 2 of 3 failed.\n");
+	return 1;
+    }
+    else
+	fprintf(stdout,"."); fflush(stdout);
+
+    SHA224_Init(&ctx);
+    for (i=0;i<1000000;i+=64)
+	SHA256_Update(&ctx, "aaaaaaaa""aaaaaaaa""aaaaaaaa""aaaaaaaa"
+			    "aaaaaaaa""aaaaaaaa""aaaaaaaa""aaaaaaaa",
+			    (1000000-i)<64?1000000-i:64);
+    SHA256_Final(md,&ctx);
+
+    if (memcmp(md,addenum_3,sizeof(addenum_3)))
+    {	fflush(stdout);
+	fprintf(stderr,"\nTEST 3 of 3 failed.\n");
+	return 1;
+    }
+    else
+	fprintf(stdout,"."); fflush(stdout);
+
+    fprintf(stdout," passed.\n"); fflush(stdout);
+
+  return 0;
+}
diff --git a/crypto/sha/sha512.c b/crypto/sha/sha512.c
new file mode 100644
index 0000000000..dc1047d59e
--- /dev/null
+++ b/crypto/sha/sha512.c
@@ -0,0 +1,478 @@
+/* crypto/sha/sha512.c */
+/* ====================================================================
+ * Copyright (c) 2004 The OpenSSL Project.  All rights reserved
+ * according to the OpenSSL license [found in ../../LICENSE].
+ * ====================================================================
+ */
+/*
+ * IMPLEMENTATION NOTES.
+ *
+ * As you might have noticed 32-bit hash algorithms:
+ *
+ * - permit SHA_LONG to be wider than 32-bit (case on CRAY);
+ * - optimized versions implement two transform functions: one operating
+ *   on [aligned] data in host byte order and one - on data in input
+ *   stream byte order;
+ * - share common byte-order neutral collector and padding function
+ *   implementations, ../md32_common.h;
+ *
+ * Neither of the above applies to this SHA-512 implementations. Reasons
+ * [in reverse order] are:
+ *
+ * - it's the only 64-bit hash algorithm for the moment of this writing,
+ *   there is no need for common collector/padding implementation [yet];
+ * - by supporting only one transform function [which operates on
+ *   *aligned* data in input stream byte order, big-endian in this case]
+ *   we minimize burden of maintenance in two ways: a) collector/padding
+ *   function is simpler; b) only one transform function to stare at;
+ * - SHA_LONG64 is required to be exactly 64-bit in order to be able to
+ *   apply a number of optimizations to mitigate potential performance
+ *   penalties caused by previous design decision;
+ *
+ * Caveat lector.
+ *
+ * Implementation relies on the fact that "long long" is 64-bit on
+ * both 32- and 64-bit platforms. If some compiler vendor comes up
+ * with 128-bit long long, adjustment to sha.h would be required.
+ * As this implementation relies on 64-bit integer type, it's totally
+ * inappropriate for platforms which don't support it, most notably
+ * 16-bit platforms.
+ *					<appro@fy.chalmers.se>
+ */
+#include <stdlib.h>
+#include <string.h>
+
+#include <openssl/opensslconf.h>
+#include <openssl/crypto.h>
+#include <openssl/sha.h>
+#include <openssl/opensslv.h>
+
+const char *SHA512_version="SHA-512" OPENSSL_VERSION_PTEXT;
+
+#if defined(_M_IX86) || defined(_M_AMD64) || defined(__i386) || defined(__x86_64)
+#define SHA512_BLOCK_CAN_MANAGE_UNALIGNED_DATA
+#endif
+
+int SHA384_Init (SHA512_CTX *c)
+	{
+	c->h[0]=U64(0xcbbb9d5dc1059ed8);
+	c->h[1]=U64(0x629a292a367cd507);
+	c->h[2]=U64(0x9159015a3070dd17);
+	c->h[3]=U64(0x152fecd8f70e5939);
+	c->h[4]=U64(0x67332667ffc00b31);
+	c->h[5]=U64(0x8eb44a8768581511);
+	c->h[6]=U64(0xdb0c2e0d64f98fa7);
+	c->h[7]=U64(0x47b5481dbefa4fa4);
+        c->Nl=0;        c->Nh=0;
+        c->num=0;       c->md_len=SHA384_DIGEST_LENGTH;
+        return 1;
+	}
+
+int SHA512_Init (SHA512_CTX *c)
+	{
+	c->h[0]=U64(0x6a09e667f3bcc908);
+	c->h[1]=U64(0xbb67ae8584caa73b);
+	c->h[2]=U64(0x3c6ef372fe94f82b);
+	c->h[3]=U64(0xa54ff53a5f1d36f1);
+	c->h[4]=U64(0x510e527fade682d1);
+	c->h[5]=U64(0x9b05688c2b3e6c1f);
+	c->h[6]=U64(0x1f83d9abfb41bd6b);
+	c->h[7]=U64(0x5be0cd19137e2179);
+        c->Nl=0;        c->Nh=0;
+        c->num=0;       c->md_len=SHA512_DIGEST_LENGTH;
+        return 1;
+	}
+
+static void sha512_block (SHA512_CTX *ctx, const void *in, size_t num);
+
+int SHA512_Final (unsigned char *md, SHA512_CTX *c)
+	{
+	unsigned char *p=(unsigned char *)c->u.p;
+	size_t n=c->num;
+
+	p[n]=0x80;	/* There always is a room for one */
+	n++;
+	if (n > (sizeof(c->u)-16))
+		memset (p+n,0,sizeof(c->u)-n), n=0,
+		sha512_block (c,p,1);
+
+	memset (p+n,0,sizeof(c->u)-16-n);
+#ifdef	B_ENDIAN
+	c->u.d[SHA_LBLOCK-2] = c->Nh;
+	c->u.d[SHA_LBLOCK-1] = c->Nl;
+#else
+	p[sizeof(c->u)-1]  = (c->Nl)&0xFF;
+	p[sizeof(c->u)-2]  = (c->Nl>>8)&0xFF;
+	p[sizeof(c->u)-3]  = (c->Nl>>16)&0xFF;
+	p[sizeof(c->u)-4]  = (c->Nl>>24)&0xFF;
+	p[sizeof(c->u)-5]  = (c->Nl>>32)&0xFF;
+	p[sizeof(c->u)-6]  = (c->Nl>>40)&0xFF;
+	p[sizeof(c->u)-7]  = (c->Nl>>48)&0xFF;
+	p[sizeof(c->u)-8]  = (c->Nl>>56)&0xFF;
+	p[sizeof(c->u)-9]  = (c->Nh)&0xFF;
+	p[sizeof(c->u)-10] = (c->Nh>>8)&0xFF;
+	p[sizeof(c->u)-11] = (c->Nh>>16)&0xFF;
+	p[sizeof(c->u)-12] = (c->Nh>>24)&0xFF;
+	p[sizeof(c->u)-13] = (c->Nh>>32)&0xFF;
+	p[sizeof(c->u)-14] = (c->Nh>>40)&0xFF;
+	p[sizeof(c->u)-15] = (c->Nh>>48)&0xFF;
+	p[sizeof(c->u)-16] = (c->Nh>>56)&0xFF;
+#endif
+
+	sha512_block (c,p,1);
+
+	if (md==0) return 0;
+
+	switch (c->md_len)
+		{
+		/* Let compiler decide if it's appropriate to unroll... */
+		case SHA384_DIGEST_LENGTH:
+			for (n=0;n<SHA384_DIGEST_LENGTH/8;n++)
+				{
+				SHA_LONG64 t = c->h[n];
+
+				*(md++)	= (t>>56)&0xFF;	*(md++)	= (t>>48)&0xFF;
+				*(md++)	= (t>>40)&0xFF;	*(md++)	= (t>>32)&0xFF;
+				*(md++)	= (t>>24)&0xFF;	*(md++)	= (t>>16)&0xFF;
+				*(md++)	= (t>>8)&0xFF;	*(md++)	= (t)&0xFF;
+				}
+			break;
+		case SHA512_DIGEST_LENGTH:
+			for (n=0;n<SHA512_DIGEST_LENGTH/8;n++)
+				{
+				SHA_LONG64 t = c->h[n];
+
+				*(md++)	= (t>>56)&0xFF;	*(md++)	= (t>>48)&0xFF;
+				*(md++)	= (t>>40)&0xFF;	*(md++)	= (t>>32)&0xFF;
+				*(md++)	= (t>>24)&0xFF;	*(md++)	= (t>>16)&0xFF;
+				*(md++)	= (t>>8)&0xFF;	*(md++)	= (t)&0xFF;
+				}
+			break;
+		/* ... as well as make sure md_len is not abused. */
+		default:	return 0;
+		}
+
+	return 1;
+	}
+
+int SHA384_Final (unsigned char *md,SHA512_CTX *c)
+{   return SHA512_Final (md,c);   }
+
+int SHA512_Update (SHA512_CTX *c, const void *_data, size_t len)
+	{
+	SHA_LONG64	l;
+	unsigned char  *p=c->u.p;
+	const unsigned char *data=(const unsigned char *)_data;
+
+	if (len==0) return  1;
+
+	l = (c->Nl+(((SHA_LONG64)len)<<3))&U64(0xffffffffffffffff);
+	if (l < c->Nl)		c->Nh++;
+	if (sizeof(len)>=8)	c->Nh+=(((SHA_LONG64)len)>>61);
+	c->Nl=l;
+
+	if (c->num != 0)
+		{
+		size_t n = sizeof(c->u) - c->num;
+
+		if (len < n)
+			{
+			memcpy (p+c->num,data,len), c->num += len;
+			return 1;
+			}
+		else	{
+			memcpy (p+c->num,data,n), c->num = 0;
+			len-=n, data+=n;
+			sha512_block (c,p,1);
+			}
+		}
+
+	if (len >= sizeof(c->u))
+		{
+#ifndef SHA512_BLOCK_CAN_MANAGE_UNALIGNED_DATA
+		if ((size_t)data%sizeof(c->u.d[0]) != 0)
+			while (len >= sizeof(c->u))
+				memcpy (p,data,sizeof(c->u)),
+				sha512_block (c,p,1),
+				len  -= sizeof(c->u),
+				data += sizeof(c->u);
+		else
+#endif
+			sha512_block (c,data,len/sizeof(c->u)),
+			data += len,
+			len  %= sizeof(c->u),
+			data -= len;
+		}
+
+	if (len != 0)	memcpy (p,data,len), c->num = (int)len;
+
+	return 1;
+	}
+
+int SHA384_Update (SHA512_CTX *c, const void *data, size_t len)
+{   return SHA512_Update (c,data,len);   }
+
+void SHA512_Transform (SHA512_CTX *c, const unsigned char *data)
+{   sha512_block (c,data,1);  }
+
+unsigned char *SHA384(const unsigned char *d, size_t n, unsigned char *md)
+	{
+	SHA512_CTX c;
+	static unsigned char m[SHA384_DIGEST_LENGTH];
+
+	if (md == NULL) md=m;
+	SHA384_Init(&c);
+	SHA512_Update(&c,d,n);
+	SHA512_Final(md,&c);
+	OPENSSL_cleanse(&c,sizeof(c));
+	return(md);
+	}
+
+unsigned char *SHA512(const unsigned char *d, size_t n, unsigned char *md)
+	{
+	SHA512_CTX c;
+	static unsigned char m[SHA512_DIGEST_LENGTH];
+
+	if (md == NULL) md=m;
+	SHA512_Init(&c);
+	SHA512_Update(&c,d,n);
+	SHA512_Final(md,&c);
+	OPENSSL_cleanse(&c,sizeof(c));
+	return(md);
+	}
+
+static const SHA_LONG64 K512[80] = {
+        U64(0x428a2f98d728ae22),U64(0x7137449123ef65cd),
+        U64(0xb5c0fbcfec4d3b2f),U64(0xe9b5dba58189dbbc),
+        U64(0x3956c25bf348b538),U64(0x59f111f1b605d019),
+        U64(0x923f82a4af194f9b),U64(0xab1c5ed5da6d8118),
+        U64(0xd807aa98a3030242),U64(0x12835b0145706fbe),
+        U64(0x243185be4ee4b28c),U64(0x550c7dc3d5ffb4e2),
+        U64(0x72be5d74f27b896f),U64(0x80deb1fe3b1696b1),
+        U64(0x9bdc06a725c71235),U64(0xc19bf174cf692694),
+        U64(0xe49b69c19ef14ad2),U64(0xefbe4786384f25e3),
+        U64(0x0fc19dc68b8cd5b5),U64(0x240ca1cc77ac9c65),
+        U64(0x2de92c6f592b0275),U64(0x4a7484aa6ea6e483),
+        U64(0x5cb0a9dcbd41fbd4),U64(0x76f988da831153b5),
+        U64(0x983e5152ee66dfab),U64(0xa831c66d2db43210),
+        U64(0xb00327c898fb213f),U64(0xbf597fc7beef0ee4),
+        U64(0xc6e00bf33da88fc2),U64(0xd5a79147930aa725),
+        U64(0x06ca6351e003826f),U64(0x142929670a0e6e70),
+        U64(0x27b70a8546d22ffc),U64(0x2e1b21385c26c926),
+        U64(0x4d2c6dfc5ac42aed),U64(0x53380d139d95b3df),
+        U64(0x650a73548baf63de),U64(0x766a0abb3c77b2a8),
+        U64(0x81c2c92e47edaee6),U64(0x92722c851482353b),
+        U64(0xa2bfe8a14cf10364),U64(0xa81a664bbc423001),
+        U64(0xc24b8b70d0f89791),U64(0xc76c51a30654be30),
+        U64(0xd192e819d6ef5218),U64(0xd69906245565a910),
+        U64(0xf40e35855771202a),U64(0x106aa07032bbd1b8),
+        U64(0x19a4c116b8d2d0c8),U64(0x1e376c085141ab53),
+        U64(0x2748774cdf8eeb99),U64(0x34b0bcb5e19b48a8),
+        U64(0x391c0cb3c5c95a63),U64(0x4ed8aa4ae3418acb),
+        U64(0x5b9cca4f7763e373),U64(0x682e6ff3d6b2b8a3),
+        U64(0x748f82ee5defb2fc),U64(0x78a5636f43172f60),
+        U64(0x84c87814a1f0ab72),U64(0x8cc702081a6439ec),
+        U64(0x90befffa23631e28),U64(0xa4506cebde82bde9),
+        U64(0xbef9a3f7b2c67915),U64(0xc67178f2e372532b),
+        U64(0xca273eceea26619c),U64(0xd186b8c721c0c207),
+        U64(0xeada7dd6cde0eb1e),U64(0xf57d4f7fee6ed178),
+        U64(0x06f067aa72176fba),U64(0x0a637dc5a2c898a6),
+        U64(0x113f9804bef90dae),U64(0x1b710b35131c471b),
+        U64(0x28db77f523047d84),U64(0x32caab7b40c72493),
+        U64(0x3c9ebe0a15c9bebc),U64(0x431d67c49c100d4c),
+        U64(0x4cc5d4becb3e42b6),U64(0x597f299cfc657e2a),
+        U64(0x5fcb6fab3ad6faec),U64(0x6c44198c4a475817) };
+
+#ifndef PEDANTIC
+# if defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
+#  if defined(__x86_64) || defined(__x86_64__)
+#   define PULL64(x) ({ SHA_LONG64 ret=*((const SHA_LONG64 *)(&(x)));	\
+				asm ("bswapq	%0"		\
+				: "=r"(ret)			\
+				: "0"(ret)); ret;		})
+#  endif
+# endif
+#endif
+
+#ifndef PULL64
+#define B(x,j)    (((SHA_LONG64)(*(((const unsigned char *)(&x))+j)))<<((7-j)*8))
+#define PULL64(x) (B(x,0)|B(x,1)|B(x,2)|B(x,3)|B(x,4)|B(x,5)|B(x,6)|B(x,7))
+#endif
+
+#ifndef PEDANTIC
+# if defined(_MSC_VER)
+#  if defined(_WIN64)	/* applies to both IA-64 and AMD64 */
+#   define ROTR(a,n)	_rotr64((a),n)
+#  endif
+# elif defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
+#  if defined(__x86_64) || defined(__x86_64__)
+#   define ROTR(a,n)	({ unsigned long ret;		\
+				asm ("rorq %1,%0"	\
+				: "=r"(ret)		\
+				: "J"(n),"0"(a)		\
+				: "cc"); ret;		})
+#  elif defined(_ARCH_PPC) && defined(__64BIT__)
+#   define ROTR(a,n)	({ unsigned long ret;		\
+				asm ("rotrdi %0,%1,%2"	\
+				: "=r"(ret)		\
+				: "r"(a),"K"(n)); ret;	})
+#  endif
+# endif
+#endif
+
+#ifndef ROTR
+#define ROTR(x,s)	(((x)>>s) | (x)<<(64-s))
+#endif
+
+#define Sigma0(x)	(ROTR((x),28) ^ ROTR((x),34) ^ ROTR((x),39))
+#define Sigma1(x)	(ROTR((x),14) ^ ROTR((x),18) ^ ROTR((x),41))
+#define sigma0(x)	(ROTR((x),1)  ^ ROTR((x),8)  ^ ((x)>>7))
+#define sigma1(x)	(ROTR((x),19) ^ ROTR((x),61) ^ ((x)>>6))
+
+#define Ch(x,y,z)	(((x) & (y)) ^ ((~(x)) & (z)))
+#define Maj(x,y,z)	(((x) & (y)) ^ ((x) & (z)) ^ ((y) & (z)))
+
+#if defined(OPENSSL_IA32_SSE2) && !defined(OPENSSL_NO_ASM)
+#define	GO_FOR_SSE2(ctx,in,num)		do {		\
+	extern int	OPENSSL_ia32cap;		\
+	void		sha512_block_sse2(void *,const void *,size_t);	\
+	if (!(OPENSSL_ia32cap & (1<<26))) break;	\
+	sha512_block_sse2(ctx->h,in,num); return;	\
+					} while (0)
+#endif
+
+#ifdef OPENSSL_SMALL_FOOTPRINT
+
+static void sha512_block (SHA512_CTX *ctx, const void *in, size_t num)
+	{
+	const SHA_LONG64 *W=in;
+	SHA_LONG64	a,b,c,d,e,f,g,h,s0,s1,T1,T2;
+	SHA_LONG64	X[16];
+	int i;
+
+#ifdef GO_FOR_SSE2
+	GO_FOR_SSE2(ctx,in,num);
+#endif
+
+			while (num--) {
+
+	a = ctx->h[0];	b = ctx->h[1];	c = ctx->h[2];	d = ctx->h[3];
+	e = ctx->h[4];	f = ctx->h[5];	g = ctx->h[6];	h = ctx->h[7];
+
+	for (i=0;i<16;i++)
+		{
+#ifdef B_ENDIAN
+		T1 = X[i] = W[i];
+#else
+		T1 = X[i] = PULL64(W[i]);
+#endif
+		T1 += h + Sigma1(e) + Ch(e,f,g) + K512[i];
+		T2 = Sigma0(a) + Maj(a,b,c);
+		h = g;	g = f;	f = e;	e = d + T1;
+		d = c;	c = b;	b = a;	a = T1 + T2;
+		}
+
+	for (;i<80;i++)
+		{
+		s0 = X[(i+1)&0x0f];	s0 = sigma0(s0);
+		s1 = X[(i+14)&0x0f];	s1 = sigma1(s1);
+
+		T1 = X[i&0xf] += s0 + s1 + X[(i+9)&0xf];
+		T1 += h + Sigma1(e) + Ch(e,f,g) + K512[i];
+		T2 = Sigma0(a) + Maj(a,b,c);
+		h = g;	g = f;	f = e;	e = d + T1;
+		d = c;	c = b;	b = a;	a = T1 + T2;
+		}
+
+	ctx->h[0] += a;	ctx->h[1] += b;	ctx->h[2] += c;	ctx->h[3] += d;
+	ctx->h[4] += e;	ctx->h[5] += f;	ctx->h[6] += g;	ctx->h[7] += h;
+
+			W+=SHA_LBLOCK;
+			}
+	}
+
+#else
+
+#define	ROUND_00_15(i,a,b,c,d,e,f,g,h)		do {	\
+	T1 += h + Sigma1(e) + Ch(e,f,g) + K512[i];	\
+	h = Sigma0(a) + Maj(a,b,c);			\
+	d += T1;	h += T1;		} while (0)
+
+#define	ROUND_16_80(i,a,b,c,d,e,f,g,h,X)	do {	\
+	s0 = X[(i+1)&0x0f];	s0 = sigma0(s0);	\
+	s1 = X[(i+14)&0x0f];	s1 = sigma1(s1);	\
+	T1 = X[(i)&0x0f] += s0 + s1 + X[(i+9)&0x0f];	\
+	ROUND_00_15(i,a,b,c,d,e,f,g,h);		} while (0)
+
+static void sha512_block (SHA512_CTX *ctx, const void *in, size_t num)
+	{
+	const SHA_LONG64 *W=in;
+	SHA_LONG64	a,b,c,d,e,f,g,h,s0,s1,T1;
+	SHA_LONG64	X[16];
+	int i;
+
+#ifdef GO_FOR_SSE2
+	GO_FOR_SSE2(ctx,in,num);
+#endif
+
+			while (num--) {
+
+	a = ctx->h[0];	b = ctx->h[1];	c = ctx->h[2];	d = ctx->h[3];
+	e = ctx->h[4];	f = ctx->h[5];	g = ctx->h[6];	h = ctx->h[7];
+
+#ifdef B_ENDIAN
+	T1 = X[0] = W[0];	ROUND_00_15(0,a,b,c,d,e,f,g,h);
+	T1 = X[1] = W[1];	ROUND_00_15(1,h,a,b,c,d,e,f,g);
+	T1 = X[2] = W[2];	ROUND_00_15(2,g,h,a,b,c,d,e,f);
+	T1 = X[3] = W[3];	ROUND_00_15(3,f,g,h,a,b,c,d,e);
+	T1 = X[4] = W[4];	ROUND_00_15(4,e,f,g,h,a,b,c,d);
+	T1 = X[5] = W[5];	ROUND_00_15(5,d,e,f,g,h,a,b,c);
+	T1 = X[6] = W[6];	ROUND_00_15(6,c,d,e,f,g,h,a,b);
+	T1 = X[7] = W[7];	ROUND_00_15(7,b,c,d,e,f,g,h,a);
+	T1 = X[8] = W[8];	ROUND_00_15(8,a,b,c,d,e,f,g,h);
+	T1 = X[9] = W[9];	ROUND_00_15(9,h,a,b,c,d,e,f,g);
+	T1 = X[10] = W[10];	ROUND_00_15(10,g,h,a,b,c,d,e,f);
+	T1 = X[11] = W[11];	ROUND_00_15(11,f,g,h,a,b,c,d,e);
+	T1 = X[12] = W[12];	ROUND_00_15(12,e,f,g,h,a,b,c,d);
+	T1 = X[13] = W[13];	ROUND_00_15(13,d,e,f,g,h,a,b,c);
+	T1 = X[14] = W[14];	ROUND_00_15(14,c,d,e,f,g,h,a,b);
+	T1 = X[15] = W[15];	ROUND_00_15(15,b,c,d,e,f,g,h,a);
+#else
+	T1 = X[0]  = PULL64(W[0]);	ROUND_00_15(0,a,b,c,d,e,f,g,h);
+	T1 = X[1]  = PULL64(W[1]);	ROUND_00_15(1,h,a,b,c,d,e,f,g);
+	T1 = X[2]  = PULL64(W[2]);	ROUND_00_15(2,g,h,a,b,c,d,e,f);
+	T1 = X[3]  = PULL64(W[3]);	ROUND_00_15(3,f,g,h,a,b,c,d,e);
+	T1 = X[4]  = PULL64(W[4]);	ROUND_00_15(4,e,f,g,h,a,b,c,d);
+	T1 = X[5]  = PULL64(W[5]);	ROUND_00_15(5,d,e,f,g,h,a,b,c);
+	T1 = X[6]  = PULL64(W[6]);	ROUND_00_15(6,c,d,e,f,g,h,a,b);
+	T1 = X[7]  = PULL64(W[7]);	ROUND_00_15(7,b,c,d,e,f,g,h,a);
+	T1 = X[8]  = PULL64(W[8]);	ROUND_00_15(8,a,b,c,d,e,f,g,h);
+	T1 = X[9]  = PULL64(W[9]);	ROUND_00_15(9,h,a,b,c,d,e,f,g);
+	T1 = X[10] = PULL64(W[10]);	ROUND_00_15(10,g,h,a,b,c,d,e,f);
+	T1 = X[11] = PULL64(W[11]);	ROUND_00_15(11,f,g,h,a,b,c,d,e);
+	T1 = X[12] = PULL64(W[12]);	ROUND_00_15(12,e,f,g,h,a,b,c,d);
+	T1 = X[13] = PULL64(W[13]);	ROUND_00_15(13,d,e,f,g,h,a,b,c);
+	T1 = X[14] = PULL64(W[14]);	ROUND_00_15(14,c,d,e,f,g,h,a,b);
+	T1 = X[15] = PULL64(W[15]);	ROUND_00_15(15,b,c,d,e,f,g,h,a);
+#endif
+
+	for (i=16;i<80;i+=8)
+		{
+		ROUND_16_80(i+0,a,b,c,d,e,f,g,h,X);
+		ROUND_16_80(i+1,h,a,b,c,d,e,f,g,X);
+		ROUND_16_80(i+2,g,h,a,b,c,d,e,f,X);
+		ROUND_16_80(i+3,f,g,h,a,b,c,d,e,X);
+		ROUND_16_80(i+4,e,f,g,h,a,b,c,d,X);
+		ROUND_16_80(i+5,d,e,f,g,h,a,b,c,X);
+		ROUND_16_80(i+6,c,d,e,f,g,h,a,b,X);
+		ROUND_16_80(i+7,b,c,d,e,f,g,h,a,X);
+		}
+
+	ctx->h[0] += a;	ctx->h[1] += b;	ctx->h[2] += c;	ctx->h[3] += d;
+	ctx->h[4] += e;	ctx->h[5] += f;	ctx->h[6] += g;	ctx->h[7] += h;
+
+			W+=SHA_LBLOCK;
+			}
+	}
+
+#endif
diff --git a/crypto/sha/sha512t.c b/crypto/sha/sha512t.c
new file mode 100644
index 0000000000..bb93070e11
--- /dev/null
+++ b/crypto/sha/sha512t.c
@@ -0,0 +1,168 @@
+/* crypto/sha/sha512t.c */
+/* ====================================================================
+ * Copyright (c) 2004 The OpenSSL Project.  All rights reserved.
+ * ====================================================================
+ */
+#include <stdio.h>
+#include <string.h>
+#include <stdlib.h>
+
+#include <openssl/sha.h>
+
+unsigned char app_c1[SHA512_DIGEST_LENGTH] = {
+	0xdd,0xaf,0x35,0xa1,0x93,0x61,0x7a,0xba,
+	0xcc,0x41,0x73,0x49,0xae,0x20,0x41,0x31,
+	0x12,0xe6,0xfa,0x4e,0x89,0xa9,0x7e,0xa2,
+	0x0a,0x9e,0xee,0xe6,0x4b,0x55,0xd3,0x9a,
+	0x21,0x92,0x99,0x2a,0x27,0x4f,0xc1,0xa8,
+	0x36,0xba,0x3c,0x23,0xa3,0xfe,0xeb,0xbd,
+	0x45,0x4d,0x44,0x23,0x64,0x3c,0xe8,0x0e,
+	0x2a,0x9a,0xc9,0x4f,0xa5,0x4c,0xa4,0x9f };
+
+unsigned char app_c2[SHA512_DIGEST_LENGTH] = {
+	0x8e,0x95,0x9b,0x75,0xda,0xe3,0x13,0xda,
+	0x8c,0xf4,0xf7,0x28,0x14,0xfc,0x14,0x3f,
+	0x8f,0x77,0x79,0xc6,0xeb,0x9f,0x7f,0xa1,
+	0x72,0x99,0xae,0xad,0xb6,0x88,0x90,0x18,
+	0x50,0x1d,0x28,0x9e,0x49,0x00,0xf7,0xe4,
+	0x33,0x1b,0x99,0xde,0xc4,0xb5,0x43,0x3a,
+	0xc7,0xd3,0x29,0xee,0xb6,0xdd,0x26,0x54,
+	0x5e,0x96,0xe5,0x5b,0x87,0x4b,0xe9,0x09 };
+
+unsigned char app_c3[SHA512_DIGEST_LENGTH] = {
+	0xe7,0x18,0x48,0x3d,0x0c,0xe7,0x69,0x64,
+	0x4e,0x2e,0x42,0xc7,0xbc,0x15,0xb4,0x63,
+	0x8e,0x1f,0x98,0xb1,0x3b,0x20,0x44,0x28,
+	0x56,0x32,0xa8,0x03,0xaf,0xa9,0x73,0xeb,
+	0xde,0x0f,0xf2,0x44,0x87,0x7e,0xa6,0x0a,
+	0x4c,0xb0,0x43,0x2c,0xe5,0x77,0xc3,0x1b,
+	0xeb,0x00,0x9c,0x5c,0x2c,0x49,0xaa,0x2e,
+	0x4e,0xad,0xb2,0x17,0xad,0x8c,0xc0,0x9b };
+
+unsigned char app_d1[SHA384_DIGEST_LENGTH] = {
+	0xcb,0x00,0x75,0x3f,0x45,0xa3,0x5e,0x8b,
+	0xb5,0xa0,0x3d,0x69,0x9a,0xc6,0x50,0x07,
+	0x27,0x2c,0x32,0xab,0x0e,0xde,0xd1,0x63,
+	0x1a,0x8b,0x60,0x5a,0x43,0xff,0x5b,0xed,
+	0x80,0x86,0x07,0x2b,0xa1,0xe7,0xcc,0x23,
+	0x58,0xba,0xec,0xa1,0x34,0xc8,0x25,0xa7 };
+
+unsigned char app_d2[SHA384_DIGEST_LENGTH] = {
+	0x09,0x33,0x0c,0x33,0xf7,0x11,0x47,0xe8,
+	0x3d,0x19,0x2f,0xc7,0x82,0xcd,0x1b,0x47,
+	0x53,0x11,0x1b,0x17,0x3b,0x3b,0x05,0xd2,
+	0x2f,0xa0,0x80,0x86,0xe3,0xb0,0xf7,0x12,
+	0xfc,0xc7,0xc7,0x1a,0x55,0x7e,0x2d,0xb9,
+	0x66,0xc3,0xe9,0xfa,0x91,0x74,0x60,0x39 };
+
+unsigned char app_d3[SHA384_DIGEST_LENGTH] = {
+	0x9d,0x0e,0x18,0x09,0x71,0x64,0x74,0xcb,
+	0x08,0x6e,0x83,0x4e,0x31,0x0a,0x4a,0x1c,
+	0xed,0x14,0x9e,0x9c,0x00,0xf2,0x48,0x52,
+	0x79,0x72,0xce,0xc5,0x70,0x4c,0x2a,0x5b,
+	0x07,0xb8,0xb3,0xdc,0x38,0xec,0xc4,0xeb,
+	0xae,0x97,0xdd,0xd8,0x7f,0x3d,0x89,0x85 };
+
+int main ()
+{ unsigned char md[SHA512_DIGEST_LENGTH];
+  int		i;
+  SHA512_CTX	ctx;
+
+#ifdef OPENSSL_IA32_SSE2
+    { extern int OPENSSL_ia32cap;
+      char      *env;
+
+	if (env=getenv("OPENSSL_ia32cap"))
+	    OPENSSL_ia32cap = strtol (env,NULL,0);
+    }
+#endif
+
+    fprintf(stdout,"Testing SHA-512 ");
+
+    SHA512((unsigned char *)"abc",3,md);
+    if (memcmp(md,app_c1,sizeof(app_c1)))
+    {	fflush(stdout);
+	fprintf(stderr,"\nTEST 1 of 3 failed.\n");
+	return 1;
+    }
+    else
+	fprintf(stdout,"."); fflush(stdout);
+
+    SHA512((unsigned char *)"abcdefgh""bcdefghi""cdefghij""defghijk"
+	   "efghijkl""fghijklm""ghijklmn""hijklmno"
+	   "ijklmnop""jklmnopq""klmnopqr""lmnopqrs"
+	   "mnopqrst""nopqrstu",112,md);
+    if (memcmp(md,app_c2,sizeof(app_c2)))
+    {	fflush(stdout);
+	fprintf(stderr,"\nTEST 2 of 3 failed.\n");
+	return 1;
+    }
+    else
+	fprintf(stdout,"."); fflush(stdout);
+
+    SHA512_Init(&ctx);
+    for (i=0;i<1000000;i+=288)
+	SHA512_Update(&ctx, "aaaaaaaa""aaaaaaaa""aaaaaaaa""aaaaaaaa"
+			    "aaaaaaaa""aaaaaaaa""aaaaaaaa""aaaaaaaa"
+			    "aaaaaaaa""aaaaaaaa""aaaaaaaa""aaaaaaaa"
+			    "aaaaaaaa""aaaaaaaa""aaaaaaaa""aaaaaaaa"
+			    "aaaaaaaa""aaaaaaaa""aaaaaaaa""aaaaaaaa"
+			    "aaaaaaaa""aaaaaaaa""aaaaaaaa""aaaaaaaa"
+			    "aaaaaaaa""aaaaaaaa""aaaaaaaa""aaaaaaaa"
+			    "aaaaaaaa""aaaaaaaa""aaaaaaaa""aaaaaaaa"
+			    "aaaaaaaa""aaaaaaaa""aaaaaaaa""aaaaaaaa",
+			    (1000000-i)<288?1000000-i:288);
+    SHA512_Final(md,&ctx);
+
+    if (memcmp(md,app_c3,sizeof(app_c3)))
+    {	fflush(stdout);
+	fprintf(stderr,"\nTEST 3 of 3 failed.\n");
+	return 1;
+    }
+    else
+	fprintf(stdout,"."); fflush(stdout);
+
+    fprintf(stdout," passed.\n"); fflush(stdout);
+
+    fprintf(stdout,"Testing SHA-384 ");
+
+    SHA384((unsigned char *)"abc",3,md);
+    if (memcmp(md,app_d1,sizeof(app_d1)))
+    {	fflush(stdout);
+	fprintf(stderr,"\nTEST 1 of 3 failed.\n");
+	return 1;
+    }
+    else
+	fprintf(stdout,"."); fflush(stdout);
+
+    SHA384((unsigned char *)"abcdefgh""bcdefghi""cdefghij""defghijk"
+	   "efghijkl""fghijklm""ghijklmn""hijklmno"
+	   "ijklmnop""jklmnopq""klmnopqr""lmnopqrs"
+	   "mnopqrst""nopqrstu",112,md);
+    if (memcmp(md,app_d2,sizeof(app_d2)))
+    {	fflush(stdout);
+	fprintf(stderr,"\nTEST 2 of 3 failed.\n");
+	return 1;
+    }
+    else
+	fprintf(stdout,"."); fflush(stdout);
+
+    SHA384_Init(&ctx);
+    for (i=0;i<1000000;i+=64)
+	SHA512_Update(&ctx, "aaaaaaaa""aaaaaaaa""aaaaaaaa""aaaaaaaa"
+			    "aaaaaaaa""aaaaaaaa""aaaaaaaa""aaaaaaaa",
+			    (1000000-i)<64?1000000-i:64);
+    SHA384_Final(md,&ctx);
+
+    if (memcmp(md,app_d3,sizeof(app_d3)))
+    {	fflush(stdout);
+	fprintf(stderr,"\nTEST 3 of 3 failed.\n");
+	return 1;
+    }
+    else
+	fprintf(stdout,"."); fflush(stdout);
+
+    fprintf(stdout," passed.\n"); fflush(stdout);
+
+  return 0;
+}
diff --git a/doc/crypto/OPENSSL_Applink.pod b/doc/crypto/OPENSSL_Applink.pod
new file mode 100644
index 0000000000..f15fb695eb
--- /dev/null
+++ b/doc/crypto/OPENSSL_Applink.pod
@@ -0,0 +1,21 @@
+=pod
+
+=head1 NAME
+
+OPENSSL_Applink
+
+=head1 SYNOPSIS
+
+ __declspec(dllexport) void **OPENSSL_Applink();
+
+=head1 DESCRIPTION
+
+OPENSSL_Applink is application-side interface which provides a glue
+between OpenSSL BIO layer and Win32 compiler run-time environment.
+Even though it appears at application side, it's essentially OpenSSL
+private interface. For this reason application developers are not
+expected to implement it, but to compile provided module with
+compiler of their choice and link it into the target application.
+The referred module is available as <openssl>/ms/applink.c.
+
+=cut
diff --git a/ms/applink.c b/ms/applink.c
new file mode 100644
index 0000000000..4333d2639d
--- /dev/null
+++ b/ms/applink.c
@@ -0,0 +1,45 @@
+#define APPLINK_STDIN	1
+#define APPLINK_STDOUT	2
+#define APPLINK_STDERR	3
+#define APPLINK_FPRINTF	4
+#define APPLINK_FGETS	5
+#define APPLINK_FREAD	6
+#define APPLINK_FWRITE	7
+#define APPLINK_FSETMOD	8
+#define APPLINK_FEOF	9
+#define APPLINK_FCLOSE 	10	/* should not be used */
+#define APPLINK_MAX	10	/* always same as last macro */
+
+#ifndef APPMACROS_ONLY
+#include <stdio.h>
+#include <io.h>
+#include <fcntl.h>
+
+static void *app_stdin()	{ return stdin;  }
+static void *app_stdout()	{ return stdout; }
+static void *app_stderr()	{ return stderr; }
+static int   app_feof(FILE *fp)	{ return feof(fp); }
+static int   app_fsetmod(FILE *fp,char mod)
+{ return _setmode (_fileno(fp),mod=='b'?_O_BINARY:_O_TEXT); }
+
+__declspec(dllexport) void **OPENSSL_Applink()
+{ static int once=1;
+  static void *OPENSSL_ApplinkTable[APPLINK_MAX+1]={(void *)APPLINK_MAX};
+
+    if (once)
+    {	OPENSSL_ApplinkTable[APPLINK_STDIN]	= app_stdin;
+	OPENSSL_ApplinkTable[APPLINK_STDOUT]	= app_stdout;
+	OPENSSL_ApplinkTable[APPLINK_STDERR]	= app_stderr;
+	OPENSSL_ApplinkTable[APPLINK_FPRINTF]	= fprintf;
+	OPENSSL_ApplinkTable[APPLINK_FGETS]	= fgets;
+	OPENSSL_ApplinkTable[APPLINK_FREAD]	= fread;
+	OPENSSL_ApplinkTable[APPLINK_FWRITE]	= fwrite;
+	OPENSSL_ApplinkTable[APPLINK_FSETMOD]	= app_fsetmod;
+	OPENSSL_ApplinkTable[APPLINK_FEOF]	= app_feof;
+	OPENSSL_ApplinkTable[APPLINK_FCLOSE]	= fclose;
+	once = 0;
+    }
+
+  return OPENSSL_ApplinkTable;
+}
+#endif
diff --git a/ms/uplink.c b/ms/uplink.c
new file mode 100644
index 0000000000..c839f9b087
--- /dev/null
+++ b/ms/uplink.c
@@ -0,0 +1,168 @@
+#if defined(_WIN64) && !defined(UNICODE)
+#define UNICODE
+#endif
+#if defined(UNICODE) && !defined(_UNICODE)
+#define _UNICODE
+#endif
+#if defined(_UNICODE) && !defined(UNICODE)
+#define UNICODE
+#endif
+#if defined(_MSC_VER) && !defined(_WIN32_WINNT)
+#define _WIN32_WINNT 0x0333	/* 3.51 */
+#endif
+
+#include <windows.h>
+#include <tchar.h>
+#include <stdio.h>
+#include <malloc.h>
+#include "uplink.h"
+
+#ifdef _MSC_VER
+#pragma comment(lib,"delayimp")
+/*
+ * CL command line should also be complemented with following:
+ *
+ *	/link /delayload:advapi32.dll /delayload:user32.dll
+ *
+ * This is required if/as we want to support Win9x. With delayloaded
+ * DLLs in question all we have to do is to make sure NT-specific
+ * functions are not actually called under Win9x.
+ */
+#endif
+
+#if defined(_WIN32_WINNT) && _WIN32_WINNT>=0x0333
+int IsService()
+{ HWINSTA h;
+  DWORD len;
+  WCHAR *name;
+
+    GetDesktopWindow(); /* return value is ignored */
+
+    h = GetProcessWindowStation();
+    if (h==NULL) return -1;
+
+    if (GetUserObjectInformationW (h,UOI_NAME,NULL,0,&len) ||
+	GetLastError() != ERROR_INSUFFICIENT_BUFFER)
+	return -1;
+
+    if (len>512) return -1;		/* paranoia */
+    len++,len&=~1;			/* paranoia */
+#ifdef _MSC_VER
+    name=(WCHAR *)_alloca(len+sizeof(WCHAR));
+#else
+    name=(WCHAR *)alloca(len+sizeof(WCHAR));
+#endif
+    if (!GetUserObjectInformationW (h,UOI_NAME,name,len,&len))
+	return -1;
+
+    len++,len&=~1;			/* paranoia */
+    name[len/sizeof(WCHAR)]=L'\0';	/* paranoia */
+#if 1
+    /* This doesn't cover "interactive" services [working with real
+     * WinSta0's] nor programs started non-interactively by Task
+     * Scheduler [those are working with SAWinSta]. */
+    if (wcsstr(name,L"Service-0x"))	return 1;
+#else
+    /* This covers all non-interactive programs such as services. */
+    if (!wcsstr(name,L"WinSta0"))	return 1;
+#endif
+    else				return 0;
+}
+#endif
+
+static TCHAR msg[128];
+
+static void unimplemented ()
+{
+#if defined(_WIN32_WINNT) && _WIN32_WINNT>=0x0333
+    /* this -------------v--- guards NT-specific calls */
+    if (GetVersion() < 0x80000000 && IsService())
+    {	HANDLE h = RegisterEventSource(0,_T("OPENSSL"));
+	TCHAR *pmsg=msg;
+	ReportEvent(h,EVENTLOG_ERROR_TYPE,0,0,0,1,0,&pmsg,0);
+	DeregisterEventSource(h);
+    }
+    else
+#endif
+    {	MSGBOXPARAMS         m;
+
+	m.cbSize             = sizeof(m);
+	m.hwndOwner          = NULL;
+	m.lpszCaption        = _T("OpenSSL: FATAL");
+	m.dwStyle            = MB_OK;
+	m.hInstance          = NULL;
+	m.lpszIcon           = IDI_ERROR;
+	m.dwContextHelpId    = 0;
+	m.lpfnMsgBoxCallback = NULL;
+	m.dwLanguageId       = MAKELANGID(LANG_ENGLISH,SUBLANG_ENGLISH_US);
+	m.lpszText           = msg;
+
+	MessageBoxIndirect (&m);
+    }
+    ExitProcess (1);
+}
+
+void OPENSSL_Uplink (void **table, int index)
+{ static HMODULE app=NULL;
+  static void **applinktable=NULL;
+  int len;
+
+    len = _stprintf (msg,_T("OPENSSL_Uplink(%p,%02X): "),table,index);
+    _tcscpy (msg+len,_T("unimplemented function"));
+    table [index] = unimplemented;
+
+    if (app==NULL && (app=GetModuleHandle(NULL))==NULL)
+    {	app=(HMODULE)-1; _tcscpy (msg+len,_T("no host application"));
+	return;
+    }
+    else if (app==(HMODULE)-1)	{ return; }
+
+    if (applinktable==NULL)
+    { void**(*applink)();
+
+	applink=(void**(*)())GetProcAddress(app,"OPENSSL_Applink");
+	if (applink==NULL)
+	{   app=(HMODULE)-1; _tcscpy (msg+len,_T("no OPENSSL_Applink"));
+	    return;
+	}
+	applinktable = (*applink)();
+	if (applinktable==NULL)
+	{   app=(HMODULE)-1; _tcscpy (msg+len,_T("no ApplinkTable"));
+	    return;
+	}
+    }
+
+    if (index > (int)applinktable[0])	{ return; }
+
+    if (applinktable[index]) table[index] = applinktable[index];
+}    
+
+#if defined(_MSC_VER) && defined(_M_IX86)
+#define LAZY(i)		\
+__declspec(naked) static void lazy##i () { 	\
+	_asm	push i				\
+	_asm	push OFFSET OPENSSL_UplinkTable	\
+	_asm	call OPENSSL_Uplink		\
+	_asm	add  esp,8			\
+	_asm	jmp  OPENSSL_UplinkTable+4*i	}
+
+#if APPLINK_MAX>20
+#error "Add more stubs..."
+#endif
+/* make some in advance... */
+LAZY(1)  LAZY(2)  LAZY(3)  LAZY(4)  LAZY(5)
+LAZY(6)  LAZY(7)  LAZY(8)  LAZY(9)  LAZY(10)
+LAZY(11) LAZY(12) LAZY(13) LAZY(14) LAZY(15)
+LAZY(16) LAZY(17) LAZY(18) LAZY(19) LAZY(20)
+void *OPENSSL_UplinkTable[] = {
+	(void *)APPLINK_MAX,
+	lazy1, lazy2, lazy3, lazy4, lazy5,
+	lazy6, lazy7, lazy8, lazy9, lazy10,
+	lazy11,lazy12,lazy13,lazy14,lazy15,
+	lazy16,lazy17,lazy18,lazy19,lazy20,
+};
+#endif
+
+#ifdef SELFTEST
+main() {  UP_fprintf(UP_stdout,"hello, world!\n"); }
+#endif
diff --git a/ms/uplink.h b/ms/uplink.h
new file mode 100644
index 0000000000..3e9911ab93
--- /dev/null
+++ b/ms/uplink.h
@@ -0,0 +1,14 @@
+#define APPMACROS_ONLY
+#include "applink.c"
+
+extern void *OPENSSL_UplinkTable[];
+#define UP_stdin  (*(void *(*)())OPENSSL_UplinkTable[APPLINK_STDIN])()
+#define UP_stdout (*(void *(*)())OPENSSL_UplinkTable[APPLINK_STDOUT])()
+#define UP_stderr (*(void *(*)())OPENSSL_UplinkTable[APPLINK_STDERR])()
+#define UP_fprintf (*(int (*)(void *,const char *,...))OPENSSL_UplinkTable[APPLINK_FPRINTF])
+#define UP_fgets  (*(char *(*)(char *,int,void *))OPENSSL_UplinkTable[APPLINK_FGETS])
+#define UP_fread  (*(size_t (*)(void *,size_t,size_t,void *))OPENSSL_UplinkTable[APPLINK_FREAD])
+#define UP_fwrite (*(size_t (*)(void *,size_t,size_t,void *))OPENSSL_UplinkTable[APPLINK_FWRITE])
+#define UP_fsetmod (*(int (*)(void *,char))OPENSSL_UplinkTable[APPLINK_FSETMOD])
+#define UP_feof   (*(int (*)(void *))OPENSSL_UplinkTable[APPLINK_FEOF])
+#define UP_fclose (*(int (*)(void *))OPENSSL_Uplink[APPLINK_FCLOSE])
diff --git a/ms/uplink.pl b/ms/uplink.pl
new file mode 100755
index 0000000000..801f6e01fa
--- /dev/null
+++ b/ms/uplink.pl
@@ -0,0 +1,177 @@
+#!/usr/bin/env perl
+#
+# For Microsoft CL this is implemented as inline assembler. So that
+# even though this script can generate even Win32 code, we'll be
+# using it primarily to generate Win64 modules. Both IA-64 and AMD64
+# are supported...
+
+# pull APPLINK_MAX value from applink.c...
+$applink_c=$0;
+$applink_c=~s|[^/\\]+$||g;
+$applink_c.="applink.c";
+open(INPUT,$applink_c) || die "can't open $applink_c: $!";
+@max=grep {/APPLINK_MAX\s+(\d+)/} <INPUT>;
+close(INPUT);
+($#max==0) or die "can't find APPLINK_MAX in $applink_c";
+
+$max[0]=~/APPLINK_MAX\s+(\d+)/;
+$N=$1;	# number of entries in OPENSSL_UplinkTable not including
+	# OPENSSL_UplinkTable[0], which contains this value...
+
+# Idea is to fill the OPENSSL_UplinkTable with pointers to stubs
+# which invoke 'void OPENSSL_Uplink (ULONG_PTR *table,int index)';
+# and then dereference themselves. Latter shall result in endless
+# loop *unless* OPENSSL_Uplink does not replace 'table[index]' with
+# something else, e.g. as 'table[index]=unimplemented;'...
+
+$arg = shift;
+#( defined shift || open STDOUT,">$arg" ) || die "can't open $arg: $!";
+
+if ($arg =~ /win32n/)	{ ia32nasm();  }
+elsif ($arg =~ /win32/)	{ ia32masm();  }
+elsif ($arg =~ /ia64/)	{ ia64ias();   }
+elsif ($arg =~ /amd64/)	{ amd64masm(); }
+else	{ die "nonsense $arg"; }
+
+sub ia32masm() {
+print <<___;
+.386P
+.model	FLAT
+
+_DATA	SEGMENT
+PUBLIC	_OPENSSL_UplinkTable
+_OPENSSL_UplinkTable	DD	$N	; amount of following entries
+___
+for ($i=1;$i<=$N;$i++) {   print "	DD	FLAT:\$lazy$i\n";   }
+print <<___;
+_DATA	ENDS
+
+_TEXT	SEGMENT
+EXTRN	_OPENSSL_Uplink:NEAR
+___
+for ($i=1;$i<=$N;$i++) {
+print <<___;
+ALIGN	4
+\$lazy$i	PROC NEAR
+	push	$i
+	push	OFFSET FLAT:_OPENSSL_UplinkTable
+	call	_OPENSSL_Uplink
+	add	esp,8
+	jmp	DWORD PTR _OPENSSL_UplinkTable+4*$i
+\$lazy$i	ENDP
+___
+}
+print <<___;
+ALIGN	4
+_TEXT	ENDS
+END
+___
+}
+
+sub ia32nasm() {
+print <<___;
+SEGMENT	.data
+GLOBAL	_OPENSSL_UplinkTable
+_OPENSSL_UplinkTable	DD	$N	; amount of following entries
+___
+for ($i=1;$i<=$N;$i++) {   print "	DD	\$lazy$i\n";   }
+print <<___;
+
+SEGMENT	.text
+EXTERN	_OPENSSL_Uplink
+___
+for ($i=1;$i<=$N;$i++) {
+print <<___;
+ALIGN	4
+\$lazy$i:
+	push	$i
+	push	_OPENSSL_UplinkTable
+	call	_OPENSSL_Uplink
+	add	esp,8
+	jmp	[_OPENSSL_UplinkTable+4*$i]
+___
+}
+print <<___;
+ALIGN	4
+END
+___
+}
+
+sub ia64ias () {
+local $V=8;	# max number of args uplink functions may accept...
+print <<___;
+.data
+.global	OPENSSL_UplinkTable#
+OPENSSL_UplinkTable:	data8	$N	// amount of following entries
+___
+for ($i=1;$i<=$N;$i++) {   print "	data8	\@fptr(lazy$i#)\n";   }
+print <<___;
+.size	OPENSSL_UplinkTable,.-OPENSSL_UplinkTable#
+
+.text
+.global	OPENSSL_Uplink#
+.type	OPENSSL_Uplink#,\@function
+___
+for ($i=1;$i<=$N;$i++) {
+print <<___;
+.proc	lazy$i
+lazy$i:
+{ .mii;	alloc	loc0=ar.pfs,$V,3,2,0
+	mov	loc1=b0
+	addl	loc2=\@ltoff(OPENSSL_UplinkTable#),gp	};;
+{ .mmi;	ld8	out0=[loc2]
+	mov	out1=$i					};;
+{ .mib;	adds	loc2=8*$i,out0
+	br.call.sptk.many	b0=OPENSSL_Uplink#	};;
+{ .mmi;	ld8	r31=[loc2];;
+	ld8	r30=[r31],8				};;
+{ .mii;	ld8	gp=[r31]
+	mov	b6=r30
+	mov	b0=loc1					};;
+{ .mib; mov	ar.pfs=loc0
+	br.many	b6					};;
+.endp	lazy$i#
+___
+}
+}
+
+sub amd64masm() {
+print <<___;
+_DATA	SEGMENT
+PUBLIC	OPENSSL_UplinkTable
+OPENSSL_UplinkTable	DQ	$N
+___
+for ($i=1;$i<=$N;$i++) {   print "	DQ	FLAT:\$lazy$i\n";   }
+print <<___;
+_DATA	ENDS
+
+TEXT	SEGMENT
+EXTERN	OPENSSL_Uplink:NEAR
+___
+for ($i=1;$i<=$N;$i++) {
+print <<___;
+ALIGN	4
+\$lazy$i	PROC NEAR
+	push	r9
+	push	r8
+	push	rdx
+	push	rcx
+	sub	rsp,40
+	mov	rcx,OFFSET FLAT:OPENSSL_UplinkTable
+	mov	rdx,$i
+	call	OPENSSL_Uplink
+	add	rsp,40
+	pop	rcx
+	pop	rdx
+	pop	r8
+	pop	r9
+	jmp	QWORD PTR OPENSSL_UplinkTable+8*$i
+\$lazy$i	ENDP
+___
+}
+print <<___;
+TEXT	ENDS
+END
+___
+}
+