openssl/openssl-1.0.1e-ppc-asm-update.patch

diff --git a/Configure b/Configure
index 9c803dc..5a5c2d8 100755
--- a/Configure
+++ b/Configure
@@ -139,8 +139,8 @@ my $s390x_asm="s390xcap.o s390xcpuid.o:bn-s390x.o s390x-mont.o s390x-gf2m.o::aes
 my $armv4_asm="armcap.o armv4cpuid.o:bn_asm.o armv4-mont.o armv4-gf2m.o::aes_cbc.o aes-armv4.o:::sha1-armv4-large.o sha256-armv4.o sha512-armv4.o:::::::ghash-armv4.o::void";
 my $parisc11_asm="pariscid.o:bn_asm.o parisc-mont.o::aes_core.o aes_cbc.o aes-parisc.o:::sha1-parisc.o sha256-parisc.o sha512-parisc.o::rc4-parisc.o:::::ghash-parisc.o::32";
 my $parisc20_asm="pariscid.o:pa-risc2W.o parisc-mont.o::aes_core.o aes_cbc.o aes-parisc.o:::sha1-parisc.o sha256-parisc.o sha512-parisc.o::rc4-parisc.o:::::ghash-parisc.o::64";
-my $ppc32_asm="ppccpuid.o ppccap.o:bn-ppc.o ppc-mont.o ppc64-mont.o::aes_core.o aes_cbc.o aes-ppc.o:::sha1-ppc.o sha256-ppc.o::::::::";
-my $ppc64_asm="ppccpuid.o ppccap.o:bn-ppc.o ppc-mont.o ppc64-mont.o::aes_core.o aes_cbc.o aes-ppc.o:::sha1-ppc.o sha256-ppc.o sha512-ppc.o::::::::";
+my $ppc64_asm="ppccpuid.o ppccap.o:bn-ppc.o ppc-mont.o ppc64-mont.o::aes_core.o aes_cbc.o aes-ppc.o vpaes-ppc.o aesp8-ppc.o:::sha1-ppc.o sha256-ppc.o sha512-ppc.o sha256p8-ppc.o sha512p8-ppc.o:::::::ghashp8-ppc.o:";
+my $ppc32_asm=$ppc64_asm;
 my $no_asm=":::::::::::::::void";
 
 # As for $BSDthreads. Idea is to maintain "collective" set of flags,
@@ -357,6 +357,7 @@ my %table=(
 ####
 "linux-generic64","gcc:-DTERMIO -O3 -Wall::-D_REENTRANT::-ldl:SIXTY_FOUR_BIT_LONG RC4_CHAR RC4_CHUNK DES_INT DES_UNROLL BF_PTR:${no_asm}:dlfcn:linux-shared:-fPIC::.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
 "linux-ppc64",	"gcc:-m64 -DB_ENDIAN -DTERMIO -O3 -Wall::-D_REENTRANT::-ldl:SIXTY_FOUR_BIT_LONG RC4_CHAR RC4_CHUNK DES_RISC1 DES_UNROLL:${ppc64_asm}:linux64:dlfcn:linux-shared:-fPIC:-m64:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR):::64",
+"linux-ppc64le","gcc:-m64 -DL_ENDIAN -DTERMIO -O3 -Wall::-D_REENTRANT::-ldl:SIXTY_FOUR_BIT_LONG RC4_CHAR RC4_CHUNK DES_RISC1 DES_UNROLL:$ppc64_asm:linux64le:dlfcn:linux-shared:-fPIC:-m64:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR):::",
 "linux-ia64",	"gcc:-DL_ENDIAN -DTERMIO -O3 -Wall::-D_REENTRANT::-ldl:SIXTY_FOUR_BIT_LONG RC4_CHUNK DES_UNROLL DES_INT:${ia64_asm}:dlfcn:linux-shared:-fPIC::.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
 "linux-ia64-ecc","ecc:-DL_ENDIAN -DTERMIO -O2 -Wall -no_cpprt::-D_REENTRANT::-ldl:SIXTY_FOUR_BIT_LONG RC4_CHUNK DES_INT:${ia64_asm}:dlfcn:linux-shared:-fPIC::.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
 "linux-ia64-icc","icc:-DL_ENDIAN -DTERMIO -O2 -Wall -no_cpprt::-D_REENTRANT::-ldl:SIXTY_FOUR_BIT_LONG RC4_CHUNK DES_RISC1 DES_INT:${ia64_asm}:dlfcn:linux-shared:-fPIC::.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)",
@@ -462,8 +463,8 @@ my %table=(
 
 #### IBM's AIX.
 "aix3-cc",  "cc:-O -DB_ENDIAN -qmaxmem=16384::(unknown):AIX::BN_LLONG RC4_CHAR:::",
-"aix-gcc",  "gcc:-O -DB_ENDIAN::-pthread:AIX::BN_LLONG RC4_CHAR:${ppc32_asm}:aix32:dlfcn:aix-shared::-shared -Wl,-G:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)::-X32",
-"aix64-gcc","gcc:-maix64 -O -DB_ENDIAN::-pthread:AIX::SIXTY_FOUR_BIT_LONG RC4_CHAR:${ppc64_asm}:aix64:dlfcn:aix-shared::-maix64 -shared -Wl,-G:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)::-X64",
+"aix-gcc",  "gcc:-O -DB_ENDIAN::-pthread:AIX::BN_LLONG RC4_CHAR:$ppc32_asm:aix32:dlfcn:aix-shared::-shared -Wl,-G:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)::-X32",
+"aix64-gcc","gcc:-maix64 -O -DB_ENDIAN::-pthread:AIX::SIXTY_FOUR_BIT_LONG RC4_CHAR:$ppc64_asm:aix64:dlfcn:aix-shared::-maix64 -shared -Wl,-G:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)::-X64",
 # Below targets assume AIX 5. Idea is to effectively disregard $OBJECT_MODE
 # at build time. $OBJECT_MODE is respected at ./config stage!
 "aix-cc",   "cc:-q32 -O -DB_ENDIAN -qmaxmem=16384 -qro -qroconst::-qthreaded -D_THREAD_SAFE:AIX::BN_LLONG RC4_CHAR:${ppc32_asm}:aix32:dlfcn:aix-shared::-q32 -G:.so.\$(SHLIB_MAJOR).\$(SHLIB_MINOR)::-X 32",
@@ -1525,7 +1526,7 @@ else	{
 	$wp_obj="wp_block.o";
 	}
 $cmll_obj=$cmll_enc	unless ($cmll_obj =~ /.o$/);
-if ($modes_obj =~ /ghash/)
+if ($modes_obj =~ /ghash\-/)
 	{
 	$cflags.=" -DGHASH_ASM";
 	}
diff --git a/config b/config
index 88b9bc6..8b80802 100755
--- a/config
+++ b/config
@@ -587,13 +587,20 @@ case "$GUESSOS" in
 	fi
 	;;
   ppc64-*-linux2)
-	echo "WARNING! If you wish to build 64-bit library, then you have to"
-	echo "         invoke './Configure linux-ppc64' *manually*."
-	if [ "$TEST" = "false" -a -t 1 ]; then
-	    echo "         You have about 5 seconds to press Ctrl-C to abort."
-	    (trap "stty `stty -g`" 2 0; stty -icanon min 0 time 50; read waste) <&1
+	if [ -z "$KERNEL_BITS" ]; then
+	    echo "WARNING! If you wish to build 64-bit library, then you have to"
+	    echo "         invoke './Configure linux-ppc64' *manually*."
+	    if [ "$TEST" = "false" -a -t 1 ]; then
+		echo "         You have about 5 seconds to press Ctrl-C to abort."
+		(trap "stty `stty -g`" 2 0; stty -icanon min 0 time 50; read waste) <&1
+	    fi
+	fi
+	if [ "$KERNEL_BITS" = "64" ]; then
+	    OUT="linux-ppc64"
+	else
+	    OUT="linux-ppc"
+	    (echo "__LP64__" | gcc -E -x c - 2>/dev/null | grep "^__LP64__" 2>&1 > /dev/null) || options="$options -m32"
 	fi
-	OUT="linux-ppc"
 	;;
   ppc-*-linux2) OUT="linux-ppc" ;;
   ppc60x-*-vxworks*) OUT="vxworks-ppc60x" ;;
diff --git a/crypto/aes/Makefile b/crypto/aes/Makefile
index 45ede0a..847f4ee 100644
--- a/crypto/aes/Makefile
+++ b/crypto/aes/Makefile
@@ -71,6 +71,10 @@ aes-sparcv9.s: asm/aes-sparcv9.pl
 
 aes-ppc.s:	asm/aes-ppc.pl
 	$(PERL) asm/aes-ppc.pl $(PERLASM_SCHEME) $@
+vpaes-ppc.s:	asm/vpaes-ppc.pl
+	$(PERL) asm/vpaes-ppc.pl $(PERLASM_SCHEME) $@
+aesp8-ppc.s:	asm/aesp8-ppc.pl
+	$(PERL) asm/aesp8-ppc.pl $(PERLASM_SCHEME) $@
 
 aes-parisc.s:	asm/aes-parisc.pl
 	$(PERL) asm/aes-parisc.pl $(PERLASM_SCHEME) $@
diff --git a/crypto/aes/asm/aes-ppc.pl b/crypto/aes/asm/aes-ppc.pl
index 7c52cbe..7a99fc3 100644
--- a/crypto/aes/asm/aes-ppc.pl
+++ b/crypto/aes/asm/aes-ppc.pl
@@ -45,6 +45,8 @@ if ($flavour =~ /64/) {
 	$PUSH	="stw";
 } else { die "nonsense $flavour"; }
 
+$LITTLE_ENDIAN = ($flavour=~/le$/) ? $SIZE_T : 0;
+
 $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
 ( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or
 ( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or
@@ -68,7 +70,7 @@ $key="r5";
 $Tbl0="r3";
 $Tbl1="r6";
 $Tbl2="r7";
-$Tbl3="r2";
+$Tbl3=$out;	# stay away from "r2"; $out is offloaded to stack
 
 $s0="r8";
 $s1="r9";
@@ -76,7 +78,7 @@ $s2="r10";
 $s3="r11";
 
 $t0="r12";
-$t1="r13";
+$t1="r0";	# stay away from "r13";
 $t2="r14";
 $t3="r15";
 
@@ -100,9 +102,6 @@ $acc13="r29";
 $acc14="r30";
 $acc15="r31";
 
-# stay away from TLS pointer
-if ($SIZE_T==8)	{ die if ($t1 ne "r13");  $t1="r0";		}
-else		{ die if ($Tbl3 ne "r2"); $Tbl3=$t0; $t0="r0";	}
 $mask80=$Tbl2;
 $mask1b=$Tbl3;
 
@@ -337,8 +336,7 @@ $code.=<<___;
 	$STU	$sp,-$FRAME($sp)
 	mflr	r0
 
-	$PUSH	$toc,`$FRAME-$SIZE_T*20`($sp)
-	$PUSH	r13,`$FRAME-$SIZE_T*19`($sp)
+	$PUSH	$out,`$FRAME-$SIZE_T*19`($sp)
 	$PUSH	r14,`$FRAME-$SIZE_T*18`($sp)
 	$PUSH	r15,`$FRAME-$SIZE_T*17`($sp)
 	$PUSH	r16,`$FRAME-$SIZE_T*16`($sp)
@@ -365,16 +363,61 @@ $code.=<<___;
 	bne	Lenc_unaligned
 
 Lenc_unaligned_ok:
+___
+$code.=<<___ if (!$LITTLE_ENDIAN);
 	lwz	$s0,0($inp)
 	lwz	$s1,4($inp)
 	lwz	$s2,8($inp)
 	lwz	$s3,12($inp)
+___
+$code.=<<___ if ($LITTLE_ENDIAN);
+	lwz	$t0,0($inp)
+	lwz	$t1,4($inp)
+	lwz	$t2,8($inp)
+	lwz	$t3,12($inp)
+	rotlwi	$s0,$t0,8
+	rotlwi	$s1,$t1,8
+	rotlwi	$s2,$t2,8
+	rotlwi	$s3,$t3,8
+	rlwimi	$s0,$t0,24,0,7
+	rlwimi	$s1,$t1,24,0,7
+	rlwimi	$s2,$t2,24,0,7
+	rlwimi	$s3,$t3,24,0,7
+	rlwimi	$s0,$t0,24,16,23
+	rlwimi	$s1,$t1,24,16,23
+	rlwimi	$s2,$t2,24,16,23
+	rlwimi	$s3,$t3,24,16,23
+___
+$code.=<<___;
 	bl	LAES_Te
 	bl	Lppc_AES_encrypt_compact
+	$POP	$out,`$FRAME-$SIZE_T*19`($sp)
+___
+$code.=<<___ if ($LITTLE_ENDIAN);
+	rotlwi	$t0,$s0,8
+	rotlwi	$t1,$s1,8
+	rotlwi	$t2,$s2,8
+	rotlwi	$t3,$s3,8
+	rlwimi	$t0,$s0,24,0,7
+	rlwimi	$t1,$s1,24,0,7
+	rlwimi	$t2,$s2,24,0,7
+	rlwimi	$t3,$s3,24,0,7
+	rlwimi	$t0,$s0,24,16,23
+	rlwimi	$t1,$s1,24,16,23
+	rlwimi	$t2,$s2,24,16,23
+	rlwimi	$t3,$s3,24,16,23
+	stw	$t0,0($out)
+	stw	$t1,4($out)
+	stw	$t2,8($out)
+	stw	$t3,12($out)
+___
+$code.=<<___ if (!$LITTLE_ENDIAN);
 	stw	$s0,0($out)
 	stw	$s1,4($out)
 	stw	$s2,8($out)
 	stw	$s3,12($out)
+___
+$code.=<<___;
 	b	Lenc_done
 
 Lenc_unaligned:
@@ -417,6 +460,7 @@ Lenc_xpage:
 
 	bl	LAES_Te
 	bl	Lppc_AES_encrypt_compact
+	$POP	$out,`$FRAME-$SIZE_T*19`($sp)
 
 	extrwi	$acc00,$s0,8,0
 	extrwi	$acc01,$s0,8,8
@@ -449,8 +493,6 @@ Lenc_xpage:
 
 Lenc_done:
 	$POP	r0,`$FRAME+$LRSAVE`($sp)
-	$POP	$toc,`$FRAME-$SIZE_T*20`($sp)
-	$POP	r13,`$FRAME-$SIZE_T*19`($sp)
 	$POP	r14,`$FRAME-$SIZE_T*18`($sp)
 	$POP	r15,`$FRAME-$SIZE_T*17`($sp)
 	$POP	r16,`$FRAME-$SIZE_T*16`($sp)
@@ -764,6 +806,7 @@ Lenc_compact_done:
 	blr
 	.long	0
 	.byte	0,12,0x14,0,0,0,0,0
+.size	.AES_encrypt,.-.AES_encrypt
 
 .globl	.AES_decrypt
 .align	7
@@ -771,8 +814,7 @@ Lenc_compact_done:
 	$STU	$sp,-$FRAME($sp)
 	mflr	r0
 
-	$PUSH	$toc,`$FRAME-$SIZE_T*20`($sp)
-	$PUSH	r13,`$FRAME-$SIZE_T*19`($sp)
+	$PUSH	$out,`$FRAME-$SIZE_T*19`($sp)
 	$PUSH	r14,`$FRAME-$SIZE_T*18`($sp)
 	$PUSH	r15,`$FRAME-$SIZE_T*17`($sp)
 	$PUSH	r16,`$FRAME-$SIZE_T*16`($sp)
@@ -799,16 +841,61 @@ Lenc_compact_done:
 	bne	Ldec_unaligned
 
 Ldec_unaligned_ok:
+___
+$code.=<<___ if (!$LITTLE_ENDIAN);
 	lwz	$s0,0($inp)
 	lwz	$s1,4($inp)
 	lwz	$s2,8($inp)
 	lwz	$s3,12($inp)
+___
+$code.=<<___ if ($LITTLE_ENDIAN);
+	lwz	$t0,0($inp)
+	lwz	$t1,4($inp)
+	lwz	$t2,8($inp)
+	lwz	$t3,12($inp)
+	rotlwi	$s0,$t0,8
+	rotlwi	$s1,$t1,8
+	rotlwi	$s2,$t2,8
+	rotlwi	$s3,$t3,8
+	rlwimi	$s0,$t0,24,0,7
+	rlwimi	$s1,$t1,24,0,7
+	rlwimi	$s2,$t2,24,0,7
+	rlwimi	$s3,$t3,24,0,7
+	rlwimi	$s0,$t0,24,16,23
+	rlwimi	$s1,$t1,24,16,23
+	rlwimi	$s2,$t2,24,16,23
+	rlwimi	$s3,$t3,24,16,23
+___
+$code.=<<___;
 	bl	LAES_Td
 	bl	Lppc_AES_decrypt_compact
+	$POP	$out,`$FRAME-$SIZE_T*19`($sp)
+___
+$code.=<<___ if ($LITTLE_ENDIAN);
+	rotlwi	$t0,$s0,8
+	rotlwi	$t1,$s1,8
+	rotlwi	$t2,$s2,8
+	rotlwi	$t3,$s3,8
+	rlwimi	$t0,$s0,24,0,7
+	rlwimi	$t1,$s1,24,0,7
+	rlwimi	$t2,$s2,24,0,7
+	rlwimi	$t3,$s3,24,0,7
+	rlwimi	$t0,$s0,24,16,23
+	rlwimi	$t1,$s1,24,16,23
+	rlwimi	$t2,$s2,24,16,23
+	rlwimi	$t3,$s3,24,16,23
+	stw	$t0,0($out)
+	stw	$t1,4($out)
+	stw	$t2,8($out)
+	stw	$t3,12($out)
+___
+$code.=<<___ if (!$LITTLE_ENDIAN);
 	stw	$s0,0($out)
 	stw	$s1,4($out)
 	stw	$s2,8($out)
 	stw	$s3,12($out)
+___
+$code.=<<___;
 	b	Ldec_done
 
 Ldec_unaligned:
@@ -851,6 +938,7 @@ Ldec_xpage:
 
 	bl	LAES_Td
 	bl	Lppc_AES_decrypt_compact
+	$POP	$out,`$FRAME-$SIZE_T*19`($sp)
 
 	extrwi	$acc00,$s0,8,0
 	extrwi	$acc01,$s0,8,8
@@ -883,8 +971,6 @@ Ldec_xpage:
 
 Ldec_done:
 	$POP	r0,`$FRAME+$LRSAVE`($sp)
-	$POP	$toc,`$FRAME-$SIZE_T*20`($sp)
-	$POP	r13,`$FRAME-$SIZE_T*19`($sp)
 	$POP	r14,`$FRAME-$SIZE_T*18`($sp)
 	$POP	r15,`$FRAME-$SIZE_T*17`($sp)
 	$POP	r16,`$FRAME-$SIZE_T*16`($sp)
@@ -1355,6 +1441,7 @@ Ldec_compact_done:
 	blr
 	.long	0
 	.byte	0,12,0x14,0,0,0,0,0
+.size	.AES_decrypt,.-.AES_decrypt
 
 .asciz	"AES for PPC, CRYPTOGAMS by <appro\@openssl.org>"
 .align	7
diff --git a/crypto/aes/asm/aesp8-ppc.pl b/crypto/aes/asm/aesp8-ppc.pl
new file mode 100755
index 0000000..3ee8979
--- /dev/null
+++ b/crypto/aes/asm/aesp8-ppc.pl
@@ -0,0 +1,1940 @@
+#!/usr/bin/env perl
+#
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+#
+# This module implements support for AES instructions as per PowerISA
+# specification version 2.07, first implemented by POWER8 processor.
+# The module is endian-agnostic in sense that it supports both big-
+# and little-endian cases. Data alignment in parallelizable modes is
+# handled with VSX loads and stores, which implies MSR.VSX flag being
+# set. It should also be noted that ISA specification doesn't prohibit
+# alignment exceptions for these instructions on page boundaries.
+# Initially alignment was handled in pure AltiVec/VMX way [when data
+# is aligned programmatically, which in turn guarantees exception-
+# free execution], but it turned to hamper performance when vcipher
+# instructions are interleaved. It's reckoned that eventual
+# misalignment penalties at page boundaries are in average lower
+# than additional overhead in pure AltiVec approach.
+
+$flavour = shift;
+
+if ($flavour =~ /64/) {
+	$SIZE_T	=8;
+	$LRSAVE	=2*$SIZE_T;
+	$STU	="stdu";
+	$POP	="ld";
+	$PUSH	="std";
+	$UCMP	="cmpld";
+	$SHL	="sldi";
+} elsif ($flavour =~ /32/) {
+	$SIZE_T	=4;
+	$LRSAVE	=$SIZE_T;
+	$STU	="stwu";
+	$POP	="lwz";
+	$PUSH	="stw";
+	$UCMP	="cmplw";
+	$SHL	="slwi";
+} else { die "nonsense $flavour"; }
+
+$LITTLE_ENDIAN = ($flavour=~/le$/) ? $SIZE_T : 0;
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or
+( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or
+die "can't locate ppc-xlate.pl";
+
+open STDOUT,"| $^X $xlate $flavour ".shift || die "can't call $xlate: $!";
+
+$FRAME=8*$SIZE_T;
+$prefix="aes_p8";
+
+$sp="r1";
+$vrsave="r12";
+
+#########################################################################
+{{{	# Key setup procedures						#
+my ($inp,$bits,$out,$ptr,$cnt,$rounds)=map("r$_",(3..8));
+my ($zero,$in0,$in1,$key,$rcon,$mask,$tmp)=map("v$_",(0..6));
+my ($stage,$outperm,$outmask,$outhead,$outtail)=map("v$_",(7..11));
+
+$code.=<<___;
+.machine	"any"
+
+.text
+
+.align	7
+rcon:
+.long	0x01000000, 0x01000000, 0x01000000, 0x01000000	?rev
+.long	0x1b000000, 0x1b000000, 0x1b000000, 0x1b000000	?rev
+.long	0x0d0e0f0c, 0x0d0e0f0c, 0x0d0e0f0c, 0x0d0e0f0c	?rev
+.long	0,0,0,0						?asis
+Lconsts:
+	mflr	r0
+	bcl	20,31,\$+4
+	mflr	$ptr	 #vvvvv "distance between . and rcon
+	addi	$ptr,$ptr,-0x48
+	mtlr	r0
+	blr
+	.long	0
+	.byte	0,12,0x14,0,0,0,0,0
+.asciz	"AES for PowerISA 2.07, CRYPTOGAMS by <appro\@openssl.org>"
+
+.globl	.${prefix}_set_encrypt_key
+.align	5
+.${prefix}_set_encrypt_key:
+Lset_encrypt_key:
+	mflr		r11
+	$PUSH		r11,$LRSAVE($sp)
+
+	li		$ptr,-1
+	${UCMP}i	$inp,0
+	beq-		Lenc_key_abort		# if ($inp==0) return -1;
+	${UCMP}i	$out,0
+	beq-		Lenc_key_abort		# if ($out==0) return -1;
+	li		$ptr,-2
+	cmpwi		$bits,128
+	blt-		Lenc_key_abort
+	cmpwi		$bits,256
+	bgt-		Lenc_key_abort
+	andi.		r0,$bits,0x3f
+	bne-		Lenc_key_abort
+
+	lis		r0,0xfff0
+	mfspr		$vrsave,256
+	mtspr		256,r0
+
+	bl		Lconsts
+	mtlr		r11
+
+	neg		r9,$inp
+	lvx		$in0,0,$inp
+	addi		$inp,$inp,15		# 15 is not typo
+	lvsr		$key,0,r9		# borrow $key
+	li		r8,0x20
+	cmpwi		$bits,192
+	lvx		$in1,0,$inp
+	le?vspltisb	$mask,0x0f		# borrow $mask
+	lvx		$rcon,0,$ptr
+	le?vxor		$key,$key,$mask		# adjust for byte swap
+	lvx		$mask,r8,$ptr
+	addi		$ptr,$ptr,0x10
+	vperm		$in0,$in0,$in1,$key	# align [and byte swap in LE]
+	li		$cnt,8
+	vxor		$zero,$zero,$zero
+	mtctr		$cnt
+
+	?lvsr		$outperm,0,$out
+	vspltisb	$outmask,-1
+	lvx		$outhead,0,$out
+	?vperm		$outmask,$zero,$outmask,$outperm
+
+	blt		Loop128
+	addi		$inp,$inp,8
+	beq		L192
+	addi		$inp,$inp,8
+	b		L256
+
+.align	4
+Loop128:
+	vperm		$key,$in0,$in0,$mask	# rotate-n-splat
+	vsldoi		$tmp,$zero,$in0,12	# >>32
+	 vperm		$outtail,$in0,$in0,$outperm	# rotate
+	 vsel		$stage,$outhead,$outtail,$outmask
+	 vmr		$outhead,$outtail
+	vcipherlast	$key,$key,$rcon
+	 stvx		$stage,0,$out
+	 addi		$out,$out,16
+
+	vxor		$in0,$in0,$tmp
+	vsldoi		$tmp,$zero,$tmp,12	# >>32
+	vxor		$in0,$in0,$tmp
+	vsldoi		$tmp,$zero,$tmp,12	# >>32
+	vxor		$in0,$in0,$tmp
+	 vadduwm	$rcon,$rcon,$rcon
+	vxor		$in0,$in0,$key
+	bdnz		Loop128
+
+	lvx		$rcon,0,$ptr		# last two round keys
+
+	vperm		$key,$in0,$in0,$mask	# rotate-n-splat
+	vsldoi		$tmp,$zero,$in0,12	# >>32
+	 vperm		$outtail,$in0,$in0,$outperm	# rotate
+	 vsel		$stage,$outhead,$outtail,$outmask
+	 vmr		$outhead,$outtail
+	vcipherlast	$key,$key,$rcon
+	 stvx		$stage,0,$out
+	 addi		$out,$out,16
+
+	vxor		$in0,$in0,$tmp
+	vsldoi		$tmp,$zero,$tmp,12	# >>32
+	vxor		$in0,$in0,$tmp
+	vsldoi		$tmp,$zero,$tmp,12	# >>32
+	vxor		$in0,$in0,$tmp
+	 vadduwm	$rcon,$rcon,$rcon
+	vxor		$in0,$in0,$key
+
+	vperm		$key,$in0,$in0,$mask	# rotate-n-splat
+	vsldoi		$tmp,$zero,$in0,12	# >>32
+	 vperm		$outtail,$in0,$in0,$outperm	# rotate
+	 vsel		$stage,$outhead,$outtail,$outmask
+	 vmr		$outhead,$outtail
+	vcipherlast	$key,$key,$rcon
+	 stvx		$stage,0,$out
+	 addi		$out,$out,16
+
+	vxor		$in0,$in0,$tmp
+	vsldoi		$tmp,$zero,$tmp,12	# >>32
+	vxor		$in0,$in0,$tmp
+	vsldoi		$tmp,$zero,$tmp,12	# >>32
+	vxor		$in0,$in0,$tmp
+	vxor		$in0,$in0,$key
+	 vperm		$outtail,$in0,$in0,$outperm	# rotate
+	 vsel		$stage,$outhead,$outtail,$outmask
+	 vmr		$outhead,$outtail
+	 stvx		$stage,0,$out
+
+	addi		$inp,$out,15		# 15 is not typo
+	addi		$out,$out,0x50
+
+	li		$rounds,10
+	b		Ldone
+
+.align	4
+L192:
+	lvx		$tmp,0,$inp
+	li		$cnt,4
+	 vperm		$outtail,$in0,$in0,$outperm	# rotate
+	 vsel		$stage,$outhead,$outtail,$outmask
+	 vmr		$outhead,$outtail
+	 stvx		$stage,0,$out
+	 addi		$out,$out,16
+	vperm		$in1,$in1,$tmp,$key	# align [and byte swap in LE]
+	vspltisb	$key,8			# borrow $key
+	mtctr		$cnt
+	vsububm		$mask,$mask,$key	# adjust the mask
+
+Loop192:
+	vperm		$key,$in1,$in1,$mask	# roate-n-splat
+	vsldoi		$tmp,$zero,$in0,12	# >>32
+	vcipherlast	$key,$key,$rcon
+
+	vxor		$in0,$in0,$tmp
+	vsldoi		$tmp,$zero,$tmp,12	# >>32
+	vxor		$in0,$in0,$tmp
+	vsldoi		$tmp,$zero,$tmp,12	# >>32
+	vxor		$in0,$in0,$tmp
+
+	 vsldoi		$stage,$zero,$in1,8
+	vspltw		$tmp,$in0,3
+	vxor		$tmp,$tmp,$in1
+	vsldoi		$in1,$zero,$in1,12	# >>32
+	 vadduwm	$rcon,$rcon,$rcon
+	vxor		$in1,$in1,$tmp
+	vxor		$in0,$in0,$key
+	vxor		$in1,$in1,$key
+	 vsldoi		$stage,$stage,$in0,8
+
+	vperm		$key,$in1,$in1,$mask	# rotate-n-splat
+	vsldoi		$tmp,$zero,$in0,12	# >>32
+	 vperm		$outtail,$stage,$stage,$outperm	# rotate
+	 vsel		$stage,$outhead,$outtail,$outmask
+	 vmr		$outhead,$outtail
+	vcipherlast	$key,$key,$rcon
+	 stvx		$stage,0,$out
+	 addi		$out,$out,16
+
+	 vsldoi		$stage,$in0,$in1,8
+	vxor		$in0,$in0,$tmp
+	vsldoi		$tmp,$zero,$tmp,12	# >>32
+	 vperm		$outtail,$stage,$stage,$outperm	# rotate
+	 vsel		$stage,$outhead,$outtail,$outmask
+	 vmr		$outhead,$outtail
+	vxor		$in0,$in0,$tmp
+	vsldoi		$tmp,$zero,$tmp,12	# >>32
+	vxor		$in0,$in0,$tmp
+	 stvx		$stage,0,$out
+	 addi		$out,$out,16
+
+	vspltw		$tmp,$in0,3
+	vxor		$tmp,$tmp,$in1
+	vsldoi		$in1,$zero,$in1,12	# >>32
+	 vadduwm	$rcon,$rcon,$rcon
+	vxor		$in1,$in1,$tmp
+	vxor		$in0,$in0,$key
+	vxor		$in1,$in1,$key
+	 vperm		$outtail,$in0,$in0,$outperm	# rotate
+	 vsel		$stage,$outhead,$outtail,$outmask
+	 vmr		$outhead,$outtail
+	 stvx		$stage,0,$out
+	 addi		$inp,$out,15		# 15 is not typo
+	 addi		$out,$out,16
+	bdnz		Loop192
+
+	li		$rounds,12
+	addi		$out,$out,0x20
+	b		Ldone
+
+.align	4
+L256:
+	lvx		$tmp,0,$inp
+	li		$cnt,7
+	li		$rounds,14
+	 vperm		$outtail,$in0,$in0,$outperm	# rotate
+	 vsel		$stage,$outhead,$outtail,$outmask
+	 vmr		$outhead,$outtail
+	 stvx		$stage,0,$out
+	 addi		$out,$out,16
+	vperm		$in1,$in1,$tmp,$key	# align [and byte swap in LE]
+	mtctr		$cnt
+
+Loop256:
+	vperm		$key,$in1,$in1,$mask	# rotate-n-splat
+	vsldoi		$tmp,$zero,$in0,12	# >>32
+	 vperm		$outtail,$in1,$in1,$outperm	# rotate
+	 vsel		$stage,$outhead,$outtail,$outmask
+	 vmr		$outhead,$outtail
+	vcipherlast	$key,$key,$rcon
+	 stvx		$stage,0,$out
+	 addi		$out,$out,16
+
+	vxor		$in0,$in0,$tmp
+	vsldoi		$tmp,$zero,$tmp,12	# >>32
+	vxor		$in0,$in0,$tmp
+	vsldoi		$tmp,$zero,$tmp,12	# >>32
+	vxor		$in0,$in0,$tmp
+	 vadduwm	$rcon,$rcon,$rcon
+	vxor		$in0,$in0,$key
+	 vperm		$outtail,$in0,$in0,$outperm	# rotate
+	 vsel		$stage,$outhead,$outtail,$outmask
+	 vmr		$outhead,$outtail
+	 stvx		$stage,0,$out
+	 addi		$inp,$out,15		# 15 is not typo
+	 addi		$out,$out,16
+	bdz		Ldone
+
+	vspltw		$key,$in0,3		# just splat
+	vsldoi		$tmp,$zero,$in1,12	# >>32
+	vsbox		$key,$key
+
+	vxor		$in1,$in1,$tmp
+	vsldoi		$tmp,$zero,$tmp,12	# >>32
+	vxor		$in1,$in1,$tmp
+	vsldoi		$tmp,$zero,$tmp,12	# >>32
+	vxor		$in1,$in1,$tmp
+
+	vxor		$in1,$in1,$key
+	b		Loop256
+
+.align	4
+Ldone:
+	lvx		$in1,0,$inp		# redundant in aligned case
+	vsel		$in1,$outhead,$in1,$outmask
+	stvx		$in1,0,$inp
+	li		$ptr,0
+	mtspr		256,$vrsave
+	stw		$rounds,0($out)
+
+Lenc_key_abort:
+	mr		r3,$ptr
+	blr
+	.long		0
+	.byte		0,12,0x14,1,0,0,3,0
+	.long		0
+.size	.${prefix}_set_encrypt_key,.-.${prefix}_set_encrypt_key
+
+.globl	.${prefix}_set_decrypt_key
+.align	5
+.${prefix}_set_decrypt_key:
+	$STU		$sp,-$FRAME($sp)
+	mflr		r10
+	$PUSH		r10,$FRAME+$LRSAVE($sp)
+	bl		Lset_encrypt_key
+	mtlr		r10
+
+	cmpwi		r3,0
+	bne-		Ldec_key_abort
+
+	slwi		$cnt,$rounds,4
+	subi		$inp,$out,240		# first round key
+	srwi		$rounds,$rounds,1
+	add		$out,$inp,$cnt		# last round key
+	mtctr		$rounds
+
+Ldeckey:
+	lwz		r0, 0($inp)
+	lwz		r6, 4($inp)
+	lwz		r7, 8($inp)
+	lwz		r8, 12($inp)
+	addi		$inp,$inp,16
+	lwz		r9, 0($out)
+	lwz		r10,4($out)
+	lwz		r11,8($out)
+	lwz		r12,12($out)
+	stw		r0, 0($out)
+	stw		r6, 4($out)
+	stw		r7, 8($out)
+	stw		r8, 12($out)
+	subi		$out,$out,16
+	stw		r9, -16($inp)
+	stw		r10,-12($inp)
+	stw		r11,-8($inp)
+	stw		r12,-4($inp)
+	bdnz		Ldeckey
+
+	xor		r3,r3,r3		# return value
+Ldec_key_abort:
+	addi		$sp,$sp,$FRAME
+	blr
+	.long		0
+	.byte		0,12,4,1,0x80,0,3,0
+	.long		0
+.size	.${prefix}_set_decrypt_key,.-.${prefix}_set_decrypt_key
+___
+}}}
+#########################################################################
+{{{	# Single block en- and decrypt procedures			#
+sub gen_block () {
+my $dir = shift;
+my $n   = $dir eq "de" ? "n" : "";
+my ($inp,$out,$key,$rounds,$idx)=map("r$_",(3..7));
+
+$code.=<<___;
+.globl	.${prefix}_${dir}crypt
+.align	5
+.${prefix}_${dir}crypt:
+	lwz		$rounds,240($key)
+	lis		r0,0xfc00
+	mfspr		$vrsave,256
+	li		$idx,15			# 15 is not typo
+	mtspr		256,r0
+
+	lvx		v0,0,$inp
+	neg		r11,$out
+	lvx		v1,$idx,$inp
+	lvsl		v2,0,$inp		# inpperm
+	le?vspltisb	v4,0x0f
+	?lvsl		v3,0,r11		# outperm
+	le?vxor		v2,v2,v4
+	li		$idx,16
+	vperm		v0,v0,v1,v2		# align [and byte swap in LE]
+	lvx		v1,0,$key
+	?lvsl		v5,0,$key		# keyperm
+	srwi		$rounds,$rounds,1
+	lvx		v2,$idx,$key
+	addi		$idx,$idx,16
+	subi		$rounds,$rounds,1
+	?vperm		v1,v1,v2,v5		# align round key
+
+	vxor		v0,v0,v1
+	lvx		v1,$idx,$key
+	addi		$idx,$idx,16
+	mtctr		$rounds
+
+Loop_${dir}c:
+	?vperm		v2,v2,v1,v5
+	v${n}cipher	v0,v0,v2
+	lvx		v2,$idx,$key
+	addi		$idx,$idx,16
+	?vperm		v1,v1,v2,v5
+	v${n}cipher	v0,v0,v1
+	lvx		v1,$idx,$key
+	addi		$idx,$idx,16
+	bdnz		Loop_${dir}c
+
+	?vperm		v2,v2,v1,v5
+	v${n}cipher	v0,v0,v2
+	lvx		v2,$idx,$key
+	?vperm		v1,v1,v2,v5
+	v${n}cipherlast	v0,v0,v1
+
+	vspltisb	v2,-1
+	vxor		v1,v1,v1
+	li		$idx,15			# 15 is not typo
+	?vperm		v2,v1,v2,v3		# outmask
+	le?vxor		v3,v3,v4
+	lvx		v1,0,$out		# outhead
+	vperm		v0,v0,v0,v3		# rotate [and byte swap in LE]
+	vsel		v1,v1,v0,v2
+	lvx		v4,$idx,$out
+	stvx		v1,0,$out
+	vsel		v0,v0,v4,v2
+	stvx		v0,$idx,$out
+
+	mtspr		256,$vrsave
+	blr
+	.long		0
+	.byte		0,12,0x14,0,0,0,3,0
+	.long		0
+.size	.${prefix}_${dir}crypt,.-.${prefix}_${dir}crypt
+___
+}
+&gen_block("en");
+&gen_block("de");
+}}}
+#########################################################################
+{{{	# CBC en- and decrypt procedures				#
+my ($inp,$out,$len,$key,$ivp,$enc,$rounds,$idx)=map("r$_",(3..10));
+my ($rndkey0,$rndkey1,$inout,$tmp)=		map("v$_",(0..3));
+my ($ivec,$inptail,$inpperm,$outhead,$outperm,$outmask,$keyperm)=
+						map("v$_",(4..10));
+$code.=<<___;
+.globl	.${prefix}_cbc_encrypt
+.align	5
+.${prefix}_cbc_encrypt:
+	${UCMP}i	$len,16
+	bltlr-
+
+	cmpwi		$enc,0			# test direction
+	lis		r0,0xffe0
+	mfspr		$vrsave,256
+	mtspr		256,r0
+
+	li		$idx,15
+	vxor		$rndkey0,$rndkey0,$rndkey0
+	le?vspltisb	$tmp,0x0f
+
+	lvx		$ivec,0,$ivp		# load [unaligned] iv
+	lvsl		$inpperm,0,$ivp
+	lvx		$inptail,$idx,$ivp
+	le?vxor		$inpperm,$inpperm,$tmp
+	vperm		$ivec,$ivec,$inptail,$inpperm
+
+	neg		r11,$inp
+	?lvsl		$keyperm,0,$key		# prepare for unaligned key
+	lwz		$rounds,240($key)
+
+	lvsr		$inpperm,0,r11		# prepare for unaligned load
+	lvx		$inptail,0,$inp
+	addi		$inp,$inp,15		# 15 is not typo
+	le?vxor		$inpperm,$inpperm,$tmp
+
+	?lvsr		$outperm,0,$out		# prepare for unaligned store
+	vspltisb	$outmask,-1
+	lvx		$outhead,0,$out
+	?vperm		$outmask,$rndkey0,$outmask,$outperm
+	le?vxor		$outperm,$outperm,$tmp
+
+	srwi		$rounds,$rounds,1
+	li		$idx,16
+	subi		$rounds,$rounds,1
+	beq		Lcbc_dec
+
+Lcbc_enc:
+	vmr		$inout,$inptail
+	lvx		$inptail,0,$inp
+	addi		$inp,$inp,16
+	mtctr		$rounds
+	subi		$len,$len,16		# len-=16
+
+	lvx		$rndkey0,0,$key
+	 vperm		$inout,$inout,$inptail,$inpperm
+	lvx		$rndkey1,$idx,$key
+	addi		$idx,$idx,16
+	?vperm		$rndkey0,$rndkey0,$rndkey1,$keyperm
+	vxor		$inout,$inout,$rndkey0
+	lvx		$rndkey0,$idx,$key
+	addi		$idx,$idx,16
+	vxor		$inout,$inout,$ivec
+
+Loop_cbc_enc:
+	?vperm		$rndkey1,$rndkey1,$rndkey0,$keyperm
+	vcipher		$inout,$inout,$rndkey1
+	lvx		$rndkey1,$idx,$key
+	addi		$idx,$idx,16
+	?vperm		$rndkey0,$rndkey0,$rndkey1,$keyperm
+	vcipher		$inout,$inout,$rndkey0
+	lvx		$rndkey0,$idx,$key
+	addi		$idx,$idx,16
+	bdnz		Loop_cbc_enc
+
+	?vperm		$rndkey1,$rndkey1,$rndkey0,$keyperm
+	vcipher		$inout,$inout,$rndkey1
+	lvx		$rndkey1,$idx,$key
+	li		$idx,16
+	?vperm		$rndkey0,$rndkey0,$rndkey1,$keyperm
+	vcipherlast	$ivec,$inout,$rndkey0
+	${UCMP}i	$len,16
+
+	vperm		$tmp,$ivec,$ivec,$outperm
+	vsel		$inout,$outhead,$tmp,$outmask
+	vmr		$outhead,$tmp
+	stvx		$inout,0,$out
+	addi		$out,$out,16
+	bge		Lcbc_enc
+
+	b		Lcbc_done
+
+.align	4
+Lcbc_dec:
+	${UCMP}i	$len,128
+	bge		_aesp8_cbc_decrypt8x
+	vmr		$tmp,$inptail
+	lvx		$inptail,0,$inp
+	addi		$inp,$inp,16
+	mtctr		$rounds
+	subi		$len,$len,16		# len-=16
+
+	lvx		$rndkey0,0,$key
+	 vperm		$tmp,$tmp,$inptail,$inpperm
+	lvx		$rndkey1,$idx,$key
+	addi		$idx,$idx,16
+	?vperm		$rndkey0,$rndkey0,$rndkey1,$keyperm
+	vxor		$inout,$tmp,$rndkey0
+	lvx		$rndkey0,$idx,$key
+	addi		$idx,$idx,16
+
+Loop_cbc_dec:
+	?vperm		$rndkey1,$rndkey1,$rndkey0,$keyperm
+	vncipher	$inout,$inout,$rndkey1
+	lvx		$rndkey1,$idx,$key
+	addi		$idx,$idx,16
+	?vperm		$rndkey0,$rndkey0,$rndkey1,$keyperm
+	vncipher	$inout,$inout,$rndkey0
+	lvx		$rndkey0,$idx,$key
+	addi		$idx,$idx,16
+	bdnz		Loop_cbc_dec
+
+	?vperm		$rndkey1,$rndkey1,$rndkey0,$keyperm
+	vncipher	$inout,$inout,$rndkey1
+	lvx		$rndkey1,$idx,$key
+	li		$idx,16
+	?vperm		$rndkey0,$rndkey0,$rndkey1,$keyperm
+	vncipherlast	$inout,$inout,$rndkey0
+	${UCMP}i	$len,16
+
+	vxor		$inout,$inout,$ivec
+	vmr		$ivec,$tmp
+	vperm		$tmp,$inout,$inout,$outperm
+	vsel		$inout,$outhead,$tmp,$outmask
+	vmr		$outhead,$tmp
+	stvx		$inout,0,$out
+	addi		$out,$out,16
+	bge		Lcbc_dec
+
+Lcbc_done:
+	addi		$out,$out,-1
+	lvx		$inout,0,$out		# redundant in aligned case
+	vsel		$inout,$outhead,$inout,$outmask
+	stvx		$inout,0,$out
+
+	neg		$enc,$ivp		# write [unaligned] iv
+	li		$idx,15			# 15 is not typo
+	vxor		$rndkey0,$rndkey0,$rndkey0
+	vspltisb	$outmask,-1
+	le?vspltisb	$tmp,0x0f
+	?lvsl		$outperm,0,$enc
+	?vperm		$outmask,$rndkey0,$outmask,$outperm
+	le?vxor		$outperm,$outperm,$tmp
+	lvx		$outhead,0,$ivp
+	vperm		$ivec,$ivec,$ivec,$outperm
+	vsel		$inout,$outhead,$ivec,$outmask
+	lvx		$inptail,$idx,$ivp
+	stvx		$inout,0,$ivp
+	vsel		$inout,$ivec,$inptail,$outmask
+	stvx		$inout,$idx,$ivp
+
+	mtspr		256,$vrsave
+	blr
+	.long		0
+	.byte		0,12,0x14,0,0,0,6,0
+	.long		0
+___
+#########################################################################
+{{	# Optimized CBC decrypt procedure				#
+my $key_="r11";
+my ($x00,$x10,$x20,$x30,$x40,$x50,$x60,$x70)=map("r$_",(0,8,26..31));
+my ($in0, $in1, $in2, $in3, $in4, $in5, $in6, $in7 )=map("v$_",(0..3,10..13));
+my ($out0,$out1,$out2,$out3,$out4,$out5,$out6,$out7)=map("v$_",(14..21));
+my $rndkey0="v23";	# v24-v25 rotating buffer for first found keys
+			# v26-v31 last 6 round keys
+my ($tmp,$keyperm)=($in3,$in4);	# aliases with "caller", redundant assignment
+
+$code.=<<___;
+.align	5
+_aesp8_cbc_decrypt8x:
+	$STU		$sp,-`($FRAME+21*16+6*$SIZE_T)`($sp)
+	li		r10,`$FRAME+8*16+15`
+	li		r11,`$FRAME+8*16+31`
+	stvx		v20,r10,$sp		# ABI says so
+	addi		r10,r10,32
+	stvx		v21,r11,$sp
+	addi		r11,r11,32
+	stvx		v22,r10,$sp
+	addi		r10,r10,32
+	stvx		v23,r11,$sp
+	addi		r11,r11,32
+	stvx		v24,r10,$sp
+	addi		r10,r10,32
+	stvx		v25,r11,$sp
+	addi		r11,r11,32
+	stvx		v26,r10,$sp
+	addi		r10,r10,32
+	stvx		v27,r11,$sp
+	addi		r11,r11,32
+	stvx		v28,r10,$sp
+	addi		r10,r10,32
+	stvx		v29,r11,$sp
+	addi		r11,r11,32
+	stvx		v30,r10,$sp
+	stvx		v31,r11,$sp
+	li		r0,-1
+	stw		$vrsave,`$FRAME+21*16-4`($sp)	# save vrsave
+	li		$x10,0x10
+	$PUSH		r26,`$FRAME+21*16+0*$SIZE_T`($sp)
+	li		$x20,0x20
+	$PUSH		r27,`$FRAME+21*16+1*$SIZE_T`($sp)
+	li		$x30,0x30
+	$PUSH		r28,`$FRAME+21*16+2*$SIZE_T`($sp)
+	li		$x40,0x40
+	$PUSH		r29,`$FRAME+21*16+3*$SIZE_T`($sp)
+	li		$x50,0x50
+	$PUSH		r30,`$FRAME+21*16+4*$SIZE_T`($sp)
+	li		$x60,0x60
+	$PUSH		r31,`$FRAME+21*16+5*$SIZE_T`($sp)
+	li		$x70,0x70
+	mtspr		256,r0
+
+	subi		$rounds,$rounds,3	# -4 in total
+	subi		$len,$len,128		# bias
+
+	lvx		$rndkey0,$x00,$key	# load key schedule
+	lvx		v30,$x10,$key
+	addi		$key,$key,0x20
+	lvx		v31,$x00,$key
+	?vperm		$rndkey0,$rndkey0,v30,$keyperm
+	addi		$key_,$sp,$FRAME+15
+	mtctr		$rounds
+
+Load_cbc_dec_key:
+	?vperm		v24,v30,v31,$keyperm
+	lvx		v30,$x10,$key
+	addi		$key,$key,0x20
+	stvx		v24,$x00,$key_		# off-load round[1]
+	?vperm		v25,v31,v30,$keyperm
+	lvx		v31,$x00,$key
+	stvx		v25,$x10,$key_		# off-load round[2]
+	addi		$key_,$key_,0x20
+	bdnz		Load_cbc_dec_key
+
+	lvx		v26,$x10,$key
+	?vperm		v24,v30,v31,$keyperm
+	lvx		v27,$x20,$key
+	stvx		v24,$x00,$key_		# off-load round[3]
+	?vperm		v25,v31,v26,$keyperm
+	lvx		v28,$x30,$key
+	stvx		v25,$x10,$key_		# off-load round[4]
+	addi		$key_,$sp,$FRAME+15	# rewind $key_
+	?vperm		v26,v26,v27,$keyperm
+	lvx		v29,$x40,$key
+	?vperm		v27,v27,v28,$keyperm
+	lvx		v30,$x50,$key
+	?vperm		v28,v28,v29,$keyperm
+	lvx		v31,$x60,$key
+	?vperm		v29,v29,v30,$keyperm
+	lvx		$out0,$x70,$key		# borrow $out0
+	?vperm		v30,v30,v31,$keyperm
+	lvx		v24,$x00,$key_		# pre-load round[1]
+	?vperm		v31,v31,$out0,$keyperm
+	lvx		v25,$x10,$key_		# pre-load round[2]
+
+	#lvx		$inptail,0,$inp		# "caller" already did this
+	#addi		$inp,$inp,15		# 15 is not typo
+	subi		$inp,$inp,15		# undo "caller"
+
+	 le?li		$idx,8
+	lvx_u		$in0,$x00,$inp		# load first 8 "words"
+	 le?lvsl	$inpperm,0,$idx
+	 le?vspltisb	$tmp,0x0f
+	lvx_u		$in1,$x10,$inp
+	 le?vxor	$inpperm,$inpperm,$tmp	# transform for lvx_u/stvx_u
+	lvx_u		$in2,$x20,$inp
+	 le?vperm	$in0,$in0,$in0,$inpperm
+	lvx_u		$in3,$x30,$inp
+	 le?vperm	$in1,$in1,$in1,$inpperm
+	lvx_u		$in4,$x40,$inp
+	 le?vperm	$in2,$in2,$in2,$inpperm
+	vxor		$out0,$in0,$rndkey0
+	lvx_u		$in5,$x50,$inp
+	 le?vperm	$in3,$in3,$in3,$inpperm
+	vxor		$out1,$in1,$rndkey0
+	lvx_u		$in6,$x60,$inp
+	 le?vperm	$in4,$in4,$in4,$inpperm
+	vxor		$out2,$in2,$rndkey0
+	lvx_u		$in7,$x70,$inp
+	addi		$inp,$inp,0x80
+	 le?vperm	$in5,$in5,$in5,$inpperm
+	vxor		$out3,$in3,$rndkey0
+	 le?vperm	$in6,$in6,$in6,$inpperm
+	vxor		$out4,$in4,$rndkey0
+	 le?vperm	$in7,$in7,$in7,$inpperm
+	vxor		$out5,$in5,$rndkey0
+	vxor		$out6,$in6,$rndkey0
+	vxor		$out7,$in7,$rndkey0
+
+	mtctr		$rounds
+	b		Loop_cbc_dec8x
+.align	5
+Loop_cbc_dec8x:
+	vncipher	$out0,$out0,v24
+	vncipher	$out1,$out1,v24
+	vncipher	$out2,$out2,v24
+	vncipher	$out3,$out3,v24
+	vncipher	$out4,$out4,v24
+	vncipher	$out5,$out5,v24
+	vncipher	$out6,$out6,v24
+	vncipher	$out7,$out7,v24
+	lvx		v24,$x20,$key_		# round[3]
+	addi		$key_,$key_,0x20
+
+	vncipher	$out0,$out0,v25
+	vncipher	$out1,$out1,v25
+	vncipher	$out2,$out2,v25
+	vncipher	$out3,$out3,v25
+	vncipher	$out4,$out4,v25
+	vncipher	$out5,$out5,v25
+	vncipher	$out6,$out6,v25
+	vncipher	$out7,$out7,v25
+	lvx		v25,$x10,$key_		# round[4]
+	bdnz		Loop_cbc_dec8x
+
+	subic		$len,$len,128		# $len-=128
+	vncipher	$out0,$out0,v24
+	vncipher	$out1,$out1,v24
+	vncipher	$out2,$out2,v24
+	vncipher	$out3,$out3,v24
+	vncipher	$out4,$out4,v24
+	vncipher	$out5,$out5,v24
+	vncipher	$out6,$out6,v24
+	vncipher	$out7,$out7,v24
+
+	subfe.		r0,r0,r0		# borrow?-1:0
+	vncipher	$out0,$out0,v25
+	vncipher	$out1,$out1,v25
+	vncipher	$out2,$out2,v25
+	vncipher	$out3,$out3,v25
+	vncipher	$out4,$out4,v25
+	vncipher	$out5,$out5,v25
+	vncipher	$out6,$out6,v25
+	vncipher	$out7,$out7,v25
+
+	and		r0,r0,$len
+	vncipher	$out0,$out0,v26
+	vncipher	$out1,$out1,v26
+	vncipher	$out2,$out2,v26
+	vncipher	$out3,$out3,v26
+	vncipher	$out4,$out4,v26
+	vncipher	$out5,$out5,v26
+	vncipher	$out6,$out6,v26
+	vncipher	$out7,$out7,v26
+
+	add		$inp,$inp,r0		# $inp is adjusted in such
+						# way that at exit from the
+						# loop inX-in7 are loaded
+						# with last "words"
+	vncipher	$out0,$out0,v27
+	vncipher	$out1,$out1,v27
+	vncipher	$out2,$out2,v27
+	vncipher	$out3,$out3,v27
+	vncipher	$out4,$out4,v27
+	vncipher	$out5,$out5,v27
+	vncipher	$out6,$out6,v27
+	vncipher	$out7,$out7,v27
+
+	addi		$key_,$sp,$FRAME+15	# rewind $key_
+	vncipher	$out0,$out0,v28
+	vncipher	$out1,$out1,v28
+	vncipher	$out2,$out2,v28
+	vncipher	$out3,$out3,v28
+	vncipher	$out4,$out4,v28
+	vncipher	$out5,$out5,v28
+	vncipher	$out6,$out6,v28
+	vncipher	$out7,$out7,v28
+	lvx		v24,$x00,$key_		# re-pre-load round[1]
+
+	vncipher	$out0,$out0,v29
+	vncipher	$out1,$out1,v29
+	vncipher	$out2,$out2,v29
+	vncipher	$out3,$out3,v29
+	vncipher	$out4,$out4,v29
+	vncipher	$out5,$out5,v29
+	vncipher	$out6,$out6,v29
+	vncipher	$out7,$out7,v29
+	lvx		v25,$x10,$key_		# re-pre-load round[2]
+
+	vncipher	$out0,$out0,v30
+	 vxor		$ivec,$ivec,v31		# xor with last round key
+	vncipher	$out1,$out1,v30
+	 vxor		$in0,$in0,v31
+	vncipher	$out2,$out2,v30
+	 vxor		$in1,$in1,v31
+	vncipher	$out3,$out3,v30
+	 vxor		$in2,$in2,v31
+	vncipher	$out4,$out4,v30
+	 vxor		$in3,$in3,v31
+	vncipher	$out5,$out5,v30
+	 vxor		$in4,$in4,v31
+	vncipher	$out6,$out6,v30
+	 vxor		$in5,$in5,v31
+	vncipher	$out7,$out7,v30
+	 vxor		$in6,$in6,v31
+
+	vncipherlast	$out0,$out0,$ivec
+	vncipherlast	$out1,$out1,$in0
+	 lvx_u		$in0,$x00,$inp		# load next input block
+	vncipherlast	$out2,$out2,$in1
+	 lvx_u		$in1,$x10,$inp
+	vncipherlast	$out3,$out3,$in2
+	 le?vperm	$in0,$in0,$in0,$inpperm
+	 lvx_u		$in2,$x20,$inp
+	vncipherlast	$out4,$out4,$in3
+	 le?vperm	$in1,$in1,$in1,$inpperm
+	 lvx_u		$in3,$x30,$inp
+	vncipherlast	$out5,$out5,$in4
+	 le?vperm	$in2,$in2,$in2,$inpperm
+	 lvx_u		$in4,$x40,$inp
+	vncipherlast	$out6,$out6,$in5
+	 le?vperm	$in3,$in3,$in3,$inpperm
+	 lvx_u		$in5,$x50,$inp
+	vncipherlast	$out7,$out7,$in6
+	 le?vperm	$in4,$in4,$in4,$inpperm
+	 lvx_u		$in6,$x60,$inp
+	vmr		$ivec,$in7
+	 le?vperm	$in5,$in5,$in5,$inpperm
+	 lvx_u		$in7,$x70,$inp
+	 addi		$inp,$inp,0x80
+
+	le?vperm	$out0,$out0,$out0,$inpperm
+	le?vperm	$out1,$out1,$out1,$inpperm
+	stvx_u		$out0,$x00,$out
+	 le?vperm	$in6,$in6,$in6,$inpperm
+	 vxor		$out0,$in0,$rndkey0
+	le?vperm	$out2,$out2,$out2,$inpperm
+	stvx_u		$out1,$x10,$out
+	 le?vperm	$in7,$in7,$in7,$inpperm
+	 vxor		$out1,$in1,$rndkey0
+	le?vperm	$out3,$out3,$out3,$inpperm
+	stvx_u		$out2,$x20,$out
+	 vxor		$out2,$in2,$rndkey0
+	le?vperm	$out4,$out4,$out4,$inpperm
+	stvx_u		$out3,$x30,$out
+	 vxor		$out3,$in3,$rndkey0
+	le?vperm	$out5,$out5,$out5,$inpperm
+	stvx_u		$out4,$x40,$out
+	 vxor		$out4,$in4,$rndkey0
+	le?vperm	$out6,$out6,$out6,$inpperm
+	stvx_u		$out5,$x50,$out
+	 vxor		$out5,$in5,$rndkey0
+	le?vperm	$out7,$out7,$out7,$inpperm
+	stvx_u		$out6,$x60,$out
+	 vxor		$out6,$in6,$rndkey0
+	stvx_u		$out7,$x70,$out
+	addi		$out,$out,0x80
+	 vxor		$out7,$in7,$rndkey0
+
+	mtctr		$rounds
+	beq		Loop_cbc_dec8x		# did $len-=128 borrow?
+
+	addic.		$len,$len,128
+	beq		Lcbc_dec8x_done
+	nop
+	nop
+
+Loop_cbc_dec8x_tail:				# up to 7 "words" tail...
+	vncipher	$out1,$out1,v24
+	vncipher	$out2,$out2,v24
+	vncipher	$out3,$out3,v24
+	vncipher	$out4,$out4,v24
+	vncipher	$out5,$out5,v24
+	vncipher	$out6,$out6,v24
+	vncipher	$out7,$out7,v24
+	lvx		v24,$x20,$key_		# round[3]
+	addi		$key_,$key_,0x20
+
+	vncipher	$out1,$out1,v25
+	vncipher	$out2,$out2,v25
+	vncipher	$out3,$out3,v25
+	vncipher	$out4,$out4,v25
+	vncipher	$out5,$out5,v25
+	vncipher	$out6,$out6,v25
+	vncipher	$out7,$out7,v25
+	lvx		v25,$x10,$key_		# round[4]
+	bdnz		Loop_cbc_dec8x_tail
+
+	vncipher	$out1,$out1,v24
+	vncipher	$out2,$out2,v24
+	vncipher	$out3,$out3,v24
+	vncipher	$out4,$out4,v24
+	vncipher	$out5,$out5,v24
+	vncipher	$out6,$out6,v24
+	vncipher	$out7,$out7,v24
+
+	vncipher	$out1,$out1,v25
+	vncipher	$out2,$out2,v25
+	vncipher	$out3,$out3,v25
+	vncipher	$out4,$out4,v25
+	vncipher	$out5,$out5,v25
+	vncipher	$out6,$out6,v25
+	vncipher	$out7,$out7,v25
+
+	vncipher	$out1,$out1,v26
+	vncipher	$out2,$out2,v26
+	vncipher	$out3,$out3,v26
+	vncipher	$out4,$out4,v26
+	vncipher	$out5,$out5,v26
+	vncipher	$out6,$out6,v26
+	vncipher	$out7,$out7,v26
+
+	vncipher	$out1,$out1,v27
+	vncipher	$out2,$out2,v27
+	vncipher	$out3,$out3,v27
+	vncipher	$out4,$out4,v27
+	vncipher	$out5,$out5,v27
+	vncipher	$out6,$out6,v27
+	vncipher	$out7,$out7,v27
+
+	vncipher	$out1,$out1,v28
+	vncipher	$out2,$out2,v28
+	vncipher	$out3,$out3,v28
+	vncipher	$out4,$out4,v28
+	vncipher	$out5,$out5,v28
+	vncipher	$out6,$out6,v28
+	vncipher	$out7,$out7,v28
+
+	vncipher	$out1,$out1,v29
+	vncipher	$out2,$out2,v29
+	vncipher	$out3,$out3,v29
+	vncipher	$out4,$out4,v29
+	vncipher	$out5,$out5,v29
+	vncipher	$out6,$out6,v29
+	vncipher	$out7,$out7,v29
+
+	vncipher	$out1,$out1,v30
+	 vxor		$ivec,$ivec,v31		# last round key
+	vncipher	$out2,$out2,v30
+	 vxor		$in1,$in1,v31
+	vncipher	$out3,$out3,v30
+	 vxor		$in2,$in2,v31
+	vncipher	$out4,$out4,v30
+	 vxor		$in3,$in3,v31
+	vncipher	$out5,$out5,v30
+	 vxor		$in4,$in4,v31
+	vncipher	$out6,$out6,v30
+	 vxor		$in5,$in5,v31
+	vncipher	$out7,$out7,v30
+	 vxor		$in6,$in6,v31
+
+	cmplwi		$len,32			# switch($len)
+	blt		Lcbc_dec8x_one
+	nop
+	beq		Lcbc_dec8x_two
+	cmplwi		$len,64
+	blt		Lcbc_dec8x_three
+	nop
+	beq		Lcbc_dec8x_four
+	cmplwi		$len,96
+	blt		Lcbc_dec8x_five
+	nop
+	beq		Lcbc_dec8x_six
+
+Lcbc_dec8x_seven:
+	vncipherlast	$out1,$out1,$ivec
+	vncipherlast	$out2,$out2,$in1
+	vncipherlast	$out3,$out3,$in2
+	vncipherlast	$out4,$out4,$in3
+	vncipherlast	$out5,$out5,$in4
+	vncipherlast	$out6,$out6,$in5
+	vncipherlast	$out7,$out7,$in6
+	vmr		$ivec,$in7
+
+	le?vperm	$out1,$out1,$out1,$inpperm
+	le?vperm	$out2,$out2,$out2,$inpperm
+	stvx_u		$out1,$x00,$out
+	le?vperm	$out3,$out3,$out3,$inpperm
+	stvx_u		$out2,$x10,$out
+	le?vperm	$out4,$out4,$out4,$inpperm
+	stvx_u		$out3,$x20,$out
+	le?vperm	$out5,$out5,$out5,$inpperm
+	stvx_u		$out4,$x30,$out
+	le?vperm	$out6,$out6,$out6,$inpperm
+	stvx_u		$out5,$x40,$out
+	le?vperm	$out7,$out7,$out7,$inpperm
+	stvx_u		$out6,$x50,$out
+	stvx_u		$out7,$x60,$out
+	addi		$out,$out,0x70
+	b		Lcbc_dec8x_done
+
+.align	5
+Lcbc_dec8x_six:
+	vncipherlast	$out2,$out2,$ivec
+	vncipherlast	$out3,$out3,$in2
+	vncipherlast	$out4,$out4,$in3
+	vncipherlast	$out5,$out5,$in4
+	vncipherlast	$out6,$out6,$in5
+	vncipherlast	$out7,$out7,$in6
+	vmr		$ivec,$in7
+
+	le?vperm	$out2,$out2,$out2,$inpperm
+	le?vperm	$out3,$out3,$out3,$inpperm
+	stvx_u		$out2,$x00,$out
+	le?vperm	$out4,$out4,$out4,$inpperm
+	stvx_u		$out3,$x10,$out
+	le?vperm	$out5,$out5,$out5,$inpperm
+	stvx_u		$out4,$x20,$out
+	le?vperm	$out6,$out6,$out6,$inpperm
+	stvx_u		$out5,$x30,$out
+	le?vperm	$out7,$out7,$out7,$inpperm
+	stvx_u		$out6,$x40,$out
+	stvx_u		$out7,$x50,$out
+	addi		$out,$out,0x60
+	b		Lcbc_dec8x_done
+
+.align	5
+Lcbc_dec8x_five:
+	vncipherlast	$out3,$out3,$ivec
+	vncipherlast	$out4,$out4,$in3
+	vncipherlast	$out5,$out5,$in4
+	vncipherlast	$out6,$out6,$in5
+	vncipherlast	$out7,$out7,$in6
+	vmr		$ivec,$in7
+
+	le?vperm	$out3,$out3,$out3,$inpperm
+	le?vperm	$out4,$out4,$out4,$inpperm
+	stvx_u		$out3,$x00,$out
+	le?vperm	$out5,$out5,$out5,$inpperm
+	stvx_u		$out4,$x10,$out
+	le?vperm	$out6,$out6,$out6,$inpperm
+	stvx_u		$out5,$x20,$out
+	le?vperm	$out7,$out7,$out7,$inpperm
+	stvx_u		$out6,$x30,$out
+	stvx_u		$out7,$x40,$out
+	addi		$out,$out,0x50
+	b		Lcbc_dec8x_done
+
+.align	5
+Lcbc_dec8x_four:
+	vncipherlast	$out4,$out4,$ivec
+	vncipherlast	$out5,$out5,$in4
+	vncipherlast	$out6,$out6,$in5
+	vncipherlast	$out7,$out7,$in6
+	vmr		$ivec,$in7
+
+	le?vperm	$out4,$out4,$out4,$inpperm
+	le?vperm	$out5,$out5,$out5,$inpperm
+	stvx_u		$out4,$x00,$out
+	le?vperm	$out6,$out6,$out6,$inpperm
+	stvx_u		$out5,$x10,$out
+	le?vperm	$out7,$out7,$out7,$inpperm
+	stvx_u		$out6,$x20,$out
+	stvx_u		$out7,$x30,$out
+	addi		$out,$out,0x40
+	b		Lcbc_dec8x_done
+
+.align	5
+Lcbc_dec8x_three:
+	vncipherlast	$out5,$out5,$ivec
+	vncipherlast	$out6,$out6,$in5
+	vncipherlast	$out7,$out7,$in6
+	vmr		$ivec,$in7
+
+	le?vperm	$out5,$out5,$out5,$inpperm
+	le?vperm	$out6,$out6,$out6,$inpperm
+	stvx_u		$out5,$x00,$out
+	le?vperm	$out7,$out7,$out7,$inpperm
+	stvx_u		$out6,$x10,$out
+	stvx_u		$out7,$x20,$out
+	addi		$out,$out,0x30
+	b		Lcbc_dec8x_done
+
+.align	5
+Lcbc_dec8x_two:
+	vncipherlast	$out6,$out6,$ivec
+	vncipherlast	$out7,$out7,$in6
+	vmr		$ivec,$in7
+
+	le?vperm	$out6,$out6,$out6,$inpperm
+	le?vperm	$out7,$out7,$out7,$inpperm
+	stvx_u		$out6,$x00,$out
+	stvx_u		$out7,$x10,$out
+	addi		$out,$out,0x20
+	b		Lcbc_dec8x_done
+
+.align	5
+Lcbc_dec8x_one:
+	vncipherlast	$out7,$out7,$ivec
+	vmr		$ivec,$in7
+
+	le?vperm	$out7,$out7,$out7,$inpperm
+	stvx_u		$out7,0,$out
+	addi		$out,$out,0x10
+
+Lcbc_dec8x_done:
+	le?vperm	$ivec,$ivec,$ivec,$inpperm
+	stvx_u		$ivec,0,$ivp		# write [unaligned] iv
+
+	li		r10,`$FRAME+15`
+	li		r11,`$FRAME+31`
+	stvx		$inpperm,r10,$sp	# wipe copies of round keys
+	addi		r10,r10,32
+	stvx		$inpperm,r11,$sp
+	addi		r11,r11,32
+	stvx		$inpperm,r10,$sp
+	addi		r10,r10,32
+	stvx		$inpperm,r11,$sp
+	addi		r11,r11,32
+	stvx		$inpperm,r10,$sp
+	addi		r10,r10,32
+	stvx		$inpperm,r11,$sp
+	addi		r11,r11,32
+	stvx		$inpperm,r10,$sp
+	addi		r10,r10,32
+	stvx		$inpperm,r11,$sp
+	addi		r11,r11,32
+
+	mtspr		256,$vrsave
+	lvx		v20,r10,$sp		# ABI says so
+	addi		r10,r10,32
+	lvx		v21,r11,$sp
+	addi		r11,r11,32
+	lvx		v22,r10,$sp
+	addi		r10,r10,32
+	lvx		v23,r11,$sp
+	addi		r11,r11,32
+	lvx		v24,r10,$sp
+	addi		r10,r10,32
+	lvx		v25,r11,$sp
+	addi		r11,r11,32
+	lvx		v26,r10,$sp
+	addi		r10,r10,32
+	lvx		v27,r11,$sp
+	addi		r11,r11,32
+	lvx		v28,r10,$sp
+	addi		r10,r10,32
+	lvx		v29,r11,$sp
+	addi		r11,r11,32
+	lvx		v30,r10,$sp
+	lvx		v31,r11,$sp
+	$POP		r26,`$FRAME+21*16+0*$SIZE_T`($sp)
+	$POP		r27,`$FRAME+21*16+1*$SIZE_T`($sp)
+	$POP		r28,`$FRAME+21*16+2*$SIZE_T`($sp)
+	$POP		r29,`$FRAME+21*16+3*$SIZE_T`($sp)
+	$POP		r30,`$FRAME+21*16+4*$SIZE_T`($sp)
+	$POP		r31,`$FRAME+21*16+5*$SIZE_T`($sp)
+	addi		$sp,$sp,`$FRAME+21*16+6*$SIZE_T`
+	blr
+	.long		0
+	.byte		0,12,0x14,0,0x80,6,6,0
+	.long		0
+.size	.${prefix}_cbc_encrypt,.-.${prefix}_cbc_encrypt
+___
+}}	}}}
+
+#########################################################################
+{{{	# CTR procedure[s]						#
+my ($inp,$out,$len,$key,$ivp,$x10,$rounds,$idx)=map("r$_",(3..10));
+my ($rndkey0,$rndkey1,$inout,$tmp)=		map("v$_",(0..3));
+my ($ivec,$inptail,$inpperm,$outhead,$outperm,$outmask,$keyperm,$one)=
+						map("v$_",(4..11));
+my $dat=$tmp;
+
+$code.=<<___;
+.globl	.${prefix}_ctr32_encrypt_blocks
+.align	5
+.${prefix}_ctr32_encrypt_blocks:
+	${UCMP}i	$len,1
+	bltlr-
+
+	lis		r0,0xfff0
+	mfspr		$vrsave,256
+	mtspr		256,r0
+
+	li		$idx,15
+	vxor		$rndkey0,$rndkey0,$rndkey0
+	le?vspltisb	$tmp,0x0f
+
+	lvx		$ivec,0,$ivp		# load [unaligned] iv
+	lvsl		$inpperm,0,$ivp
+	lvx		$inptail,$idx,$ivp
+	 vspltisb	$one,1
+	le?vxor		$inpperm,$inpperm,$tmp
+	vperm		$ivec,$ivec,$inptail,$inpperm
+	 vsldoi		$one,$rndkey0,$one,1
+
+	neg		r11,$inp
+	?lvsl		$keyperm,0,$key		# prepare for unaligned key
+	lwz		$rounds,240($key)
+
+	lvsr		$inpperm,0,r11		# prepare for unaligned load
+	lvx		$inptail,0,$inp
+	addi		$inp,$inp,15		# 15 is not typo
+	le?vxor		$inpperm,$inpperm,$tmp
+
+	srwi		$rounds,$rounds,1
+	li		$idx,16
+	subi		$rounds,$rounds,1
+
+	${UCMP}i	$len,8
+	bge		_aesp8_ctr32_encrypt8x
+
+	?lvsr		$outperm,0,$out		# prepare for unaligned store
+	vspltisb	$outmask,-1
+	lvx		$outhead,0,$out
+	?vperm		$outmask,$rndkey0,$outmask,$outperm
+	le?vxor		$outperm,$outperm,$tmp
+
+	lvx		$rndkey0,0,$key
+	mtctr		$rounds
+	lvx		$rndkey1,$idx,$key
+	addi		$idx,$idx,16
+	?vperm		$rndkey0,$rndkey0,$rndkey1,$keyperm
+	vxor		$inout,$ivec,$rndkey0
+	lvx		$rndkey0,$idx,$key
+	addi		$idx,$idx,16
+	b		Loop_ctr32_enc
+
+.align	5
+Loop_ctr32_enc:
+	?vperm		$rndkey1,$rndkey1,$rndkey0,$keyperm
+	vcipher		$inout,$inout,$rndkey1
+	lvx		$rndkey1,$idx,$key
+	addi		$idx,$idx,16
+	?vperm		$rndkey0,$rndkey0,$rndkey1,$keyperm
+	vcipher		$inout,$inout,$rndkey0
+	lvx		$rndkey0,$idx,$key
+	addi		$idx,$idx,16
+	bdnz		Loop_ctr32_enc
+
+	vadduwm		$ivec,$ivec,$one
+	 vmr		$dat,$inptail
+	 lvx		$inptail,0,$inp
+	 addi		$inp,$inp,16
+	 subic.		$len,$len,1		# blocks--
+
+	?vperm		$rndkey1,$rndkey1,$rndkey0,$keyperm
+	vcipher		$inout,$inout,$rndkey1
+	lvx		$rndkey1,$idx,$key
+	 vperm		$dat,$dat,$inptail,$inpperm
+	 li		$idx,16
+	?vperm		$rndkey1,$rndkey0,$rndkey1,$keyperm
+	 lvx		$rndkey0,0,$key
+	vxor		$dat,$dat,$rndkey1	# last round key
+	vcipherlast	$inout,$inout,$dat
+
+	 lvx		$rndkey1,$idx,$key
+	 addi		$idx,$idx,16
+	vperm		$inout,$inout,$inout,$outperm
+	vsel		$dat,$outhead,$inout,$outmask
+	 mtctr		$rounds
+	 ?vperm		$rndkey0,$rndkey0,$rndkey1,$keyperm
+	vmr		$outhead,$inout
+	 vxor		$inout,$ivec,$rndkey0
+	 lvx		$rndkey0,$idx,$key
+	 addi		$idx,$idx,16
+	stvx		$dat,0,$out
+	addi		$out,$out,16
+	bne		Loop_ctr32_enc
+
+	addi		$out,$out,-1
+	lvx		$inout,0,$out		# redundant in aligned case
+	vsel		$inout,$outhead,$inout,$outmask
+	stvx		$inout,0,$out
+
+	mtspr		256,$vrsave
+	blr
+	.long		0
+	.byte		0,12,0x14,0,0,0,6,0
+	.long		0
+___
+#########################################################################
+{{	# Optimized CTR procedure					#
+my $key_="r11";
+my ($x00,$x10,$x20,$x30,$x40,$x50,$x60,$x70)=map("r$_",(0,8,26..31));
+my ($in0, $in1, $in2, $in3, $in4, $in5, $in6, $in7 )=map("v$_",(0..3,10,12..14));
+my ($out0,$out1,$out2,$out3,$out4,$out5,$out6,$out7)=map("v$_",(15..22));
+my $rndkey0="v23";	# v24-v25 rotating buffer for first found keys
+			# v26-v31 last 6 round keys
+my ($tmp,$keyperm)=($in3,$in4);	# aliases with "caller", redundant assignment
+my ($two,$three,$four)=($outhead,$outperm,$outmask);
+
+$code.=<<___;
+.align	5
+_aesp8_ctr32_encrypt8x:
+	$STU		$sp,-`($FRAME+21*16+6*$SIZE_T)`($sp)
+	li		r10,`$FRAME+8*16+15`
+	li		r11,`$FRAME+8*16+31`
+	stvx		v20,r10,$sp		# ABI says so
+	addi		r10,r10,32
+	stvx		v21,r11,$sp
+	addi		r11,r11,32
+	stvx		v22,r10,$sp
+	addi		r10,r10,32
+	stvx		v23,r11,$sp
+	addi		r11,r11,32
+	stvx		v24,r10,$sp
+	addi		r10,r10,32
+	stvx		v25,r11,$sp
+	addi		r11,r11,32
+	stvx		v26,r10,$sp
+	addi		r10,r10,32
+	stvx		v27,r11,$sp
+	addi		r11,r11,32
+	stvx		v28,r10,$sp
+	addi		r10,r10,32
+	stvx		v29,r11,$sp
+	addi		r11,r11,32
+	stvx		v30,r10,$sp
+	stvx		v31,r11,$sp
+	li		r0,-1
+	stw		$vrsave,`$FRAME+21*16-4`($sp)	# save vrsave
+	li		$x10,0x10
+	$PUSH		r26,`$FRAME+21*16+0*$SIZE_T`($sp)
+	li		$x20,0x20
+	$PUSH		r27,`$FRAME+21*16+1*$SIZE_T`($sp)
+	li		$x30,0x30
+	$PUSH		r28,`$FRAME+21*16+2*$SIZE_T`($sp)
+	li		$x40,0x40
+	$PUSH		r29,`$FRAME+21*16+3*$SIZE_T`($sp)
+	li		$x50,0x50
+	$PUSH		r30,`$FRAME+21*16+4*$SIZE_T`($sp)
+	li		$x60,0x60
+	$PUSH		r31,`$FRAME+21*16+5*$SIZE_T`($sp)
+	li		$x70,0x70
+	mtspr		256,r0
+
+	subi		$rounds,$rounds,3	# -4 in total
+
+	lvx		$rndkey0,$x00,$key	# load key schedule
+	lvx		v30,$x10,$key
+	addi		$key,$key,0x20
+	lvx		v31,$x00,$key
+	?vperm		$rndkey0,$rndkey0,v30,$keyperm
+	addi		$key_,$sp,$FRAME+15
+	mtctr		$rounds
+
+Load_ctr32_enc_key:
+	?vperm		v24,v30,v31,$keyperm
+	lvx		v30,$x10,$key
+	addi		$key,$key,0x20
+	stvx		v24,$x00,$key_		# off-load round[1]
+	?vperm		v25,v31,v30,$keyperm
+	lvx		v31,$x00,$key
+	stvx		v25,$x10,$key_		# off-load round[2]
+	addi		$key_,$key_,0x20
+	bdnz		Load_ctr32_enc_key
+
+	lvx		v26,$x10,$key
+	?vperm		v24,v30,v31,$keyperm
+	lvx		v27,$x20,$key
+	stvx		v24,$x00,$key_		# off-load round[3]
+	?vperm		v25,v31,v26,$keyperm
+	lvx		v28,$x30,$key
+	stvx		v25,$x10,$key_		# off-load round[4]
+	addi		$key_,$sp,$FRAME+15	# rewind $key_
+	?vperm		v26,v26,v27,$keyperm
+	lvx		v29,$x40,$key
+	?vperm		v27,v27,v28,$keyperm
+	lvx		v30,$x50,$key
+	?vperm		v28,v28,v29,$keyperm
+	lvx		v31,$x60,$key
+	?vperm		v29,v29,v30,$keyperm
+	lvx		$out0,$x70,$key		# borrow $out0
+	?vperm		v30,v30,v31,$keyperm
+	lvx		v24,$x00,$key_		# pre-load round[1]
+	?vperm		v31,v31,$out0,$keyperm
+	lvx		v25,$x10,$key_		# pre-load round[2]
+
+	vadduwm		$two,$one,$one
+	subi		$inp,$inp,15		# undo "caller"
+	$SHL		$len,$len,4
+
+	vadduwm		$out1,$ivec,$one	# counter values ...
+	vadduwm		$out2,$ivec,$two
+	vxor		$out0,$ivec,$rndkey0	# ... xored with rndkey[0]
+	 le?li		$idx,8
+	vadduwm		$out3,$out1,$two
+	vxor		$out1,$out1,$rndkey0
+	 le?lvsl	$inpperm,0,$idx
+	vadduwm		$out4,$out2,$two
+	vxor		$out2,$out2,$rndkey0
+	 le?vspltisb	$tmp,0x0f
+	vadduwm		$out5,$out3,$two
+	vxor		$out3,$out3,$rndkey0
+	 le?vxor	$inpperm,$inpperm,$tmp	# transform for lvx_u/stvx_u
+	vadduwm		$out6,$out4,$two
+	vxor		$out4,$out4,$rndkey0
+	vadduwm		$out7,$out5,$two
+	vxor		$out5,$out5,$rndkey0
+	vadduwm		$ivec,$out6,$two	# next counter value
+	vxor		$out6,$out6,$rndkey0
+	vxor		$out7,$out7,$rndkey0
+
+	mtctr		$rounds
+	b		Loop_ctr32_enc8x
+.align	5
+Loop_ctr32_enc8x:
+	vcipher 	$out0,$out0,v24
+	vcipher 	$out1,$out1,v24
+	vcipher 	$out2,$out2,v24
+	vcipher 	$out3,$out3,v24
+	vcipher 	$out4,$out4,v24
+	vcipher 	$out5,$out5,v24
+	vcipher 	$out6,$out6,v24
+	vcipher 	$out7,$out7,v24
+Loop_ctr32_enc8x_middle:
+	lvx		v24,$x20,$key_		# round[3]
+	addi		$key_,$key_,0x20
+
+	vcipher 	$out0,$out0,v25
+	vcipher 	$out1,$out1,v25
+	vcipher 	$out2,$out2,v25
+	vcipher 	$out3,$out3,v25
+	vcipher 	$out4,$out4,v25
+	vcipher 	$out5,$out5,v25
+	vcipher 	$out6,$out6,v25
+	vcipher 	$out7,$out7,v25
+	lvx		v25,$x10,$key_		# round[4]
+	bdnz		Loop_ctr32_enc8x
+
+	subic		r11,$len,256		# $len-256, borrow $key_
+	vcipher 	$out0,$out0,v24
+	vcipher 	$out1,$out1,v24
+	vcipher 	$out2,$out2,v24
+	vcipher 	$out3,$out3,v24
+	vcipher 	$out4,$out4,v24
+	vcipher 	$out5,$out5,v24
+	vcipher 	$out6,$out6,v24
+	vcipher 	$out7,$out7,v24
+
+	subfe		r0,r0,r0		# borrow?-1:0
+	vcipher 	$out0,$out0,v25
+	vcipher 	$out1,$out1,v25
+	vcipher 	$out2,$out2,v25
+	vcipher 	$out3,$out3,v25
+	vcipher 	$out4,$out4,v25
+	vcipher		$out5,$out5,v25
+	vcipher		$out6,$out6,v25
+	vcipher		$out7,$out7,v25
+
+	and		r0,r0,r11
+	addi		$key_,$sp,$FRAME+15	# rewind $key_
+	vcipher		$out0,$out0,v26
+	vcipher		$out1,$out1,v26
+	vcipher		$out2,$out2,v26
+	vcipher		$out3,$out3,v26
+	vcipher		$out4,$out4,v26
+	vcipher		$out5,$out5,v26
+	vcipher		$out6,$out6,v26
+	vcipher		$out7,$out7,v26
+	lvx		v24,$x00,$key_		# re-pre-load round[1]
+
+	subic		$len,$len,129		# $len-=129
+	vcipher		$out0,$out0,v27
+	addi		$len,$len,1		# $len-=128 really
+	vcipher		$out1,$out1,v27
+	vcipher		$out2,$out2,v27
+	vcipher		$out3,$out3,v27
+	vcipher		$out4,$out4,v27
+	vcipher		$out5,$out5,v27
+	vcipher		$out6,$out6,v27
+	vcipher		$out7,$out7,v27
+	lvx		v25,$x10,$key_		# re-pre-load round[2]
+
+	vcipher		$out0,$out0,v28
+	 lvx_u		$in0,$x00,$inp		# load input
+	vcipher		$out1,$out1,v28
+	 lvx_u		$in1,$x10,$inp
+	vcipher		$out2,$out2,v28
+	 lvx_u		$in2,$x20,$inp
+	vcipher		$out3,$out3,v28
+	 lvx_u		$in3,$x30,$inp
+	vcipher		$out4,$out4,v28
+	 lvx_u		$in4,$x40,$inp
+	vcipher		$out5,$out5,v28
+	 lvx_u		$in5,$x50,$inp
+	vcipher		$out6,$out6,v28
+	 lvx_u		$in6,$x60,$inp
+	vcipher		$out7,$out7,v28
+	 lvx_u		$in7,$x70,$inp
+	 addi		$inp,$inp,0x80
+
+	vcipher		$out0,$out0,v29
+	 le?vperm	$in0,$in0,$in0,$inpperm
+	vcipher		$out1,$out1,v29
+	 le?vperm	$in1,$in1,$in1,$inpperm
+	vcipher		$out2,$out2,v29
+	 le?vperm	$in2,$in2,$in2,$inpperm
+	vcipher		$out3,$out3,v29
+	 le?vperm	$in3,$in3,$in3,$inpperm
+	vcipher		$out4,$out4,v29
+	 le?vperm	$in4,$in4,$in4,$inpperm
+	vcipher		$out5,$out5,v29
+	 le?vperm	$in5,$in5,$in5,$inpperm
+	vcipher		$out6,$out6,v29
+	 le?vperm	$in6,$in6,$in6,$inpperm
+	vcipher		$out7,$out7,v29
+	 le?vperm	$in7,$in7,$in7,$inpperm
+
+	add		$inp,$inp,r0		# $inp is adjusted in such
+						# way that at exit from the
+						# loop inX-in7 are loaded
+						# with last "words"
+	subfe.		r0,r0,r0		# borrow?-1:0
+	vcipher		$out0,$out0,v30
+	 vxor		$in0,$in0,v31		# xor with last round key
+	vcipher		$out1,$out1,v30
+	 vxor		$in1,$in1,v31
+	vcipher		$out2,$out2,v30
+	 vxor		$in2,$in2,v31
+	vcipher		$out3,$out3,v30
+	 vxor		$in3,$in3,v31
+	vcipher		$out4,$out4,v30
+	 vxor		$in4,$in4,v31
+	vcipher		$out5,$out5,v30
+	 vxor		$in5,$in5,v31
+	vcipher		$out6,$out6,v30
+	 vxor		$in6,$in6,v31
+	vcipher		$out7,$out7,v30
+	 vxor		$in7,$in7,v31
+
+	bne		Lctr32_enc8x_break	# did $len-129 borrow?
+
+	vcipherlast	$in0,$out0,$in0
+	vcipherlast	$in1,$out1,$in1
+	 vadduwm	$out1,$ivec,$one	# counter values ...
+	vcipherlast	$in2,$out2,$in2
+	 vadduwm	$out2,$ivec,$two
+	 vxor		$out0,$ivec,$rndkey0	# ... xored with rndkey[0]
+	vcipherlast	$in3,$out3,$in3
+	 vadduwm	$out3,$out1,$two
+	 vxor		$out1,$out1,$rndkey0
+	vcipherlast	$in4,$out4,$in4
+	 vadduwm	$out4,$out2,$two
+	 vxor		$out2,$out2,$rndkey0
+	vcipherlast	$in5,$out5,$in5
+	 vadduwm	$out5,$out3,$two
+	 vxor		$out3,$out3,$rndkey0
+	vcipherlast	$in6,$out6,$in6
+	 vadduwm	$out6,$out4,$two
+	 vxor		$out4,$out4,$rndkey0
+	vcipherlast	$in7,$out7,$in7
+	 vadduwm	$out7,$out5,$two
+	 vxor		$out5,$out5,$rndkey0
+	le?vperm	$in0,$in0,$in0,$inpperm
+	 vadduwm	$ivec,$out6,$two	# next counter value
+	 vxor		$out6,$out6,$rndkey0
+	le?vperm	$in1,$in1,$in1,$inpperm
+	 vxor		$out7,$out7,$rndkey0
+	mtctr		$rounds
+
+	 vcipher	$out0,$out0,v24
+	stvx_u		$in0,$x00,$out
+	le?vperm	$in2,$in2,$in2,$inpperm
+	 vcipher	$out1,$out1,v24
+	stvx_u		$in1,$x10,$out
+	le?vperm	$in3,$in3,$in3,$inpperm
+	 vcipher	$out2,$out2,v24
+	stvx_u		$in2,$x20,$out
+	le?vperm	$in4,$in4,$in4,$inpperm
+	 vcipher	$out3,$out3,v24
+	stvx_u		$in3,$x30,$out
+	le?vperm	$in5,$in5,$in5,$inpperm
+	 vcipher	$out4,$out4,v24
+	stvx_u		$in4,$x40,$out
+	le?vperm	$in6,$in6,$in6,$inpperm
+	 vcipher	$out5,$out5,v24
+	stvx_u		$in5,$x50,$out
+	le?vperm	$in7,$in7,$in7,$inpperm
+	 vcipher	$out6,$out6,v24
+	stvx_u		$in6,$x60,$out
+	 vcipher	$out7,$out7,v24
+	stvx_u		$in7,$x70,$out
+	addi		$out,$out,0x80
+
+	b		Loop_ctr32_enc8x_middle
+
+.align	5
+Lctr32_enc8x_break:
+	cmpwi		$len,-0x60
+	blt		Lctr32_enc8x_one
+	nop
+	beq		Lctr32_enc8x_two
+	cmpwi		$len,-0x40
+	blt		Lctr32_enc8x_three
+	nop
+	beq		Lctr32_enc8x_four
+	cmpwi		$len,-0x20
+	blt		Lctr32_enc8x_five
+	nop
+	beq		Lctr32_enc8x_six
+	cmpwi		$len,0x00
+	blt		Lctr32_enc8x_seven
+
+Lctr32_enc8x_eight:
+	vcipherlast	$out0,$out0,$in0
+	vcipherlast	$out1,$out1,$in1
+	vcipherlast	$out2,$out2,$in2
+	vcipherlast	$out3,$out3,$in3
+	vcipherlast	$out4,$out4,$in4
+	vcipherlast	$out5,$out5,$in5
+	vcipherlast	$out6,$out6,$in6
+	vcipherlast	$out7,$out7,$in7
+
+	le?vperm	$out0,$out0,$out0,$inpperm
+	le?vperm	$out1,$out1,$out1,$inpperm
+	stvx_u		$out0,$x00,$out
+	le?vperm	$out2,$out2,$out2,$inpperm
+	stvx_u		$out1,$x10,$out
+	le?vperm	$out3,$out3,$out3,$inpperm
+	stvx_u		$out2,$x20,$out
+	le?vperm	$out4,$out4,$out4,$inpperm
+	stvx_u		$out3,$x30,$out
+	le?vperm	$out5,$out5,$out5,$inpperm
+	stvx_u		$out4,$x40,$out
+	le?vperm	$out6,$out6,$out6,$inpperm
+	stvx_u		$out5,$x50,$out
+	le?vperm	$out7,$out7,$out7,$inpperm
+	stvx_u		$out6,$x60,$out
+	stvx_u		$out7,$x70,$out
+	addi		$out,$out,0x80
+	b		Lctr32_enc8x_done
+
+.align	5
+Lctr32_enc8x_seven:
+	vcipherlast	$out0,$out0,$in1
+	vcipherlast	$out1,$out1,$in2
+	vcipherlast	$out2,$out2,$in3
+	vcipherlast	$out3,$out3,$in4
+	vcipherlast	$out4,$out4,$in5
+	vcipherlast	$out5,$out5,$in6
+	vcipherlast	$out6,$out6,$in7
+
+	le?vperm	$out0,$out0,$out0,$inpperm
+	le?vperm	$out1,$out1,$out1,$inpperm
+	stvx_u		$out0,$x00,$out
+	le?vperm	$out2,$out2,$out2,$inpperm
+	stvx_u		$out1,$x10,$out
+	le?vperm	$out3,$out3,$out3,$inpperm
+	stvx_u		$out2,$x20,$out
+	le?vperm	$out4,$out4,$out4,$inpperm
+	stvx_u		$out3,$x30,$out
+	le?vperm	$out5,$out5,$out5,$inpperm
+	stvx_u		$out4,$x40,$out
+	le?vperm	$out6,$out6,$out6,$inpperm
+	stvx_u		$out5,$x50,$out
+	stvx_u		$out6,$x60,$out
+	addi		$out,$out,0x70
+	b		Lctr32_enc8x_done
+
+.align	5
+Lctr32_enc8x_six:
+	vcipherlast	$out0,$out0,$in2
+	vcipherlast	$out1,$out1,$in3
+	vcipherlast	$out2,$out2,$in4
+	vcipherlast	$out3,$out3,$in5
+	vcipherlast	$out4,$out4,$in6
+	vcipherlast	$out5,$out5,$in7
+
+	le?vperm	$out0,$out0,$out0,$inpperm
+	le?vperm	$out1,$out1,$out1,$inpperm
+	stvx_u		$out0,$x00,$out
+	le?vperm	$out2,$out2,$out2,$inpperm
+	stvx_u		$out1,$x10,$out
+	le?vperm	$out3,$out3,$out3,$inpperm
+	stvx_u		$out2,$x20,$out
+	le?vperm	$out4,$out4,$out4,$inpperm
+	stvx_u		$out3,$x30,$out
+	le?vperm	$out5,$out5,$out5,$inpperm
+	stvx_u		$out4,$x40,$out
+	stvx_u		$out5,$x50,$out
+	addi		$out,$out,0x60
+	b		Lctr32_enc8x_done
+
+.align	5
+Lctr32_enc8x_five:
+	vcipherlast	$out0,$out0,$in3
+	vcipherlast	$out1,$out1,$in4
+	vcipherlast	$out2,$out2,$in5
+	vcipherlast	$out3,$out3,$in6
+	vcipherlast	$out4,$out4,$in7
+
+	le?vperm	$out0,$out0,$out0,$inpperm
+	le?vperm	$out1,$out1,$out1,$inpperm
+	stvx_u		$out0,$x00,$out
+	le?vperm	$out2,$out2,$out2,$inpperm
+	stvx_u		$out1,$x10,$out
+	le?vperm	$out3,$out3,$out3,$inpperm
+	stvx_u		$out2,$x20,$out
+	le?vperm	$out4,$out4,$out4,$inpperm
+	stvx_u		$out3,$x30,$out
+	stvx_u		$out4,$x40,$out
+	addi		$out,$out,0x50
+	b		Lctr32_enc8x_done
+
+.align	5
+Lctr32_enc8x_four:
+	vcipherlast	$out0,$out0,$in4
+	vcipherlast	$out1,$out1,$in5
+	vcipherlast	$out2,$out2,$in6
+	vcipherlast	$out3,$out3,$in7
+
+	le?vperm	$out0,$out0,$out0,$inpperm
+	le?vperm	$out1,$out1,$out1,$inpperm
+	stvx_u		$out0,$x00,$out
+	le?vperm	$out2,$out2,$out2,$inpperm
+	stvx_u		$out1,$x10,$out
+	le?vperm	$out3,$out3,$out3,$inpperm
+	stvx_u		$out2,$x20,$out
+	stvx_u		$out3,$x30,$out
+	addi		$out,$out,0x40
+	b		Lctr32_enc8x_done
+
+.align	5
+Lctr32_enc8x_three:
+	vcipherlast	$out0,$out0,$in5
+	vcipherlast	$out1,$out1,$in6
+	vcipherlast	$out2,$out2,$in7
+
+	le?vperm	$out0,$out0,$out0,$inpperm
+	le?vperm	$out1,$out1,$out1,$inpperm
+	stvx_u		$out0,$x00,$out
+	le?vperm	$out2,$out2,$out2,$inpperm
+	stvx_u		$out1,$x10,$out
+	stvx_u		$out2,$x20,$out
+	addi		$out,$out,0x30
+	b		Lcbc_dec8x_done
+
+.align	5
+Lctr32_enc8x_two:
+	vcipherlast	$out0,$out0,$in6
+	vcipherlast	$out1,$out1,$in7
+
+	le?vperm	$out0,$out0,$out0,$inpperm
+	le?vperm	$out1,$out1,$out1,$inpperm
+	stvx_u		$out0,$x00,$out
+	stvx_u		$out1,$x10,$out
+	addi		$out,$out,0x20
+	b		Lcbc_dec8x_done
+
+.align	5
+Lctr32_enc8x_one:
+	vcipherlast	$out0,$out0,$in7
+
+	le?vperm	$out0,$out0,$out0,$inpperm
+	stvx_u		$out0,0,$out
+	addi		$out,$out,0x10
+
+Lctr32_enc8x_done:
+	li		r10,`$FRAME+15`
+	li		r11,`$FRAME+31`
+	stvx		$inpperm,r10,$sp	# wipe copies of round keys
+	addi		r10,r10,32
+	stvx		$inpperm,r11,$sp
+	addi		r11,r11,32
+	stvx		$inpperm,r10,$sp
+	addi		r10,r10,32
+	stvx		$inpperm,r11,$sp
+	addi		r11,r11,32
+	stvx		$inpperm,r10,$sp
+	addi		r10,r10,32
+	stvx		$inpperm,r11,$sp
+	addi		r11,r11,32
+	stvx		$inpperm,r10,$sp
+	addi		r10,r10,32
+	stvx		$inpperm,r11,$sp
+	addi		r11,r11,32
+
+	mtspr		256,$vrsave
+	lvx		v20,r10,$sp		# ABI says so
+	addi		r10,r10,32
+	lvx		v21,r11,$sp
+	addi		r11,r11,32
+	lvx		v22,r10,$sp
+	addi		r10,r10,32
+	lvx		v23,r11,$sp
+	addi		r11,r11,32
+	lvx		v24,r10,$sp
+	addi		r10,r10,32
+	lvx		v25,r11,$sp
+	addi		r11,r11,32
+	lvx		v26,r10,$sp
+	addi		r10,r10,32
+	lvx		v27,r11,$sp
+	addi		r11,r11,32
+	lvx		v28,r10,$sp
+	addi		r10,r10,32
+	lvx		v29,r11,$sp
+	addi		r11,r11,32
+	lvx		v30,r10,$sp
+	lvx		v31,r11,$sp
+	$POP		r26,`$FRAME+21*16+0*$SIZE_T`($sp)
+	$POP		r27,`$FRAME+21*16+1*$SIZE_T`($sp)
+	$POP		r28,`$FRAME+21*16+2*$SIZE_T`($sp)
+	$POP		r29,`$FRAME+21*16+3*$SIZE_T`($sp)
+	$POP		r30,`$FRAME+21*16+4*$SIZE_T`($sp)
+	$POP		r31,`$FRAME+21*16+5*$SIZE_T`($sp)
+	addi		$sp,$sp,`$FRAME+21*16+6*$SIZE_T`
+	blr
+	.long		0
+	.byte		0,12,0x14,0,0x80,6,6,0
+	.long		0
+.size	.${prefix}_ctr32_encrypt_blocks,.-.${prefix}_ctr32_encrypt_blocks
+___
+}}	}}}
+
+my $consts=1;
+foreach(split("\n",$code)) {
+        s/\`([^\`]*)\`/eval($1)/geo;
+
+	# constants table endian-specific conversion
+	if ($consts && m/\.(long|byte)\s+(.+)\s+(\?[a-z]*)$/o) {
+	    my $conv=$3;
+	    my @bytes=();
+
+	    # convert to endian-agnostic format
+	    if ($1 eq "long") {
+	      foreach (split(/,\s*/,$2)) {
+		my $l = /^0/?oct:int;
+		push @bytes,($l>>24)&0xff,($l>>16)&0xff,($l>>8)&0xff,$l&0xff;
+	      }
+	    } else {
+		@bytes = map(/^0/?oct:int,split(/,\s*/,$2));
+	    }
+
+	    # little-endian conversion
+	    if ($flavour =~ /le$/o) {
+		SWITCH: for($conv)  {
+		    /\?inv/ && do   { @bytes=map($_^0xf,@bytes); last; };
+		    /\?rev/ && do   { @bytes=reverse(@bytes);    last; }; 
+		}
+	    }
+
+	    #emit
+	    print ".byte\t",join(',',map (sprintf("0x%02x",$_),@bytes)),"\n";
+	    next;
+	}
+	$consts=0 if (m/Lconsts:/o);	# end of table
+
+	# instructions prefixed with '?' are endian-specific and need
+	# to be adjusted accordingly...
+	if ($flavour =~ /le$/o) {	# little-endian
+	    s/le\?//o		or
+	    s/be\?/#be#/o	or
+	    s/\?lvsr/lvsl/o	or
+	    s/\?lvsl/lvsr/o	or
+	    s/\?(vperm\s+v[0-9]+,\s*)(v[0-9]+,\s*)(v[0-9]+,\s*)(v[0-9]+)/$1$3$2$4/o or
+	    s/\?(vsldoi\s+v[0-9]+,\s*)(v[0-9]+,)\s*(v[0-9]+,\s*)([0-9]+)/$1$3$2 16-$4/o or
+	    s/\?(vspltw\s+v[0-9]+,\s*)(v[0-9]+,)\s*([0-9])/$1$2 3-$3/o;
+	} else {			# big-endian
+	    s/le\?/#le#/o	or
+	    s/be\?//o		or
+	    s/\?([a-z]+)/$1/o;
+	}
+
+        print $_,"\n";
+}
+
+close STDOUT;
diff --git a/crypto/aes/asm/vpaes-ppc.pl b/crypto/aes/asm/vpaes-ppc.pl
new file mode 100644
index 0000000..7fda60e
--- /dev/null
+++ b/crypto/aes/asm/vpaes-ppc.pl
@@ -0,0 +1,1512 @@
+#!/usr/bin/env perl
+
+######################################################################
+## Constant-time SSSE3 AES core implementation.
+## version 0.1
+##
+## By Mike Hamburg (Stanford University), 2009
+## Public domain.
+##
+## For details see http://shiftleft.org/papers/vector_aes/ and
+## http://crypto.stanford.edu/vpaes/.
+
+# CBC encrypt/decrypt performance in cycles per byte processed with
+# 128-bit key.
+#
+#		aes-ppc.pl		this
+# G4e		35.5/52.1/(23.8)	11.9(*)/15.4
+# POWER6	42.7/54.3/(28.2)	63.0/92.8(**)
+# POWER7	32.3/42.9/(18.4)	18.5/23.3
+#
+# (*)	This is ~10% worse than reported in paper. The reason is
+#	twofold. This module doesn't make any assumption about
+#	key schedule (or data for that matter) alignment and handles
+#	it in-line. Secondly it, being transliterated from
+#	vpaes-x86_64.pl, relies on "nested inversion" better suited
+#	for Intel CPUs.
+# (**)	Inadequate POWER6 performance is due to astronomic AltiVec
+#	latency, 9 cycles per simple logical operation.
+
+$flavour = shift;
+
+if ($flavour =~ /64/) {
+	$SIZE_T	=8;
+	$LRSAVE	=2*$SIZE_T;
+	$STU	="stdu";
+	$POP	="ld";
+	$PUSH	="std";
+	$UCMP	="cmpld";
+} elsif ($flavour =~ /32/) {
+	$SIZE_T	=4;
+	$LRSAVE	=$SIZE_T;
+	$STU	="stwu";
+	$POP	="lwz";
+	$PUSH	="stw";
+	$UCMP	="cmplw";
+} else { die "nonsense $flavour"; }
+
+$sp="r1";
+$FRAME=6*$SIZE_T+13*16;	# 13*16 is for v20-v31 offload
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or
+( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or
+die "can't locate ppc-xlate.pl";
+
+open STDOUT,"| $^X $xlate $flavour ".shift || die "can't call $xlate: $!";
+
+$code.=<<___;
+.machine	"any"
+
+.text
+
+.align	7	# totally strategic alignment
+_vpaes_consts:
+Lk_mc_forward:	# mc_forward
+	.long	0x01020300, 0x05060704, 0x090a0b08, 0x0d0e0f0c	?inv
+	.long	0x05060704, 0x090a0b08, 0x0d0e0f0c, 0x01020300	?inv
+	.long	0x090a0b08, 0x0d0e0f0c, 0x01020300, 0x05060704	?inv
+	.long	0x0d0e0f0c, 0x01020300, 0x05060704, 0x090a0b08	?inv
+Lk_mc_backward:	# mc_backward
+	.long	0x03000102, 0x07040506, 0x0b08090a, 0x0f0c0d0e	?inv
+	.long	0x0f0c0d0e, 0x03000102, 0x07040506, 0x0b08090a	?inv
+	.long	0x0b08090a, 0x0f0c0d0e, 0x03000102, 0x07040506	?inv
+	.long	0x07040506, 0x0b08090a, 0x0f0c0d0e, 0x03000102	?inv
+Lk_sr:		# sr
+	.long	0x00010203, 0x04050607, 0x08090a0b, 0x0c0d0e0f	?inv
+	.long	0x00050a0f, 0x04090e03, 0x080d0207, 0x0c01060b	?inv
+	.long	0x0009020b, 0x040d060f, 0x08010a03, 0x0c050e07	?inv
+	.long	0x000d0a07, 0x04010e0b, 0x0805020f, 0x0c090603	?inv
+
+##
+## "Hot" constants
+##
+Lk_inv:		# inv, inva
+	.long	0xf001080d, 0x0f06050e, 0x020c0b0a, 0x09030704	?rev
+	.long	0xf0070b0f, 0x060a0401, 0x09080502, 0x0c0e0d03	?rev
+Lk_ipt:		# input transform (lo, hi)
+	.long	0x00702a5a, 0x98e8b2c2, 0x08782252, 0x90e0baca	?rev
+	.long	0x004d7c31, 0x7d30014c, 0x81ccfdb0, 0xfcb180cd	?rev
+Lk_sbo:		# sbou, sbot
+	.long	0x00c7bd6f, 0x176dd2d0, 0x78a802c5, 0x7abfaa15	?rev
+	.long	0x006abb5f, 0xa574e4cf, 0xfa352b41, 0xd1901e8e	?rev
+Lk_sb1:		# sb1u, sb1t
+	.long	0x0023e2fa, 0x15d41836, 0xefd92e0d, 0xc1ccf73b	?rev
+	.long	0x003e50cb, 0x8fe19bb1, 0x44f52a14, 0x6e7adfa5	?rev
+Lk_sb2:		# sb2u, sb2t
+	.long	0x0029e10a, 0x4088eb69, 0x4a2382ab, 0xc863a1c2	?rev
+	.long	0x0024710b, 0xc6937ae2, 0xcd2f98bc, 0x55e9b75e	?rev
+
+##
+##  Decryption stuff
+##
+Lk_dipt:	# decryption input transform
+	.long	0x005f540b, 0x045b500f, 0x1a454e11, 0x1e414a15	?rev
+	.long	0x00650560, 0xe683e386, 0x94f191f4, 0x72177712	?rev
+Lk_dsbo:	# decryption sbox final output
+	.long	0x0040f97e, 0x53ea8713, 0x2d3e94d4, 0xb96daac7	?rev
+	.long	0x001d4493, 0x0f56d712, 0x9c8ec5d8, 0x59814bca	?rev
+Lk_dsb9:	# decryption sbox output *9*u, *9*t
+	.long	0x00d6869a, 0x53031c85, 0xc94c994f, 0x501fd5ca	?rev
+	.long	0x0049d7ec, 0x89173bc0, 0x65a5fbb2, 0x9e2c5e72	?rev
+Lk_dsbd:	# decryption sbox output *D*u, *D*t
+	.long	0x00a2b1e6, 0xdfcc577d, 0x39442a88, 0x139b6ef5	?rev
+	.long	0x00cbc624, 0xf7fae23c, 0xd3efde15, 0x0d183129	?rev
+Lk_dsbb:	# decryption sbox output *B*u, *B*t
+	.long	0x0042b496, 0x926422d0, 0x04d4f2b0, 0xf6462660	?rev
+	.long	0x006759cd, 0xa69894c1, 0x6baa5532, 0x3e0cfff3	?rev
+Lk_dsbe:	# decryption sbox output *E*u, *E*t
+	.long	0x00d0d426, 0x9692f246, 0xb0f6b464, 0x04604222	?rev
+	.long	0x00c1aaff, 0xcda6550c, 0x323e5998, 0x6bf36794	?rev
+
+##
+##  Key schedule constants
+##
+Lk_dksd:	# decryption key schedule: invskew x*D
+	.long	0x0047e4a3, 0x5d1ab9fe, 0xf9be1d5a, 0xa4e34007	?rev
+	.long	0x008336b5, 0xf477c241, 0x1e9d28ab, 0xea69dc5f	?rev
+Lk_dksb:	# decryption key schedule: invskew x*B
+	.long	0x00d55085, 0x1fca4f9a, 0x994cc91c, 0x8653d603	?rev
+	.long	0x004afcb6, 0xa7ed5b11, 0xc882347e, 0x6f2593d9	?rev
+Lk_dkse:	# decryption key schedule: invskew x*E + 0x63
+	.long	0x00d6c91f, 0xca1c03d5, 0x86504f99, 0x4c9a8553	?rev
+	.long	0xe87bdc4f, 0x059631a2, 0x8714b320, 0x6af95ecd	?rev
+Lk_dks9:	# decryption key schedule: invskew x*9
+	.long	0x00a7d97e, 0xc86f11b6, 0xfc5b2582, 0x3493ed4a	?rev
+	.long	0x00331427, 0x62517645, 0xcefddae9, 0xac9fb88b	?rev
+
+Lk_rcon:	# rcon
+	.long	0xb6ee9daf, 0xb991831f, 0x817d7c4d, 0x08982a70	?asis
+Lk_s63:
+	.long	0x5b5b5b5b, 0x5b5b5b5b, 0x5b5b5b5b, 0x5b5b5b5b	?asis
+
+Lk_opt:		# output transform
+	.long	0x0060b6d6, 0x29499fff, 0x0868bede, 0x214197f7	?rev
+	.long	0x00ecbc50, 0x51bded01, 0xe00c5cb0, 0xb15d0de1	?rev
+Lk_deskew:	# deskew tables: inverts the sbox's "skew"
+	.long	0x00e3a447, 0x40a3e407, 0x1af9be5d, 0x5ab9fe1d	?rev
+	.long	0x0069ea83, 0xdcb5365f, 0x771e9df4, 0xabc24128	?rev
+.align	5
+Lconsts:
+	mflr	r0
+	bcl	20,31,\$+4
+	mflr	r12	#vvvvv "distance between . and _vpaes_consts
+	addi	r12,r12,-0x308
+	mtlr	r0
+	blr
+	.long	0
+	.byte	0,12,0x14,0,0,0,0,0
+.asciz  "Vector Permutation AES for AltiVec, Mike Hamburg (Stanford University)"
+.align	6
+___
+
+my ($inptail,$inpperm,$outhead,$outperm,$outmask,$keyperm) = map("v$_",(26..31));
+{
+my ($inp,$out,$key) = map("r$_",(3..5));
+
+my ($invlo,$invhi,$iptlo,$ipthi,$sbou,$sbot) = map("v$_",(10..15));
+my ($sb1u,$sb1t,$sb2u,$sb2t) = map("v$_",(16..19));
+my ($sb9u,$sb9t,$sbdu,$sbdt,$sbbu,$sbbt,$sbeu,$sbet)=map("v$_",(16..23));
+
+$code.=<<___;
+##
+##  _aes_preheat
+##
+##  Fills register %r10 -> .aes_consts (so you can -fPIC)
+##  and %xmm9-%xmm15 as specified below.
+##
+.align	4
+_vpaes_encrypt_preheat:
+	mflr	r8
+	bl	Lconsts
+	mtlr	r8
+	li	r11, 0xc0		# Lk_inv
+	li	r10, 0xd0
+	li	r9,  0xe0		# Lk_ipt
+	li	r8,  0xf0
+	vxor	v7, v7, v7		# 0x00..00
+	vspltisb	v8,4		# 0x04..04
+	vspltisb	v9,0x0f		# 0x0f..0f
+	lvx	$invlo, r12, r11
+	li	r11, 0x100
+	lvx	$invhi, r12, r10
+	li	r10, 0x110
+	lvx	$iptlo, r12, r9
+	li	r9,  0x120
+	lvx	$ipthi, r12, r8
+	li	r8,  0x130
+	lvx	$sbou, r12, r11
+	li	r11, 0x140
+	lvx	$sbot, r12, r10
+	li	r10, 0x150
+	lvx	$sb1u, r12, r9
+	lvx	$sb1t, r12, r8
+	lvx	$sb2u, r12, r11
+	lvx	$sb2t, r12, r10
+	blr
+	.long	0
+	.byte	0,12,0x14,0,0,0,0,0
+
+##
+##  _aes_encrypt_core
+##
+##  AES-encrypt %xmm0.
+##
+##  Inputs:
+##     %xmm0 = input
+##     %xmm9-%xmm15 as in _vpaes_preheat
+##    (%rdx) = scheduled keys
+##
+##  Output in %xmm0
+##  Clobbers  %xmm1-%xmm6, %r9, %r10, %r11, %rax
+##
+##
+.align 5
+_vpaes_encrypt_core:
+	lwz	r8, 240($key)		# pull rounds
+	li	r9, 16
+	lvx	v5, 0, $key		# vmovdqu	(%r9),	%xmm5		# round0 key
+	li	r11, 0x10
+	lvx	v6, r9, $key
+	addi	r9, r9, 16
+	?vperm	v5, v5, v6, $keyperm	# align round key
+	addi	r10, r11, 0x40
+	vsrb	v1, v0, v8		# vpsrlb	\$4,	%xmm0,	%xmm0
+	vperm	v0, $iptlo, $iptlo, v0	# vpshufb	%xmm1,	%xmm2,	%xmm1
+	vperm	v1, $ipthi, $ipthi, v1	# vpshufb	%xmm0,	%xmm3,	%xmm2
+	vxor	v0, v0, v5		# vpxor	%xmm5,	%xmm1,	%xmm0
+	vxor	v0, v0, v1		# vpxor	%xmm2,	%xmm0,	%xmm0
+	mtctr	r8
+	b	Lenc_entry
+
+.align 4
+Lenc_loop:
+	# middle of middle round
+	vperm	v4, $sb1t, v7, v2	# vpshufb	%xmm2,	%xmm13,	%xmm4	# 4 = sb1u
+	lvx	v1, r12, r11		# vmovdqa	-0x40(%r11,%r10), %xmm1	# .Lk_mc_forward[]
+	addi	r11, r11, 16
+	vperm	v0, $sb1u, v7, v3	# vpshufb	%xmm3,	%xmm12,	%xmm0	# 0 = sb1t
+	vxor	v4, v4, v5		# vpxor		%xmm5,	%xmm4,	%xmm4	# 4 = sb1u + k
+	andi.	r11, r11, 0x30		# and		\$0x30, %r11	# ... mod 4
+	vperm	v5, $sb2t, v7, v2	# vpshufb	%xmm2,	%xmm15,	%xmm5	# 4 = sb2u
+	vxor	v0, v0, v4		# vpxor		%xmm4,	%xmm0,	%xmm0	# 0 = A
+	vperm	v2, $sb2u, v7, v3	# vpshufb	%xmm3,	%xmm14,	%xmm2	# 2 = sb2t
+	lvx	v4, r12, r10		# vmovdqa	(%r11,%r10), %xmm4	# .Lk_mc_backward[]
+	addi	r10, r11, 0x40
+	vperm	v3, v0, v7, v1		# vpshufb	%xmm1,	%xmm0,	%xmm3	# 0 = B
+	vxor	v2, v2, v5		# vpxor		%xmm5,	%xmm2,	%xmm2	# 2 = 2A
+	vperm	v0, v0, v7, v4		# vpshufb	%xmm4,	%xmm0,	%xmm0	# 3 = D
+	vxor	v3, v3, v2		# vpxor		%xmm2,	%xmm3,	%xmm3	# 0 = 2A+B
+	vperm	v4, v3, v7, v1		# vpshufb	%xmm1,	%xmm3,	%xmm4	# 0 = 2B+C
+	vxor	v0, v0, v3		# vpxor		%xmm3,	%xmm0,	%xmm0	# 3 = 2A+B+D
+	vxor	v0, v0, v4		# vpxor		%xmm4,	%xmm0, %xmm0	# 0 = 2A+3B+C+D
+
+Lenc_entry:
+	# top of round
+	vsrb	v1, v0, v8		# vpsrlb	\$4,	%xmm0,	%xmm0	# 1 = i
+	vperm	v5, $invhi, $invhi, v0	# vpshufb	%xmm1,	%xmm11,	%xmm5	# 2 = a/k
+	vxor	v0, v0, v1		# vpxor		%xmm0,	%xmm1,	%xmm1	# 0 = j
+	vperm	v3, $invlo, $invlo, v1	# vpshufb	%xmm0, 	%xmm10,	%xmm3  	# 3 = 1/i
+	vperm	v4, $invlo, $invlo, v0	# vpshufb	%xmm1, 	%xmm10,	%xmm4  	# 4 = 1/j
+	vand	v0, v0, v9
+	vxor	v3, v3, v5		# vpxor		%xmm5,	%xmm3,	%xmm3	# 3 = iak = 1/i + a/k
+	vxor	v4, v4, v5		# vpxor		%xmm5,	%xmm4,	%xmm4  	# 4 = jak = 1/j + a/k
+	vperm	v2, $invlo, v7, v3	# vpshufb	%xmm3,	%xmm10,	%xmm2  	# 2 = 1/iak
+	vmr	v5, v6
+	lvx	v6, r9, $key		# vmovdqu	(%r9), %xmm5
+	vperm	v3, $invlo, v7, v4	# vpshufb	%xmm4,	%xmm10,	%xmm3	# 3 = 1/jak
+	addi	r9, r9, 16
+	vxor	v2, v2, v0		# vpxor		%xmm1,	%xmm2,	%xmm2  	# 2 = io
+	?vperm	v5, v5, v6, $keyperm	# align round key
+	vxor	v3, v3, v1		# vpxor		%xmm0,	%xmm3,	%xmm3	# 3 = jo
+	bdnz	Lenc_loop
+
+	# middle of last round
+	addi	r10, r11, 0x80
+					# vmovdqa	-0x60(%r10), %xmm4	# 3 : sbou	.Lk_sbo
+					# vmovdqa	-0x50(%r10), %xmm0	# 0 : sbot	.Lk_sbo+16
+	vperm	v4, $sbou, v7, v2	# vpshufb	%xmm2,	%xmm4,	%xmm4	# 4 = sbou
+	lvx	v1, r12, r10		# vmovdqa	0x40(%r11,%r10), %xmm1	# .Lk_sr[]
+	vperm	v0, $sbot, v7, v3	# vpshufb	%xmm3,	%xmm0,	%xmm0	# 0 = sb1t
+	vxor	v4, v4, v5		# vpxor		%xmm5,	%xmm4,	%xmm4	# 4 = sb1u + k
+	vxor	v0, v0, v4		# vpxor		%xmm4,	%xmm0,	%xmm0	# 0 = A
+	vperm	v0, v0, v7, v1		# vpshufb	%xmm1,	%xmm0,	%xmm0
+	blr
+	.long	0
+	.byte	0,12,0x14,0,0,0,0,0
+
+.globl	.vpaes_encrypt
+.align	5
+.vpaes_encrypt:
+	$STU	$sp,-$FRAME($sp)
+	li	r10,`15+6*$SIZE_T`
+	li	r11,`31+6*$SIZE_T`
+	mflr	r6
+	mfspr	r7, 256			# save vrsave
+	stvx	v20,r10,$sp
+	addi	r10,r10,32
+	stvx	v21,r11,$sp
+	addi	r11,r11,32
+	stvx	v22,r10,$sp
+	addi	r10,r10,32
+	stvx	v23,r11,$sp
+	addi	r11,r11,32
+	stvx	v24,r10,$sp
+	addi	r10,r10,32
+	stvx	v25,r11,$sp
+	addi	r11,r11,32
+	stvx	v26,r10,$sp
+	addi	r10,r10,32
+	stvx	v27,r11,$sp
+	addi	r11,r11,32
+	stvx	v28,r10,$sp
+	addi	r10,r10,32
+	stvx	v29,r11,$sp
+	addi	r11,r11,32
+	stvx	v30,r10,$sp
+	stvx	v31,r11,$sp
+	stw	r7,`$FRAME-4`($sp)	# save vrsave
+	li	r0, -1
+	$PUSH	r6,`$FRAME+$LRSAVE`($sp)
+	mtspr	256, r0			# preserve all AltiVec registers
+
+	bl	_vpaes_encrypt_preheat
+
+	?lvsl	$inpperm, 0, $inp	# prepare for unaligned access
+	lvx	v0, 0, $inp
+	addi	$inp, $inp, 15		# 15 is not a typo
+	 ?lvsr	$outperm, 0, $out
+	?lvsl	$keyperm, 0, $key	# prepare for unaligned access
+	 vnor	$outmask, v7, v7	# 0xff..ff
+	lvx	$inptail, 0, $inp	# redundant in aligned case
+	 ?vperm	$outmask, v7, $outmask, $outperm
+	 lvx	$outhead, 0, $out
+	?vperm	v0, v0, $inptail, $inpperm
+
+	bl	_vpaes_encrypt_core
+
+	vperm	v0, v0, v0, $outperm	# rotate right/left
+	vsel	v1, $outhead, v0, $outmask
+	vmr	$outhead, v0
+	stvx	v1, 0, $out
+	addi	$out, $out, 15		# 15 is not a typo
+	########
+
+	lvx	v1, 0, $out		# redundant in aligned case
+	vsel	v1, $outhead, v1, $outmask
+	stvx	v1, 0, $out
+
+	li	r10,`15+6*$SIZE_T`
+	li	r11,`31+6*$SIZE_T`
+	mtlr	r6
+	mtspr	256, r7			# restore vrsave
+	lvx	v20,r10,$sp
+	addi	r10,r10,32
+	lvx	v21,r11,$sp
+	addi	r11,r11,32
+	lvx	v22,r10,$sp
+	addi	r10,r10,32
+	lvx	v23,r11,$sp
+	addi	r11,r11,32
+	lvx	v24,r10,$sp
+	addi	r10,r10,32
+	lvx	v25,r11,$sp
+	addi	r11,r11,32
+	lvx	v26,r10,$sp
+	addi	r10,r10,32
+	lvx	v27,r11,$sp
+	addi	r11,r11,32
+	lvx	v28,r10,$sp
+	addi	r10,r10,32
+	lvx	v29,r11,$sp
+	addi	r11,r11,32
+	lvx	v30,r10,$sp
+	lvx	v31,r11,$sp
+	addi	$sp,$sp,$FRAME
+	blr
+	.long	0
+	.byte	0,12,0x04,1,0x80,0,3,0
+	.long	0
+.size	.vpaes_encrypt,.-.vpaes_encrypt
+
+.align	4
+_vpaes_decrypt_preheat:
+	mflr	r8
+	bl	Lconsts
+	mtlr	r8
+	li	r11, 0xc0		# Lk_inv
+	li	r10, 0xd0
+	li	r9,  0x160		# Ldipt
+	li	r8,  0x170
+	vxor	v7, v7, v7		# 0x00..00
+	vspltisb	v8,4		# 0x04..04
+	vspltisb	v9,0x0f		# 0x0f..0f
+	lvx	$invlo, r12, r11
+	li	r11, 0x180
+	lvx	$invhi, r12, r10
+	li	r10, 0x190
+	lvx	$iptlo, r12, r9
+	li	r9,  0x1a0
+	lvx	$ipthi, r12, r8
+	li	r8,  0x1b0
+	lvx	$sbou, r12, r11
+	li	r11, 0x1c0
+	lvx	$sbot, r12, r10
+	li	r10, 0x1d0
+	lvx	$sb9u, r12, r9
+	li	r9,  0x1e0
+	lvx	$sb9t, r12, r8
+	li	r8,  0x1f0
+	lvx	$sbdu, r12, r11
+	li	r11, 0x200
+	lvx	$sbdt, r12, r10
+	li	r10, 0x210
+	lvx	$sbbu, r12, r9
+	lvx	$sbbt, r12, r8
+	lvx	$sbeu, r12, r11
+	lvx	$sbet, r12, r10
+	blr
+	.long	0
+	.byte	0,12,0x14,0,0,0,0,0
+
+##
+##  Decryption core
+##
+##  Same API as encryption core.
+##
+.align	4
+_vpaes_decrypt_core:
+	lwz	r8, 240($key)		# pull rounds
+	li	r9, 16
+	lvx	v5, 0, $key		# vmovdqu	(%r9),	%xmm4		# round0 key
+	li	r11, 0x30
+	lvx	v6, r9, $key
+	addi	r9, r9, 16
+	?vperm	v5, v5, v6, $keyperm	# align round key
+	vsrb	v1, v0, v8		# vpsrlb	\$4,	%xmm0,	%xmm0
+	vperm	v0, $iptlo, $iptlo, v0	# vpshufb	%xmm1,	%xmm2,	%xmm2
+	vperm	v1, $ipthi, $ipthi, v1	# vpshufb	%xmm0,	%xmm1,	%xmm0
+	vxor	v0, v0, v5		# vpxor	%xmm4,	%xmm2,	%xmm2
+	vxor	v0, v0, v1		# vpxor	%xmm2,	%xmm0,	%xmm0
+	mtctr	r8
+	b	Ldec_entry
+
+.align 4
+Ldec_loop:
+#
+#  Inverse mix columns
+#
+	lvx	v0, r12, r11		# v5 and v0 are flipped
+					# vmovdqa	-0x20(%r10),%xmm4		# 4 : sb9u
+					# vmovdqa	-0x10(%r10),%xmm1		# 0 : sb9t
+	vperm	v4, $sb9u, v7, v2	# vpshufb	%xmm2,	%xmm4,	%xmm4		# 4 = sb9u
+	subi	r11, r11, 16
+	vperm	v1, $sb9t, v7, v3	# vpshufb	%xmm3,	%xmm1,	%xmm1		# 0 = sb9t
+	andi.	r11, r11, 0x30
+	vxor	v5, v5, v4		# vpxor		%xmm4,	%xmm0,	%xmm0
+					# vmovdqa	0x00(%r10),%xmm4		# 4 : sbdu
+	vxor	v5, v5, v1		# vpxor		%xmm1,	%xmm0,	%xmm0		# 0 = ch
+					# vmovdqa	0x10(%r10),%xmm1		# 0 : sbdt
+
+	vperm	v4, $sbdu, v7, v2	# vpshufb	%xmm2,	%xmm4,	%xmm4		# 4 = sbdu
+	vperm 	v5, v5, v7, v0		# vpshufb	%xmm5,	%xmm0,	%xmm0		# MC ch
+	vperm	v1, $sbdt, v7, v3	# vpshufb	%xmm3,	%xmm1,	%xmm1		# 0 = sbdt
+	vxor	v5, v5, v4		# vpxor		%xmm4,	%xmm0,	%xmm0		# 4 = ch
+					# vmovdqa	0x20(%r10),	%xmm4		# 4 : sbbu
+	vxor	v5, v5, v1		# vpxor		%xmm1,	%xmm0,	%xmm0		# 0 = ch
+					# vmovdqa	0x30(%r10),	%xmm1		# 0 : sbbt
+
+	vperm	v4, $sbbu, v7, v2	# vpshufb	%xmm2,	%xmm4,	%xmm4		# 4 = sbbu
+	vperm	v5, v5, v7, v0		# vpshufb	%xmm5,	%xmm0,	%xmm0		# MC ch
+	vperm	v1, $sbbt, v7, v3	# vpshufb	%xmm3,	%xmm1,	%xmm1		# 0 = sbbt
+	vxor	v5, v5, v4		# vpxor		%xmm4,	%xmm0,	%xmm0		# 4 = ch
+					# vmovdqa	0x40(%r10),	%xmm4		# 4 : sbeu
+	vxor	v5, v5, v1		# vpxor		%xmm1,	%xmm0,	%xmm0		# 0 = ch
+					# vmovdqa	0x50(%r10),	%xmm1		# 0 : sbet
+
+	vperm	v4, $sbeu, v7, v2	# vpshufb	%xmm2,	%xmm4,	%xmm4		# 4 = sbeu
+	vperm	v5, v5, v7, v0		# vpshufb	%xmm5,	%xmm0,	%xmm0		# MC ch
+	vperm	v1, $sbet, v7, v3	# vpshufb	%xmm3,	%xmm1,	%xmm1		# 0 = sbet
+	vxor	v0, v5, v4		# vpxor		%xmm4,	%xmm0,	%xmm0		# 4 = ch
+	vxor	v0, v0, v1		# vpxor		%xmm1,	%xmm0,	%xmm0		# 0 = ch
+
+Ldec_entry:
+	# top of round
+	vsrb	v1, v0, v8		# vpsrlb	\$4,	%xmm0,	%xmm0	# 1 = i
+	vperm	v2, $invhi, $invhi, v0	# vpshufb	%xmm1,	%xmm11,	%xmm2	# 2 = a/k
+	vxor	v0, v0, v1		# vpxor		%xmm0,	%xmm1,	%xmm1	# 0 = j
+	vperm	v3, $invlo, $invlo, v1	# vpshufb	%xmm0, 	%xmm10,	%xmm3	# 3 = 1/i
+	vperm	v4, $invlo, $invlo, v0	# vpshufb	%xmm1,	%xmm10,	%xmm4	# 4 = 1/j
+	vand	v0, v0, v9
+	vxor	v3, v3, v2		# vpxor		%xmm2,	%xmm3,	%xmm3	# 3 = iak = 1/i + a/k
+	vxor	v4, v4, v2		# vpxor		%xmm2, 	%xmm4,	%xmm4	# 4 = jak = 1/j + a/k
+	vperm	v2, $invlo, v7, v3	# vpshufb	%xmm3,	%xmm10,	%xmm2	# 2 = 1/iak
+	vmr	v5, v6
+	lvx	v6, r9, $key		# vmovdqu	(%r9),	%xmm0
+	vperm	v3, $invlo, v7, v4	# vpshufb	%xmm4,  %xmm10,	%xmm3	# 3 = 1/jak
+	addi	r9, r9, 16
+	vxor	v2, v2, v0		# vpxor		%xmm1,	%xmm2,	%xmm2	# 2 = io
+	?vperm	v5, v5, v6, $keyperm	# align round key
+	vxor	v3, v3, v1		# vpxor		%xmm0,  %xmm3,	%xmm3	# 3 = jo
+	bdnz	Ldec_loop
+
+	# middle of last round
+	addi	r10, r11, 0x80
+					# vmovdqa	0x60(%r10),	%xmm4	# 3 : sbou
+	vperm	v4, $sbou, v7, v2	# vpshufb	%xmm2,	%xmm4,	%xmm4	# 4 = sbou
+					# vmovdqa	0x70(%r10),	%xmm1	# 0 : sbot
+	lvx	v2, r12, r10		# vmovdqa	-0x160(%r11),	%xmm2	# .Lk_sr-.Lk_dsbd=-0x160
+	vperm	v1, $sbot, v7, v3	# vpshufb	%xmm3,	%xmm1,	%xmm1	# 0 = sb1t
+	vxor	v4, v4, v5		# vpxor		%xmm0,	%xmm4,	%xmm4	# 4 = sb1u + k
+	vxor	v0, v1, v4		# vpxor		%xmm4,	%xmm1,	%xmm0	# 0 = A
+	vperm	v0, v0, v7, v2		# vpshufb	%xmm2,	%xmm0,	%xmm0
+	blr
+	.long	0
+	.byte	0,12,0x14,0,0,0,0,0
+
+.globl	.vpaes_decrypt
+.align	5
+.vpaes_decrypt:
+	$STU	$sp,-$FRAME($sp)
+	li	r10,`15+6*$SIZE_T`
+	li	r11,`31+6*$SIZE_T`
+	mflr	r6
+	mfspr	r7, 256			# save vrsave
+	stvx	v20,r10,$sp
+	addi	r10,r10,32
+	stvx	v21,r11,$sp
+	addi	r11,r11,32
+	stvx	v22,r10,$sp
+	addi	r10,r10,32
+	stvx	v23,r11,$sp
+	addi	r11,r11,32
+	stvx	v24,r10,$sp
+	addi	r10,r10,32
+	stvx	v25,r11,$sp
+	addi	r11,r11,32
+	stvx	v26,r10,$sp
+	addi	r10,r10,32
+	stvx	v27,r11,$sp
+	addi	r11,r11,32
+	stvx	v28,r10,$sp
+	addi	r10,r10,32
+	stvx	v29,r11,$sp
+	addi	r11,r11,32
+	stvx	v30,r10,$sp
+	stvx	v31,r11,$sp
+	stw	r7,`$FRAME-4`($sp)	# save vrsave
+	li	r0, -1
+	$PUSH	r6,`$FRAME+$LRSAVE`($sp)
+	mtspr	256, r0			# preserve all AltiVec registers
+
+	bl	_vpaes_decrypt_preheat
+
+	?lvsl	$inpperm, 0, $inp	# prepare for unaligned access
+	lvx	v0, 0, $inp
+	addi	$inp, $inp, 15		# 15 is not a typo
+	 ?lvsr	$outperm, 0, $out
+	?lvsl	$keyperm, 0, $key
+	 vnor	$outmask, v7, v7	# 0xff..ff
+	lvx	$inptail, 0, $inp	# redundant in aligned case
+	 ?vperm	$outmask, v7, $outmask, $outperm
+	 lvx	$outhead, 0, $out
+	?vperm	v0, v0, $inptail, $inpperm
+
+	bl	_vpaes_decrypt_core
+
+	vperm	v0, v0, v0, $outperm	# rotate right/left
+	vsel	v1, $outhead, v0, $outmask
+	vmr	$outhead, v0
+	stvx	v1, 0, $out
+	addi	$out, $out, 15		# 15 is not a typo
+	########
+
+	lvx	v1, 0, $out		# redundant in aligned case
+	vsel	v1, $outhead, v1, $outmask
+	stvx	v1, 0, $out
+
+	li	r10,`15+6*$SIZE_T`
+	li	r11,`31+6*$SIZE_T`
+	mtlr	r6
+	mtspr	256, r7			# restore vrsave
+	lvx	v20,r10,$sp
+	addi	r10,r10,32
+	lvx	v21,r11,$sp
+	addi	r11,r11,32
+	lvx	v22,r10,$sp
+	addi	r10,r10,32
+	lvx	v23,r11,$sp
+	addi	r11,r11,32
+	lvx	v24,r10,$sp
+	addi	r10,r10,32
+	lvx	v25,r11,$sp
+	addi	r11,r11,32
+	lvx	v26,r10,$sp
+	addi	r10,r10,32
+	lvx	v27,r11,$sp
+	addi	r11,r11,32
+	lvx	v28,r10,$sp
+	addi	r10,r10,32
+	lvx	v29,r11,$sp
+	addi	r11,r11,32
+	lvx	v30,r10,$sp
+	lvx	v31,r11,$sp
+	addi	$sp,$sp,$FRAME
+	blr
+	.long	0
+	.byte	0,12,0x04,1,0x80,0,3,0
+	.long	0
+.size	.vpaes_decrypt,.-.vpaes_decrypt
+
+.globl	.vpaes_cbc_encrypt
+.align	5
+.vpaes_cbc_encrypt:
+	${UCMP}i r5,16
+	bltlr-
+
+	$STU	$sp,-`($FRAME+2*$SIZE_T)`($sp)
+	mflr	r0
+	li	r10,`15+6*$SIZE_T`
+	li	r11,`31+6*$SIZE_T`
+	mfspr	r12, 256
+	stvx	v20,r10,$sp
+	addi	r10,r10,32
+	stvx	v21,r11,$sp
+	addi	r11,r11,32
+	stvx	v22,r10,$sp
+	addi	r10,r10,32
+	stvx	v23,r11,$sp
+	addi	r11,r11,32
+	stvx	v24,r10,$sp
+	addi	r10,r10,32
+	stvx	v25,r11,$sp
+	addi	r11,r11,32
+	stvx	v26,r10,$sp
+	addi	r10,r10,32
+	stvx	v27,r11,$sp
+	addi	r11,r11,32
+	stvx	v28,r10,$sp
+	addi	r10,r10,32
+	stvx	v29,r11,$sp
+	addi	r11,r11,32
+	stvx	v30,r10,$sp
+	stvx	v31,r11,$sp
+	stw	r12,`$FRAME-4`($sp)	# save vrsave
+	$PUSH	r30,`$FRAME+$SIZE_T*0`($sp)
+	$PUSH	r31,`$FRAME+$SIZE_T*1`($sp)
+	li	r9, -16
+	$PUSH	r0, `$FRAME+$SIZE_T*2+$LRSAVE`($sp)
+
+	and	r30, r5, r9		# copy length&-16
+	mr	r5, r6			# copy pointer to key
+	mr	r31, r7			# copy pointer to iv
+	blt	Lcbc_abort
+	cmpwi	r8, 0			# test direction
+	li	r6, -1
+	mr	r7, r12			# copy vrsave
+	mtspr	256, r6			# preserve all AltiVec registers
+
+	lvx	v24, 0, r31		# load [potentially unaligned] iv
+	li	r9, 15
+	?lvsl	$inpperm, 0, r31
+	lvx	v25, r9, r31
+	?vperm	v24, v24, v25, $inpperm
+
+	neg	r8, $inp		# prepare for unaligned access
+	 vxor	v7, v7, v7
+	?lvsl	$keyperm, 0, $key
+	 ?lvsr	$outperm, 0, $out
+	?lvsr	$inpperm, 0, r8		# -$inp
+	 vnor	$outmask, v7, v7	# 0xff..ff
+	lvx	$inptail, 0, $inp
+	 ?vperm	$outmask, v7, $outmask, $outperm
+	addi	$inp, $inp, 15		# 15 is not a typo
+	 lvx	$outhead, 0, $out
+
+	beq	Lcbc_decrypt
+
+	bl	_vpaes_encrypt_preheat
+	li	r0, 16
+
+Lcbc_enc_loop:
+	vmr	v0, $inptail
+	lvx	$inptail, 0, $inp
+	addi	$inp, $inp, 16
+	?vperm	v0, v0, $inptail, $inpperm
+	vxor	v0, v0, v24		# ^= iv
+
+	bl	_vpaes_encrypt_core
+
+	vmr	v24, v0			# put aside iv
+	sub.	r30, r30, r0		# len -= 16
+	vperm	v0, v0, v0, $outperm	# rotate right/left
+	vsel	v1, $outhead, v0, $outmask
+	vmr	$outhead, v0
+	stvx	v1, 0, $out
+	addi	$out, $out, 16
+	bne	Lcbc_enc_loop
+
+	b	Lcbc_done
+
+.align	5
+Lcbc_decrypt:
+	bl	_vpaes_decrypt_preheat
+	li	r0, 16
+
+Lcbc_dec_loop:
+	vmr	v0, $inptail
+	lvx	$inptail, 0, $inp
+	addi	$inp, $inp, 16
+	?vperm	v0, v0, $inptail, $inpperm
+	vmr	v25, v0			# put aside input
+
+	bl	_vpaes_decrypt_core
+
+	vxor	v0, v0, v24		# ^= iv
+	vmr	v24, v25
+	sub.	r30, r30, r0		# len -= 16
+	vperm	v0, v0, v0, $outperm	# rotate right/left
+	vsel	v1, $outhead, v0, $outmask
+	vmr	$outhead, v0
+	stvx	v1, 0, $out
+	addi	$out, $out, 16
+	bne	Lcbc_dec_loop
+
+Lcbc_done:
+	addi	$out, $out, -1
+	lvx	v1, 0, $out		# redundant in aligned case
+	vsel	v1, $outhead, v1, $outmask
+	stvx	v1, 0, $out
+
+	neg	r8, r31			# write [potentially unaligned] iv
+	?lvsl	$outperm, 0, r8
+	li	r6, 15
+	vnor	$outmask, v7, v7	# 0xff..ff
+	?vperm	$outmask, v7, $outmask, $outperm
+	lvx	$outhead, 0, r31
+	vperm	v24, v24, v24, $outperm	# rotate right/left
+	vsel	v0, $outhead, v24, $outmask
+	lvx	v1, r6, r31
+	stvx	v0, 0, r31
+	vsel	v1, v24, v1, $outmask
+	stvx	v1, r6, r31
+
+	mtspr	256, r7			# restore vrsave
+	li	r10,`15+6*$SIZE_T`
+	li	r11,`31+6*$SIZE_T`
+	lvx	v20,r10,$sp
+	addi	r10,r10,32
+	lvx	v21,r11,$sp
+	addi	r11,r11,32
+	lvx	v22,r10,$sp
+	addi	r10,r10,32
+	lvx	v23,r11,$sp
+	addi	r11,r11,32
+	lvx	v24,r10,$sp
+	addi	r10,r10,32
+	lvx	v25,r11,$sp
+	addi	r11,r11,32
+	lvx	v26,r10,$sp
+	addi	r10,r10,32
+	lvx	v27,r11,$sp
+	addi	r11,r11,32
+	lvx	v28,r10,$sp
+	addi	r10,r10,32
+	lvx	v29,r11,$sp
+	addi	r11,r11,32
+	lvx	v30,r10,$sp
+	lvx	v31,r11,$sp
+Lcbc_abort:
+	$POP	r0, `$FRAME+$SIZE_T*2+$LRSAVE`($sp)
+	$POP	r30,`$FRAME+$SIZE_T*0`($sp)
+	$POP	r31,`$FRAME+$SIZE_T*1`($sp)
+	mtlr	r0
+	addi	$sp,$sp,`$FRAME+$SIZE_T*2`
+	blr
+	.long	0
+	.byte	0,12,0x04,1,0x80,2,6,0
+	.long	0
+.size	.vpaes_cbc_encrypt,.-.vpaes_cbc_encrypt
+___
+}
+{
+my ($inp,$bits,$out)=map("r$_",(3..5));
+my $dir="cr1";
+my ($invlo,$invhi,$iptlo,$ipthi,$rcon) = map("v$_",(10..13,24));
+
+$code.=<<___;
+########################################################
+##                                                    ##
+##                  AES key schedule                  ##
+##                                                    ##
+########################################################
+.align	4
+_vpaes_key_preheat:
+	mflr	r8
+	bl	Lconsts
+	mtlr	r8
+	li	r11, 0xc0		# Lk_inv
+	li	r10, 0xd0
+	li	r9,  0xe0		# L_ipt
+	li	r8,  0xf0
+
+	vspltisb	v8,4		# 0x04..04
+	vxor	v9,v9,v9		# 0x00..00
+	lvx	$invlo, r12, r11	# Lk_inv
+	li	r11, 0x120
+	lvx	$invhi, r12, r10
+	li	r10, 0x130
+	lvx	$iptlo, r12, r9		# Lk_ipt
+	li	r9, 0x220
+	lvx	$ipthi, r12, r8
+	li	r8, 0x230
+
+	lvx	v14, r12, r11		# Lk_sb1
+	li	r11, 0x240
+	lvx	v15, r12, r10
+	li	r10, 0x250
+
+	lvx	v16, r12, r9		# Lk_dksd
+	li	r9, 0x260
+	lvx	v17, r12, r8
+	li	r8, 0x270
+	lvx	v18, r12, r11		# Lk_dksb
+	li	r11, 0x280
+	lvx	v19, r12, r10
+	li	r10, 0x290
+	lvx	v20, r12, r9		# Lk_dkse
+	li	r9, 0x2a0
+	lvx	v21, r12, r8
+	li	r8, 0x2b0
+	lvx	v22, r12, r11		# Lk_dks9
+	lvx	v23, r12, r10
+
+	lvx	v24, r12, r9		# Lk_rcon
+	lvx	v25, 0, r12		# Lk_mc_forward[0]
+	lvx	v26, r12, r8		# Lks63
+	blr
+	.long	0
+	.byte	0,12,0x14,0,0,0,0,0
+
+.align	4
+_vpaes_schedule_core:
+	mflr	r7
+
+	bl	_vpaes_key_preheat	# load the tables
+
+	#lvx	v0, 0, $inp		# vmovdqu	(%rdi),	%xmm0		# load key (unaligned)
+	neg	r8, $inp		# prepare for unaligned access
+	lvx	v0, 0, $inp
+	addi	$inp, $inp, 15		# 15 is not typo
+	?lvsr	$inpperm, 0, r8		# -$inp
+	lvx	v6, 0, $inp		# v6 serves as inptail
+	addi	$inp, $inp, 8
+	?vperm	v0, v0, v6, $inpperm
+
+	# input transform
+	vmr	v3, v0			# vmovdqa	%xmm0,	%xmm3
+	bl	_vpaes_schedule_transform
+	vmr	v7, v0			# vmovdqa	%xmm0,	%xmm7
+
+	bne	$dir, Lschedule_am_decrypting
+
+	# encrypting, output zeroth round key after transform
+	li	r8, 0x30		# mov	\$0x30,%r8d
+	addi	r10, r12, 0x80		# lea	.Lk_sr(%rip),%r10
+
+	?lvsr	$outperm, 0, $out	# prepare for unaligned access
+	vnor	$outmask, v9, v9	# 0xff..ff
+	lvx	$outhead, 0, $out
+	?vperm	$outmask, v9, $outmask, $outperm
+
+	#stvx	v0, 0, $out		# vmovdqu	%xmm0,	(%rdx)
+	vperm	v1, v0, v0, $outperm	# rotate right/left
+	vsel	v2, $outhead, v1, $outmask
+	vmr	$outhead, v1
+	stvx	v2, 0, $out
+	b	Lschedule_go
+
+Lschedule_am_decrypting:
+	srwi	r8, $bits, 1		# shr	\$1,%r8d
+	andi.	r8, r8, 32		# and	\$32,%r8d
+	xori	r8, r8, 32		# xor	\$32,%r8d	# nbits==192?0:32
+	addi	r10, r12, 0x80		# lea	.Lk_sr(%rip),%r10
+	# decrypting, output zeroth round key after shiftrows
+	lvx	v1, r8, r10		# vmovdqa	(%r8,%r10),	%xmm1
+	vperm	v4, v3, v3, v1		# vpshufb	%xmm1,	%xmm3,	%xmm3
+
+	neg	r0, $out		# prepare for unaligned access
+	?lvsl	$outperm, 0, r0
+	addi	$out, $out, 15		# 15 is not typo
+	vnor	$outmask, v9, v9	# 0xff..ff
+	lvx	$outhead, 0, $out
+	?vperm	$outmask, $outmask, v9, $outperm
+
+	#stvx	v4, 0, $out		# vmovdqu	%xmm3,	(%rdx)
+	vperm	v4, v4, v4, $outperm	# rotate right/left
+	vsel	v2, $outhead, v4, $outmask
+	vmr	$outhead, v4
+	stvx	v2, 0, $out
+	xori	r8, r8, 0x30		# xor	\$0x30, %r8
+
+Lschedule_go:
+	cmplwi	$bits, 192		# cmp	\$192,	%esi
+	bgt	Lschedule_256
+	beq	Lschedule_192
+	# 128: fall though
+
+##
+##  .schedule_128
+##
+##  128-bit specific part of key schedule.
+##
+##  This schedule is really simple, because all its parts
+##  are accomplished by the subroutines.
+##
+Lschedule_128:
+	li	r0, 10			# mov	\$10, %esi
+	mtctr	r0
+
+Loop_schedule_128:
+	bl 	_vpaes_schedule_round
+	bdz 	Lschedule_mangle_last	# dec	%esi
+	bl	_vpaes_schedule_mangle	# write output
+	b 	Loop_schedule_128
+
+##
+##  .aes_schedule_192
+##
+##  192-bit specific part of key schedule.
+##
+##  The main body of this schedule is the same as the 128-bit
+##  schedule, but with more smearing.  The long, high side is
+##  stored in %xmm7 as before, and the short, low side is in
+##  the high bits of %xmm6.
+##
+##  This schedule is somewhat nastier, however, because each
+##  round produces 192 bits of key material, or 1.5 round keys.
+##  Therefore, on each cycle we do 2 rounds and produce 3 round
+##  keys.
+##
+.align	4
+Lschedule_192:
+	li	r0, 4			# mov	\$4,	%esi
+	lvx	v0, 0, $inp
+	?vperm	v0, v6, v0, $inpperm
+	?vsldoi	v0, v3, v0, 8		# vmovdqu	8(%rdi),%xmm0		# load key part 2 (very unaligned)
+	bl	_vpaes_schedule_transform	# input transform
+	?vsldoi	v6, v0, v9, 8
+	?vsldoi	v6, v9, v6, 8		# clobber "low" side with zeros
+	mtctr	r0
+
+Loop_schedule_192:
+	bl	_vpaes_schedule_round
+	?vsldoi	v0, v6, v0, 8		# vpalignr	\$8,%xmm6,%xmm0,%xmm0
+	bl	_vpaes_schedule_mangle	# save key n
+	bl	_vpaes_schedule_192_smear
+	bl	_vpaes_schedule_mangle	# save key n+1
+	bl	_vpaes_schedule_round
+	bdz 	Lschedule_mangle_last	# dec	%esi
+	bl	_vpaes_schedule_mangle	# save key n+2
+	bl	_vpaes_schedule_192_smear
+	b	Loop_schedule_192
+
+##
+##  .aes_schedule_256
+##
+##  256-bit specific part of key schedule.
+##
+##  The structure here is very similar to the 128-bit
+##  schedule, but with an additional "low side" in
+##  %xmm6.  The low side's rounds are the same as the
+##  high side's, except no rcon and no rotation.
+##
+.align	4
+Lschedule_256:
+	li	r0, 7			# mov	\$7, %esi
+	addi	$inp, $inp, 8
+	lvx	v0, 0, $inp		# vmovdqu	16(%rdi),%xmm0		# load key part 2 (unaligned)
+	?vperm	v0, v6, v0, $inpperm
+	bl	_vpaes_schedule_transform	# input transform
+	mtctr	r0
+
+Loop_schedule_256:
+	bl	_vpaes_schedule_mangle	# output low result
+	vmr	v6, v0			# vmovdqa	%xmm0,	%xmm6		# save cur_lo in xmm6
+
+	# high round
+	bl	_vpaes_schedule_round
+	bdz 	Lschedule_mangle_last	# dec	%esi
+	bl	_vpaes_schedule_mangle	
+
+	# low round. swap xmm7 and xmm6
+	?vspltw	v0, v0, 3		# vpshufd	\$0xFF,	%xmm0,	%xmm0
+	vmr	v5, v7			# vmovdqa	%xmm7,	%xmm5
+	vmr	v7, v6			# vmovdqa	%xmm6,	%xmm7
+	bl	_vpaes_schedule_low_round
+	vmr	v7, v5			# vmovdqa	%xmm5,	%xmm7
+	
+	b	Loop_schedule_256
+##
+##  .aes_schedule_mangle_last
+##
+##  Mangler for last round of key schedule
+##  Mangles %xmm0
+##    when encrypting, outputs out(%xmm0) ^ 63
+##    when decrypting, outputs unskew(%xmm0)
+##
+##  Always called right before return... jumps to cleanup and exits
+##
+.align	4
+Lschedule_mangle_last:
+	# schedule last round key from xmm0
+	li	r11, 0x2e0		# lea	.Lk_deskew(%rip),%r11
+	li	r9,  0x2f0
+	bne	$dir, Lschedule_mangle_last_dec
+
+	# encrypting
+	lvx	v1, r8, r10		# vmovdqa	(%r8,%r10),%xmm1
+	li	r11, 0x2c0		# lea		.Lk_opt(%rip),	%r11	# prepare to output transform
+	li	r9,  0x2d0		# prepare to output transform
+	vperm	v0, v0, v0, v1		# vpshufb	%xmm1,	%xmm0,	%xmm0	# output permute
+
+	lvx	$iptlo, r11, r12	# reload $ipt
+	lvx	$ipthi, r9, r12
+	addi	$out, $out, 16		# add	\$16,	%rdx
+	vxor	v0, v0, v26		# vpxor		.Lk_s63(%rip),	%xmm0,	%xmm0
+	bl	_vpaes_schedule_transform	# output transform
+
+	#stvx	v0, r0, $out		# vmovdqu	%xmm0,	(%rdx)		# save last key
+	vperm	v0, v0, v0, $outperm	# rotate right/left
+	vsel	v2, $outhead, v0, $outmask
+	vmr	$outhead, v0
+	stvx	v2, 0, $out
+
+	addi	$out, $out, 15		# 15 is not typo
+	lvx	v1, 0, $out		# redundant in aligned case
+	vsel	v1, $outhead, v1, $outmask
+	stvx	v1, 0, $out
+	b	Lschedule_mangle_done
+
+.align	4
+Lschedule_mangle_last_dec:
+	lvx	$iptlo, r11, r12	# reload $ipt
+	lvx	$ipthi, r9,  r12
+	addi	$out, $out, -16		# add	\$-16,	%rdx 
+	vxor	v0, v0, v26		# vpxor	.Lk_s63(%rip),	%xmm0,	%xmm0
+	bl	_vpaes_schedule_transform	# output transform
+
+	#stvx	v0, r0, $out		# vmovdqu	%xmm0,	(%rdx)		# save last key
+	vperm	v0, v0, v0, $outperm	# rotate right/left
+	vsel	v2, $outhead, v0, $outmask
+	vmr	$outhead, v0
+	stvx	v2, 0, $out
+
+	addi	$out, $out, -15		# -15 is not typo
+	lvx	v1, 0, $out		# redundant in aligned case
+	vsel	v1, $outhead, v1, $outmask
+	stvx	v1, 0, $out
+
+Lschedule_mangle_done:
+	mtlr	r7
+	# cleanup
+	vxor	v0, v0, v0		# vpxor		%xmm0,	%xmm0,	%xmm0
+	vxor	v1, v1, v1		# vpxor		%xmm1,	%xmm1,	%xmm1
+	vxor	v2, v2, v2		# vpxor		%xmm2,	%xmm2,	%xmm2
+	vxor	v3, v3, v3		# vpxor		%xmm3,	%xmm3,	%xmm3
+	vxor	v4, v4, v4		# vpxor		%xmm4,	%xmm4,	%xmm4
+	vxor	v5, v5, v5		# vpxor		%xmm5,	%xmm5,	%xmm5
+	vxor	v6, v6, v6		# vpxor		%xmm6,	%xmm6,	%xmm6
+	vxor	v7, v7, v7		# vpxor		%xmm7,	%xmm7,	%xmm7
+
+	blr
+	.long	0
+	.byte	0,12,0x14,0,0,0,0,0
+
+##
+##  .aes_schedule_192_smear
+##
+##  Smear the short, low side in the 192-bit key schedule.
+##
+##  Inputs:
+##    %xmm7: high side, b  a  x  y
+##    %xmm6:  low side, d  c  0  0
+##    %xmm13: 0
+##
+##  Outputs:
+##    %xmm6: b+c+d  b+c  0  0
+##    %xmm0: b+c+d  b+c  b  a
+##
+.align	4
+_vpaes_schedule_192_smear:
+	?vspltw	v0, v7, 3
+	?vsldoi	v1, v9, v6, 12		# vpshufd	\$0x80,	%xmm6,	%xmm1	# d c 0 0 -> c 0 0 0
+	?vsldoi	v0, v7, v0, 8		# vpshufd	\$0xFE,	%xmm7,	%xmm0	# b a _ _ -> b b b a
+	vxor	v6, v6, v1		# vpxor		%xmm1,	%xmm6,	%xmm6	# -> c+d c 0 0
+	vxor	v6, v6, v0		# vpxor		%xmm0,	%xmm6,	%xmm6	# -> b+c+d b+c b a
+	vmr	v0, v6
+	?vsldoi	v6, v6, v9, 8
+	?vsldoi	v6, v9, v6, 8		# clobber low side with zeros
+	blr
+	.long	0
+	.byte	0,12,0x14,0,0,0,0,0
+
+##
+##  .aes_schedule_round
+##
+##  Runs one main round of the key schedule on %xmm0, %xmm7
+##
+##  Specifically, runs subbytes on the high dword of %xmm0
+##  then rotates it by one byte and xors into the low dword of
+##  %xmm7.
+##
+##  Adds rcon from low byte of %xmm8, then rotates %xmm8 for
+##  next rcon.
+##
+##  Smears the dwords of %xmm7 by xoring the low into the
+##  second low, result into third, result into highest.
+##
+##  Returns results in %xmm7 = %xmm0.
+##  Clobbers %xmm1-%xmm4, %r11.
+##
+.align	4
+_vpaes_schedule_round:
+	# extract rcon from xmm8
+	#vxor	v4, v4, v4		# vpxor		%xmm4,	%xmm4,	%xmm4
+	?vsldoi	v1, $rcon, v9, 15	# vpalignr	\$15,	%xmm8,	%xmm4,	%xmm1
+	?vsldoi	$rcon, $rcon, $rcon, 15	# vpalignr	\$15,	%xmm8,	%xmm8,	%xmm8
+	vxor	v7, v7, v1		# vpxor		%xmm1,	%xmm7,	%xmm7
+
+	# rotate
+	?vspltw	v0, v0, 3		# vpshufd	\$0xFF,	%xmm0,	%xmm0
+	?vsldoi	v0, v0, v0, 1		# vpalignr	\$1,	%xmm0,	%xmm0,	%xmm0
+
+	# fall through...
+
+	# low round: same as high round, but no rotation and no rcon.
+_vpaes_schedule_low_round:
+	# smear xmm7
+	?vsldoi	v1, v9, v7, 12		# vpslldq	\$4,	%xmm7,	%xmm1
+	vxor	v7, v7, v1		# vpxor		%xmm1,	%xmm7,	%xmm7
+	vspltisb	v1, 0x0f	# 0x0f..0f
+	?vsldoi	v4, v9, v7, 8		# vpslldq	\$8,	%xmm7,	%xmm4
+
+	# subbytes
+	vand	v1, v1, v0		# vpand		%xmm9,	%xmm0,	%xmm1		# 0 = k
+	vsrb	v0, v0, v8		# vpsrlb	\$4,	%xmm0,	%xmm0		# 1 = i
+	 vxor	v7, v7, v4		# vpxor		%xmm4,	%xmm7,	%xmm7
+	vperm	v2, $invhi, v9, v1	# vpshufb	%xmm1,	%xmm11,	%xmm2		# 2 = a/k
+	vxor	v1, v1, v0		# vpxor		%xmm0,	%xmm1,	%xmm1		# 0 = j
+	vperm	v3, $invlo, v9, v0	# vpshufb	%xmm0, 	%xmm10,	%xmm3		# 3 = 1/i
+	vxor	v3, v3, v2		# vpxor		%xmm2,	%xmm3,	%xmm3		# 3 = iak = 1/i + a/k
+	vperm	v4, $invlo, v9, v1	# vpshufb	%xmm1,	%xmm10,	%xmm4		# 4 = 1/j
+	 vxor	v7, v7, v26		# vpxor		.Lk_s63(%rip),	%xmm7,	%xmm7
+	vperm	v3, $invlo, v9, v3	# vpshufb	%xmm3,	%xmm10,	%xmm3		# 2 = 1/iak
+	vxor	v4, v4, v2		# vpxor		%xmm2,	%xmm4,	%xmm4		# 4 = jak = 1/j + a/k
+	vperm	v2, $invlo, v9, v4	# vpshufb	%xmm4,	%xmm10,	%xmm2		# 3 = 1/jak
+	vxor	v3, v3, v1		# vpxor		%xmm1,	%xmm3,	%xmm3		# 2 = io
+	vxor	v2, v2, v0		# vpxor		%xmm0,	%xmm2,	%xmm2		# 3 = jo
+	vperm	v4, v15, v9, v3		# vpshufb	%xmm3,	%xmm13,	%xmm4		# 4 = sbou
+	vperm	v1, v14, v9, v2		# vpshufb	%xmm2,	%xmm12,	%xmm1		# 0 = sb1t
+	vxor	v1, v1, v4		# vpxor		%xmm4,	%xmm1,	%xmm1		# 0 = sbox output
+
+	# add in smeared stuff
+	vxor	v0, v1, v7		# vpxor		%xmm7,	%xmm1,	%xmm0
+	vxor	v7, v1, v7		# vmovdqa	%xmm0,	%xmm7
+	blr
+	.long	0
+	.byte	0,12,0x14,0,0,0,0,0
+
+##
+##  .aes_schedule_transform
+##
+##  Linear-transform %xmm0 according to tables at (%r11)
+##
+##  Requires that %xmm9 = 0x0F0F... as in preheat
+##  Output in %xmm0
+##  Clobbers %xmm2
+##
+.align	4
+_vpaes_schedule_transform:
+	#vand	v1, v0, v9		# vpand		%xmm9,	%xmm0,	%xmm1
+	vsrb	v2, v0, v8		# vpsrlb	\$4,	%xmm0,	%xmm0
+					# vmovdqa	(%r11),	%xmm2 	# lo
+	vperm	v0, $iptlo, $iptlo, v0	# vpshufb	%xmm1,	%xmm2,	%xmm2
+					# vmovdqa	16(%r11),	%xmm1 # hi
+	vperm	v2, $ipthi, $ipthi, v2	# vpshufb	%xmm0,	%xmm1,	%xmm0
+	vxor	v0, v0, v2		# vpxor		%xmm2,	%xmm0,	%xmm0
+	blr
+	.long	0
+	.byte	0,12,0x14,0,0,0,0,0
+
+##
+##  .aes_schedule_mangle
+##
+##  Mangle xmm0 from (basis-transformed) standard version
+##  to our version.
+##
+##  On encrypt,
+##    xor with 0x63
+##    multiply by circulant 0,1,1,1
+##    apply shiftrows transform
+##
+##  On decrypt,
+##    xor with 0x63
+##    multiply by "inverse mixcolumns" circulant E,B,D,9
+##    deskew
+##    apply shiftrows transform
+##
+##
+##  Writes out to (%rdx), and increments or decrements it
+##  Keeps track of round number mod 4 in %r8
+##  Preserves xmm0
+##  Clobbers xmm1-xmm5
+##
+.align	4
+_vpaes_schedule_mangle:
+	#vmr	v4, v0			# vmovdqa	%xmm0,	%xmm4	# save xmm0 for later
+					# vmovdqa	.Lk_mc_forward(%rip),%xmm5
+	bne	$dir, Lschedule_mangle_dec
+
+	# encrypting
+	vxor	v4, v0, v26		# vpxor	.Lk_s63(%rip),	%xmm0,	%xmm4
+	addi	$out, $out, 16		# add	\$16,	%rdx
+	vperm	v4, v4, v4, v25		# vpshufb	%xmm5,	%xmm4,	%xmm4
+	vperm	v1, v4, v4, v25		# vpshufb	%xmm5,	%xmm4,	%xmm1
+	vperm	v3, v1, v1, v25		# vpshufb	%xmm5,	%xmm1,	%xmm3
+	vxor	v4, v4, v1		# vpxor		%xmm1,	%xmm4,	%xmm4
+	lvx	v1, r8, r10		# vmovdqa	(%r8,%r10),	%xmm1
+	vxor	v3, v3, v4		# vpxor		%xmm4,	%xmm3,	%xmm3
+
+	vperm	v3, v3, v3, v1		# vpshufb	%xmm1,	%xmm3,	%xmm3
+	addi	r8, r8, -16		# add	\$-16,	%r8
+	andi.	r8, r8, 0x30		# and	\$0x30,	%r8
+
+	#stvx	v3, 0, $out		# vmovdqu	%xmm3,	(%rdx)
+	vperm	v1, v3, v3, $outperm	# rotate right/left
+	vsel	v2, $outhead, v1, $outmask
+	vmr	$outhead, v1
+	stvx	v2, 0, $out
+	blr
+
+.align	4
+Lschedule_mangle_dec:
+	# inverse mix columns
+					# lea	.Lk_dksd(%rip),%r11
+	vsrb	v1, v0, v8		# vpsrlb	\$4,	%xmm4,	%xmm1	# 1 = hi
+	#and	v4, v0, v9		# vpand		%xmm9,	%xmm4,	%xmm4	# 4 = lo
+
+					# vmovdqa	0x00(%r11),	%xmm2
+	vperm	v2, v16, v16, v0	# vpshufb	%xmm4,	%xmm2,	%xmm2
+					# vmovdqa	0x10(%r11),	%xmm3
+	vperm	v3, v17, v17, v1	# vpshufb	%xmm1,	%xmm3,	%xmm3
+	vxor	v3, v3, v2		# vpxor		%xmm2,	%xmm3,	%xmm3
+	vperm	v3, v3, v9, v25		# vpshufb	%xmm5,	%xmm3,	%xmm3
+
+					# vmovdqa	0x20(%r11),	%xmm2
+	vperm	v2, v18, v18, v0	# vpshufb	%xmm4,	%xmm2,	%xmm2
+	vxor	v2, v2, v3		# vpxor		%xmm3,	%xmm2,	%xmm2
+					# vmovdqa	0x30(%r11),	%xmm3
+	vperm	v3, v19, v19, v1	# vpshufb	%xmm1,	%xmm3,	%xmm3
+	vxor	v3, v3, v2		# vpxor		%xmm2,	%xmm3,	%xmm3
+	vperm	v3, v3, v9, v25		# vpshufb	%xmm5,	%xmm3,	%xmm3
+
+					# vmovdqa	0x40(%r11),	%xmm2
+	vperm	v2, v20, v20, v0	# vpshufb	%xmm4,	%xmm2,	%xmm2
+	vxor	v2, v2, v3		# vpxor		%xmm3,	%xmm2,	%xmm2
+					# vmovdqa	0x50(%r11),	%xmm3
+	vperm	v3, v21, v21, v1	# vpshufb	%xmm1,	%xmm3,	%xmm3
+	vxor	v3, v3, v2		# vpxor		%xmm2,	%xmm3,	%xmm3
+
+					# vmovdqa	0x60(%r11),	%xmm2
+	vperm	v2, v22, v22, v0	# vpshufb	%xmm4,	%xmm2,	%xmm2
+	vperm	v3, v3, v9, v25		# vpshufb	%xmm5,	%xmm3,	%xmm3
+					# vmovdqa	0x70(%r11),	%xmm4
+	vperm	v4, v23, v23, v1	# vpshufb	%xmm1,	%xmm4,	%xmm4
+	lvx	v1, r8, r10		# vmovdqa	(%r8,%r10),	%xmm1
+	vxor	v2, v2, v3		# vpxor		%xmm3,	%xmm2,	%xmm2
+	vxor	v3, v4, v2		# vpxor		%xmm2,	%xmm4,	%xmm3
+
+	addi	$out, $out, -16		# add	\$-16,	%rdx
+
+	vperm	v3, v3, v3, v1		# vpshufb	%xmm1,	%xmm3,	%xmm3
+	addi	r8, r8, -16		# add	\$-16,	%r8
+	andi.	r8, r8, 0x30		# and	\$0x30,	%r8
+
+	#stvx	v3, 0, $out		# vmovdqu	%xmm3,	(%rdx)
+	vperm	v1, v3, v3, $outperm	# rotate right/left
+	vsel	v2, $outhead, v1, $outmask
+	vmr	$outhead, v1
+	stvx	v2, 0, $out
+	blr
+	.long	0
+	.byte	0,12,0x14,0,0,0,0,0
+
+.globl	.vpaes_set_encrypt_key
+.align	5
+.vpaes_set_encrypt_key:
+	$STU	$sp,-$FRAME($sp)
+	li	r10,`15+6*$SIZE_T`
+	li	r11,`31+6*$SIZE_T`
+	mflr	r0
+	mfspr	r6, 256			# save vrsave
+	stvx	v20,r10,$sp
+	addi	r10,r10,32
+	stvx	v21,r11,$sp
+	addi	r11,r11,32
+	stvx	v22,r10,$sp
+	addi	r10,r10,32
+	stvx	v23,r11,$sp
+	addi	r11,r11,32
+	stvx	v24,r10,$sp
+	addi	r10,r10,32
+	stvx	v25,r11,$sp
+	addi	r11,r11,32
+	stvx	v26,r10,$sp
+	addi	r10,r10,32
+	stvx	v27,r11,$sp
+	addi	r11,r11,32
+	stvx	v28,r10,$sp
+	addi	r10,r10,32
+	stvx	v29,r11,$sp
+	addi	r11,r11,32
+	stvx	v30,r10,$sp
+	stvx	v31,r11,$sp
+	stw	r6,`$FRAME-4`($sp)	# save vrsave
+	li	r7, -1
+	$PUSH	r0, `$FRAME+$LRSAVE`($sp)
+	mtspr	256, r7			# preserve all AltiVec registers
+
+	srwi	r9, $bits, 5		# shr	\$5,%eax
+	addi	r9, r9, 6		# add	\$5,%eax
+	stw	r9, 240($out)		# mov	%eax,240(%rdx)	# AES_KEY->rounds = nbits/32+5;
+
+	cmplw	$dir, $bits, $bits	# set encrypt direction
+	li	r8, 0x30		# mov	\$0x30,%r8d
+	bl	_vpaes_schedule_core
+
+	$POP	r0, `$FRAME+$LRSAVE`($sp)
+	li	r10,`15+6*$SIZE_T`
+	li	r11,`31+6*$SIZE_T`
+	mtspr	256, r6			# restore vrsave
+	mtlr	r0
+	xor	r3, r3, r3
+	lvx	v20,r10,$sp
+	addi	r10,r10,32
+	lvx	v21,r11,$sp
+	addi	r11,r11,32
+	lvx	v22,r10,$sp
+	addi	r10,r10,32
+	lvx	v23,r11,$sp
+	addi	r11,r11,32
+	lvx	v24,r10,$sp
+	addi	r10,r10,32
+	lvx	v25,r11,$sp
+	addi	r11,r11,32
+	lvx	v26,r10,$sp
+	addi	r10,r10,32
+	lvx	v27,r11,$sp
+	addi	r11,r11,32
+	lvx	v28,r10,$sp
+	addi	r10,r10,32
+	lvx	v29,r11,$sp
+	addi	r11,r11,32
+	lvx	v30,r10,$sp
+	lvx	v31,r11,$sp
+	addi	$sp,$sp,$FRAME
+	blr
+	.long	0
+	.byte	0,12,0x04,1,0x80,0,3,0
+	.long	0
+.size	.vpaes_set_encrypt_key,.-.vpaes_set_encrypt_key
+
+.globl	.vpaes_set_decrypt_key
+.align	4
+.vpaes_set_decrypt_key:
+	$STU	$sp,-$FRAME($sp)
+	li	r10,`15+6*$SIZE_T`
+	li	r11,`31+6*$SIZE_T`
+	mflr	r0
+	mfspr	r6, 256			# save vrsave
+	stvx	v20,r10,$sp
+	addi	r10,r10,32
+	stvx	v21,r11,$sp
+	addi	r11,r11,32
+	stvx	v22,r10,$sp
+	addi	r10,r10,32
+	stvx	v23,r11,$sp
+	addi	r11,r11,32
+	stvx	v24,r10,$sp
+	addi	r10,r10,32
+	stvx	v25,r11,$sp
+	addi	r11,r11,32
+	stvx	v26,r10,$sp
+	addi	r10,r10,32
+	stvx	v27,r11,$sp
+	addi	r11,r11,32
+	stvx	v28,r10,$sp
+	addi	r10,r10,32
+	stvx	v29,r11,$sp
+	addi	r11,r11,32
+	stvx	v30,r10,$sp
+	stvx	v31,r11,$sp
+	stw	r6,`$FRAME-4`($sp)	# save vrsave
+	li	r7, -1
+	$PUSH	r0, `$FRAME+$LRSAVE`($sp)
+	mtspr	256, r7			# preserve all AltiVec registers
+
+	srwi	r9, $bits, 5		# shr	\$5,%eax
+	addi	r9, r9, 6		# add	\$5,%eax
+	stw	r9, 240($out)		# mov	%eax,240(%rdx)	# AES_KEY->rounds = nbits/32+5;
+
+	slwi	r9, r9, 4		# shl	\$4,%eax
+	add	$out, $out, r9		# lea	(%rdx,%rax),%rdx
+
+	cmplwi	$dir, $bits, 0		# set decrypt direction
+	srwi	r8, $bits, 1		# shr	\$1,%r8d
+	andi.	r8, r8, 32		# and	\$32,%r8d
+	xori	r8, r8, 32		# xor	\$32,%r8d	# nbits==192?0:32
+	bl	_vpaes_schedule_core
+
+	$POP	r0,  `$FRAME+$LRSAVE`($sp)
+	li	r10,`15+6*$SIZE_T`
+	li	r11,`31+6*$SIZE_T`
+	mtspr	256, r6			# restore vrsave
+	mtlr	r0
+	xor	r3, r3, r3
+	lvx	v20,r10,$sp
+	addi	r10,r10,32
+	lvx	v21,r11,$sp
+	addi	r11,r11,32
+	lvx	v22,r10,$sp
+	addi	r10,r10,32
+	lvx	v23,r11,$sp
+	addi	r11,r11,32
+	lvx	v24,r10,$sp
+	addi	r10,r10,32
+	lvx	v25,r11,$sp
+	addi	r11,r11,32
+	lvx	v26,r10,$sp
+	addi	r10,r10,32
+	lvx	v27,r11,$sp
+	addi	r11,r11,32
+	lvx	v28,r10,$sp
+	addi	r10,r10,32
+	lvx	v29,r11,$sp
+	addi	r11,r11,32
+	lvx	v30,r10,$sp
+	lvx	v31,r11,$sp
+	addi	$sp,$sp,$FRAME
+	blr
+	.long	0
+	.byte	0,12,0x04,1,0x80,0,3,0
+	.long	0
+.size	.vpaes_set_decrypt_key,.-.vpaes_set_decrypt_key
+___
+}
+
+my $consts=1;
+foreach  (split("\n",$code)) {
+	s/\`([^\`]*)\`/eval $1/geo;
+
+	# constants table endian-specific conversion
+	if ($consts && m/\.long\s+(.+)\s+(\?[a-z]*)$/o) {
+	    my $conv=$2;
+	    my @bytes=();
+
+	    # convert to endian-agnostic format
+	    foreach (split(/,\s+/,$1)) {
+		my $l = /^0/?oct:int;
+		push @bytes,($l>>24)&0xff,($l>>16)&0xff,($l>>8)&0xff,$l&0xff;
+	    }
+
+	    # little-endian conversion
+	    if ($flavour =~ /le$/o) {
+		SWITCH: for($conv)  {
+		    /\?inv/ && do   { @bytes=map($_^0xf,@bytes); last; };
+		    /\?rev/ && do   { @bytes=reverse(@bytes);    last; }; 
+		}
+	    }
+
+	    #emit
+	    print ".byte\t",join(',',map (sprintf("0x%02x",$_),@bytes)),"\n";
+	    next;
+	}
+	$consts=0 if (m/Lconsts:/o);	# end of table
+
+	# instructions prefixed with '?' are endian-specific and need
+	# to be adjusted accordingly...
+	if ($flavour =~ /le$/o) {	# little-endian
+	    s/\?lvsr/lvsl/o or
+	    s/\?lvsl/lvsr/o or
+	    s/\?(vperm\s+v[0-9]+,\s*)(v[0-9]+,\s*)(v[0-9]+,\s*)(v[0-9]+)/$1$3$2$4/o or
+	    s/\?(vsldoi\s+v[0-9]+,\s*)(v[0-9]+,)\s*(v[0-9]+,\s*)([0-9]+)/$1$3$2 16-$4/o or
+	    s/\?(vspltw\s+v[0-9]+,\s*)(v[0-9]+,)\s*([0-9])/$1$2 3-$3/o;
+	} else {			# big-endian
+	    s/\?([a-z]+)/$1/o;
+	}
+
+	print $_,"\n";
+}
+
+close STDOUT;
diff --git a/crypto/bn/asm/ppc-mont.pl b/crypto/bn/asm/ppc-mont.pl
index f9b6992..da69c6a 100644
--- a/crypto/bn/asm/ppc-mont.pl
+++ b/crypto/bn/asm/ppc-mont.pl
@@ -325,6 +325,7 @@ Lcopy:				; copy or in-place refresh
 	.long	0
 	.byte	0,12,4,0,0x80,12,6,0
 	.long	0
+.size	.bn_mul_mont_int,.-.bn_mul_mont_int
 
 .asciz  "Montgomery Multiplication for PPC, CRYPTOGAMS by <appro\@openssl.org>"
 ___
diff --git a/crypto/bn/asm/ppc.pl b/crypto/bn/asm/ppc.pl
index 1249ce2..04df1fe 100644
--- a/crypto/bn/asm/ppc.pl
+++ b/crypto/bn/asm/ppc.pl
@@ -392,6 +392,7 @@ $data=<<EOF;
 	.long	0
 	.byte	0,12,0x14,0,0,0,2,0
 	.long	0
+.size	.bn_sqr_comba4,.-.bn_sqr_comba4
 
 #
 #	NOTE:	The following label name should be changed to
@@ -819,6 +820,7 @@ $data=<<EOF;
 	.long	0
 	.byte	0,12,0x14,0,0,0,2,0
 	.long	0
+.size	.bn_sqr_comba8,.-.bn_sqr_comba8
 
 #
 #	NOTE:	The following label name should be changed to
@@ -972,6 +974,7 @@ $data=<<EOF;
 	.long	0
 	.byte	0,12,0x14,0,0,0,3,0
 	.long	0
+.size	.bn_mul_comba4,.-.bn_mul_comba4
 
 #
 #	NOTE:	The following label name should be changed to
@@ -1510,6 +1513,7 @@ $data=<<EOF;
 	.long	0
 	.byte	0,12,0x14,0,0,0,3,0
 	.long	0
+.size	.bn_mul_comba8,.-.bn_mul_comba8
 
 #
 #	NOTE:	The following label name should be changed to
@@ -1560,6 +1564,7 @@ Lppcasm_sub_adios:
 	.long	0
 	.byte	0,12,0x14,0,0,0,4,0
 	.long	0
+.size	.bn_sub_words,.-.bn_sub_words
 
 #
 #	NOTE:	The following label name should be changed to
@@ -1605,6 +1610,7 @@ Lppcasm_add_adios:
 	.long	0
 	.byte	0,12,0x14,0,0,0,4,0
 	.long	0
+.size	.bn_add_words,.-.bn_add_words
 
 #
 #	NOTE:	The following label name should be changed to
@@ -1720,6 +1726,7 @@ Lppcasm_div9:
 	.long	0
 	.byte	0,12,0x14,0,0,0,3,0
 	.long	0
+.size	.bn_div_words,.-.bn_div_words
 
 #
 #	NOTE:	The following label name should be changed to
@@ -1761,6 +1768,7 @@ Lppcasm_sqr_adios:
 	.long	0
 	.byte	0,12,0x14,0,0,0,3,0
 	.long	0
+.size	.bn_sqr_words,.-.bn_sqr_words
 
 #
 #	NOTE:	The following label name should be changed to
@@ -1866,6 +1874,7 @@ Lppcasm_mw_OVER:
 	.long	0
 	.byte	0,12,0x14,0,0,0,4,0
 	.long	0
+.size	bn_mul_words,.-bn_mul_words
 
 #
 #	NOTE:	The following label name should be changed to
@@ -1991,6 +2000,7 @@ Lppcasm_maw_adios:
 	.long	0
 	.byte	0,12,0x14,0,0,0,4,0
 	.long	0
+.size	.bn_mul_add_words,.-.bn_mul_add_words
 	.align	4
 EOF
 $data =~ s/\`([^\`]*)\`/eval $1/gem;
diff --git a/crypto/bn/asm/ppc64-mont.pl b/crypto/bn/asm/ppc64-mont.pl
index a14e769..68e3733 100644
--- a/crypto/bn/asm/ppc64-mont.pl
+++ b/crypto/bn/asm/ppc64-mont.pl
@@ -1,7 +1,7 @@
 #!/usr/bin/env perl
 
 # ====================================================================
-# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
 # project. The module is, however, dual licensed under OpenSSL and
 # CRYPTOGAMS licenses depending on where you obtain it. For further
 # details see http://www.openssl.org/~appro/cryptogams/.
@@ -65,6 +65,14 @@
 # others alternative would be to break dependence on upper halves of
 # GPRs by sticking to 32-bit integer operations...
 
+# December 2012
+
+# Remove above mentioned dependence on GPRs' upper halves in 32-bit
+# build. No signal masking overhead, but integer instructions are
+# *more* numerous... It's still "universally" faster than 32-bit
+# ppc-mont.pl, but improvement coefficient is not as impressive
+# for longer keys...
+
 $flavour = shift;
 
 if ($flavour =~ /32/) {
@@ -110,6 +118,9 @@ $tp="r10";
 $j="r11";
 $i="r12";
 # non-volatile registers
+$c1="r19";
+$n1="r20";
+$a1="r21";
 $nap_d="r22";	# interleaved ap and np in double format
 $a0="r23";	# ap[0]
 $t0="r24";	# temporary registers
@@ -180,8 +191,8 @@ $T3a="f30";	$T3b="f31";
 #		.				.
 #		+-------------------------------+
 #		.				.
-#   -12*size_t	+-------------------------------+
-#		| 10 saved gpr, r22-r31		|
+#   -13*size_t	+-------------------------------+
+#		| 13 saved gpr, r19-r31		|
 #		.				.
 #		.				.
 #   -12*8	+-------------------------------+
@@ -215,6 +226,9 @@ $code=<<___;
 	mr	$i,$sp
 	$STUX	$sp,$sp,$tp	; alloca
 
+	$PUSH	r19,`-12*8-13*$SIZE_T`($i)
+	$PUSH	r20,`-12*8-12*$SIZE_T`($i)
+	$PUSH	r21,`-12*8-11*$SIZE_T`($i)
 	$PUSH	r22,`-12*8-10*$SIZE_T`($i)
 	$PUSH	r23,`-12*8-9*$SIZE_T`($i)
 	$PUSH	r24,`-12*8-8*$SIZE_T`($i)
@@ -237,40 +251,26 @@ $code=<<___;
 	stfd	f29,`-3*8`($i)
 	stfd	f30,`-2*8`($i)
 	stfd	f31,`-1*8`($i)
-___
-$code.=<<___ if ($SIZE_T==8);
-	ld	$a0,0($ap)	; pull ap[0] value
-	ld	$n0,0($n0)	; pull n0[0] value
-	ld	$t3,0($bp)	; bp[0]
-___
-$code.=<<___ if ($SIZE_T==4);
-	mr	$t1,$n0
-	lwz	$a0,0($ap)	; pull ap[0,1] value
-	lwz	$t0,4($ap)
-	lwz	$n0,0($t1)	; pull n0[0,1] value
-	lwz	$t1,4($t1)
-	lwz	$t3,0($bp)	; bp[0,1]
-	lwz	$t2,4($bp)
-	insrdi	$a0,$t0,32,0
-	insrdi	$n0,$t1,32,0
-	insrdi	$t3,$t2,32,0
-___
-$code.=<<___;
+
 	addi	$tp,$sp,`$FRAME+$TRANSFER+8+64`
 	li	$i,-64
 	add	$nap_d,$tp,$num
 	and	$nap_d,$nap_d,$i	; align to 64 bytes
-
-	mulld	$t7,$a0,$t3	; ap[0]*bp[0]
 	; nap_d is off by 1, because it's used with stfdu/lfdu
 	addi	$nap_d,$nap_d,-8
 	srwi	$j,$num,`3+1`	; counter register, num/2
-	mulld	$t7,$t7,$n0	; tp[0]*n0
 	addi	$j,$j,-1
 	addi	$tp,$sp,`$FRAME+$TRANSFER-8`
 	li	$carry,0
 	mtctr	$j
+___
+
+$code.=<<___ if ($SIZE_T==8);
+	ld	$a0,0($ap)		; pull ap[0] value
+	ld	$t3,0($bp)		; bp[0]
+	ld	$n0,0($n0)		; pull n0[0] value
 
+	mulld	$t7,$a0,$t3		; ap[0]*bp[0]
 	; transfer bp[0] to FPU as 4x16-bit values
 	extrdi	$t0,$t3,16,48
 	extrdi	$t1,$t3,16,32
@@ -280,6 +280,8 @@ $code.=<<___;
 	std	$t1,`$FRAME+8`($sp)
 	std	$t2,`$FRAME+16`($sp)
 	std	$t3,`$FRAME+24`($sp)
+
+	mulld	$t7,$t7,$n0		; tp[0]*n0
 	; transfer (ap[0]*bp[0])*n0 to FPU as 4x16-bit values
 	extrdi	$t4,$t7,16,48
 	extrdi	$t5,$t7,16,32
@@ -289,21 +291,61 @@ $code.=<<___;
 	std	$t5,`$FRAME+40`($sp)
 	std	$t6,`$FRAME+48`($sp)
 	std	$t7,`$FRAME+56`($sp)
-___
-$code.=<<___ if ($SIZE_T==8);
-	lwz	$t0,4($ap)		; load a[j] as 32-bit word pair
-	lwz	$t1,0($ap)
-	lwz	$t2,12($ap)		; load a[j+1] as 32-bit word pair
+
+	extrdi	$t0,$a0,32,32		; lwz	$t0,4($ap)
+	extrdi	$t1,$a0,32,0		; lwz	$t1,0($ap)
+	lwz	$t2,12($ap)		; load a[1] as 32-bit word pair
 	lwz	$t3,8($ap)
-	lwz	$t4,4($np)		; load n[j] as 32-bit word pair
+	lwz	$t4,4($np)		; load n[0] as 32-bit word pair
 	lwz	$t5,0($np)
-	lwz	$t6,12($np)		; load n[j+1] as 32-bit word pair
+	lwz	$t6,12($np)		; load n[1] as 32-bit word pair
 	lwz	$t7,8($np)
 ___
 $code.=<<___ if ($SIZE_T==4);
-	lwz	$t0,0($ap)		; load a[j..j+3] as 32-bit word pairs
-	lwz	$t1,4($ap)
-	lwz	$t2,8($ap)
+	lwz	$a0,0($ap)		; pull ap[0,1] value
+	mr	$n1,$n0
+	lwz	$a1,4($ap)
+	li	$c1,0
+	lwz	$t1,0($bp)		; bp[0,1]
+	lwz	$t3,4($bp)
+	lwz	$n0,0($n1)		; pull n0[0,1] value
+	lwz	$n1,4($n1)
+
+	mullw	$t4,$a0,$t1		; mulld ap[0]*bp[0]
+	mulhwu	$t5,$a0,$t1
+	mullw	$t6,$a1,$t1
+	mullw	$t7,$a0,$t3
+	add	$t5,$t5,$t6
+	add	$t5,$t5,$t7
+	; transfer bp[0] to FPU as 4x16-bit values
+	extrwi	$t0,$t1,16,16
+	extrwi	$t1,$t1,16,0
+	extrwi	$t2,$t3,16,16
+	extrwi	$t3,$t3,16,0
+	std	$t0,`$FRAME+0`($sp)	; yes, std in 32-bit build
+	std	$t1,`$FRAME+8`($sp)
+	std	$t2,`$FRAME+16`($sp)
+	std	$t3,`$FRAME+24`($sp)
+
+	mullw	$t0,$t4,$n0		; mulld tp[0]*n0
+	mulhwu	$t1,$t4,$n0
+	mullw	$t2,$t5,$n0
+	mullw	$t3,$t4,$n1
+	add	$t1,$t1,$t2
+	add	$t1,$t1,$t3
+	; transfer (ap[0]*bp[0])*n0 to FPU as 4x16-bit values
+	extrwi	$t4,$t0,16,16
+	extrwi	$t5,$t0,16,0
+	extrwi	$t6,$t1,16,16
+	extrwi	$t7,$t1,16,0
+	std	$t4,`$FRAME+32`($sp)	; yes, std in 32-bit build
+	std	$t5,`$FRAME+40`($sp)
+	std	$t6,`$FRAME+48`($sp)
+	std	$t7,`$FRAME+56`($sp)
+
+	mr	$t0,$a0			; lwz	$t0,0($ap)
+	mr	$t1,$a1			; lwz	$t1,4($ap)
+	lwz	$t2,8($ap)		; load a[j..j+3] as 32-bit word pairs
 	lwz	$t3,12($ap)
 	lwz	$t4,0($np)		; load n[j..j+3] as 32-bit word pairs
 	lwz	$t5,4($np)
@@ -319,7 +361,7 @@ $code.=<<___;
 	lfd	$nb,`$FRAME+40`($sp)
 	lfd	$nc,`$FRAME+48`($sp)
 	lfd	$nd,`$FRAME+56`($sp)
-	std	$t0,`$FRAME+64`($sp)
+	std	$t0,`$FRAME+64`($sp)	; yes, std even in 32-bit build
 	std	$t1,`$FRAME+72`($sp)
 	std	$t2,`$FRAME+80`($sp)
 	std	$t3,`$FRAME+88`($sp)
@@ -441,7 +483,7 @@ $code.=<<___ if ($SIZE_T==4);
 	lwz	$t7,12($np)
 ___
 $code.=<<___;
-	std	$t0,`$FRAME+64`($sp)
+	std	$t0,`$FRAME+64`($sp)	; yes, std even in 32-bit build
 	std	$t1,`$FRAME+72`($sp)
 	std	$t2,`$FRAME+80`($sp)
 	std	$t3,`$FRAME+88`($sp)
@@ -449,6 +491,9 @@ $code.=<<___;
 	std	$t5,`$FRAME+104`($sp)
 	std	$t6,`$FRAME+112`($sp)
 	std	$t7,`$FRAME+120`($sp)
+___
+if ($SIZE_T==8 or $flavour =~ /osx/) {
+$code.=<<___;
 	ld	$t0,`$FRAME+0`($sp)
 	ld	$t1,`$FRAME+8`($sp)
 	ld	$t2,`$FRAME+16`($sp)
@@ -457,6 +502,20 @@ $code.=<<___;
 	ld	$t5,`$FRAME+40`($sp)
 	ld	$t6,`$FRAME+48`($sp)
 	ld	$t7,`$FRAME+56`($sp)
+___
+} else {
+$code.=<<___;
+	lwz	$t1,`$FRAME+0`($sp)
+	lwz	$t0,`$FRAME+4`($sp)
+	lwz	$t3,`$FRAME+8`($sp)
+	lwz	$t2,`$FRAME+12`($sp)
+	lwz	$t5,`$FRAME+16`($sp)
+	lwz	$t4,`$FRAME+20`($sp)
+	lwz	$t7,`$FRAME+24`($sp)
+	lwz	$t6,`$FRAME+28`($sp)
+___
+}
+$code.=<<___;
 	lfd	$A0,`$FRAME+64`($sp)
 	lfd	$A1,`$FRAME+72`($sp)
 	lfd	$A2,`$FRAME+80`($sp)
@@ -488,7 +547,9 @@ $code.=<<___;
 	fmadd	$T0b,$A0,$bb,$dotb
 	stfd	$A2,24($nap_d)		; save a[j+1] in double format
 	stfd	$A3,32($nap_d)
-
+___
+if ($SIZE_T==8 or $flavour =~ /osx/) {
+$code.=<<___;
 	fmadd	$T1a,$A0,$bc,$T1a
 	fmadd	$T1b,$A0,$bd,$T1b
 	fmadd	$T2a,$A1,$bc,$T2a
@@ -561,11 +622,123 @@ $code.=<<___;
 	stfd	$T3b,`$FRAME+56`($sp)
 	 std	$t0,8($tp)		; tp[j-1]
 	 stdu	$t4,16($tp)		; tp[j]
+___
+} else {
+$code.=<<___;
+	fmadd	$T1a,$A0,$bc,$T1a
+	fmadd	$T1b,$A0,$bd,$T1b
+	 addc	$t0,$t0,$carry
+	 adde	$t1,$t1,$c1
+	 srwi	$carry,$t0,16
+	fmadd	$T2a,$A1,$bc,$T2a
+	fmadd	$T2b,$A1,$bd,$T2b
+	stfd	$N0,40($nap_d)		; save n[j] in double format
+	stfd	$N1,48($nap_d)
+	 srwi	$c1,$t1,16
+	 insrwi	$carry,$t1,16,0
+	fmadd	$T3a,$A2,$bc,$T3a
+	fmadd	$T3b,$A2,$bd,$T3b
+	 addc	$t2,$t2,$carry
+	 adde	$t3,$t3,$c1
+	 srwi	$carry,$t2,16
+	fmul	$dota,$A3,$bc
+	fmul	$dotb,$A3,$bd
+	stfd	$N2,56($nap_d)		; save n[j+1] in double format
+	stfdu	$N3,64($nap_d)
+	 insrwi	$t0,$t2,16,0		; 0..31 bits
+	 srwi	$c1,$t3,16
+	 insrwi	$carry,$t3,16,0
+
+	fmadd	$T1a,$N1,$na,$T1a
+	fmadd	$T1b,$N1,$nb,$T1b
+	 lwz	$t3,`$FRAME+32`($sp)	; permuted $t1
+	 lwz	$t2,`$FRAME+36`($sp)	; permuted $t0
+	 addc	$t4,$t4,$carry
+	 adde	$t5,$t5,$c1
+	 srwi	$carry,$t4,16
+	fmadd	$T2a,$N2,$na,$T2a
+	fmadd	$T2b,$N2,$nb,$T2b
+	 srwi	$c1,$t5,16
+	 insrwi	$carry,$t5,16,0
+	fmadd	$T3a,$N3,$na,$T3a
+	fmadd	$T3b,$N3,$nb,$T3b
+	 addc	$t6,$t6,$carry
+	 adde	$t7,$t7,$c1
+	 srwi	$carry,$t6,16
+	fmadd	$T0a,$N0,$na,$T0a
+	fmadd	$T0b,$N0,$nb,$T0b
+	 insrwi	$t4,$t6,16,0		; 32..63 bits
+	 srwi	$c1,$t7,16
+	 insrwi	$carry,$t7,16,0
+
+	fmadd	$T1a,$N0,$nc,$T1a
+	fmadd	$T1b,$N0,$nd,$T1b
+	 lwz	$t7,`$FRAME+40`($sp)	; permuted $t3
+	 lwz	$t6,`$FRAME+44`($sp)	; permuted $t2
+	 addc	$t2,$t2,$carry
+	 adde	$t3,$t3,$c1
+	 srwi	$carry,$t2,16
+	fmadd	$T2a,$N1,$nc,$T2a
+	fmadd	$T2b,$N1,$nd,$T2b
+	 stw	$t0,12($tp)		; tp[j-1]
+	 stw	$t4,8($tp)
+	 srwi	$c1,$t3,16
+	 insrwi	$carry,$t3,16,0
+	fmadd	$T3a,$N2,$nc,$T3a
+	fmadd	$T3b,$N2,$nd,$T3b
+	 lwz	$t1,`$FRAME+48`($sp)	; permuted $t5
+	 lwz	$t0,`$FRAME+52`($sp)	; permuted $t4
+	 addc	$t6,$t6,$carry
+	 adde	$t7,$t7,$c1
+	 srwi	$carry,$t6,16
+	fmadd	$dota,$N3,$nc,$dota
+	fmadd	$dotb,$N3,$nd,$dotb
+	 insrwi	$t2,$t6,16,0		; 64..95 bits
+	 srwi	$c1,$t7,16
+	 insrwi	$carry,$t7,16,0
+
+	fctid	$T0a,$T0a
+	fctid	$T0b,$T0b
+	 lwz	$t5,`$FRAME+56`($sp)	; permuted $t7
+	 lwz	$t4,`$FRAME+60`($sp)	; permuted $t6
+	 addc	$t0,$t0,$carry
+	 adde	$t1,$t1,$c1
+	 srwi	$carry,$t0,16
+	fctid	$T1a,$T1a
+	fctid	$T1b,$T1b
+	 srwi	$c1,$t1,16
+	 insrwi	$carry,$t1,16,0
+	fctid	$T2a,$T2a
+	fctid	$T2b,$T2b
+	 addc	$t4,$t4,$carry
+	 adde	$t5,$t5,$c1
+	 srwi	$carry,$t4,16
+	fctid	$T3a,$T3a
+	fctid	$T3b,$T3b
+	 insrwi	$t0,$t4,16,0		; 96..127 bits
+	 srwi	$c1,$t5,16
+	 insrwi	$carry,$t5,16,0
+
+	stfd	$T0a,`$FRAME+0`($sp)
+	stfd	$T0b,`$FRAME+8`($sp)
+	stfd	$T1a,`$FRAME+16`($sp)
+	stfd	$T1b,`$FRAME+24`($sp)
+	stfd	$T2a,`$FRAME+32`($sp)
+	stfd	$T2b,`$FRAME+40`($sp)
+	stfd	$T3a,`$FRAME+48`($sp)
+	stfd	$T3b,`$FRAME+56`($sp)
+	 stw	$t2,20($tp)		; tp[j]
+	 stwu	$t0,16($tp)
+___
+}
+$code.=<<___;
 	bdnz-	L1st
 
 	fctid	$dota,$dota
 	fctid	$dotb,$dotb
-
+___
+if ($SIZE_T==8 or $flavour =~ /osx/) {
+$code.=<<___;
 	ld	$t0,`$FRAME+0`($sp)
 	ld	$t1,`$FRAME+8`($sp)
 	ld	$t2,`$FRAME+16`($sp)
@@ -611,33 +784,117 @@ $code.=<<___;
 	insrdi	$t6,$t7,48,0
 	srdi	$ovf,$t7,48
 	std	$t6,8($tp)		; tp[num-1]
+___
+} else {
+$code.=<<___;
+	lwz	$t1,`$FRAME+0`($sp)
+	lwz	$t0,`$FRAME+4`($sp)
+	lwz	$t3,`$FRAME+8`($sp)
+	lwz	$t2,`$FRAME+12`($sp)
+	lwz	$t5,`$FRAME+16`($sp)
+	lwz	$t4,`$FRAME+20`($sp)
+	lwz	$t7,`$FRAME+24`($sp)
+	lwz	$t6,`$FRAME+28`($sp)
+	stfd	$dota,`$FRAME+64`($sp)
+	stfd	$dotb,`$FRAME+72`($sp)
 
+	addc	$t0,$t0,$carry
+	adde	$t1,$t1,$c1
+	srwi	$carry,$t0,16
+	insrwi	$carry,$t1,16,0
+	srwi	$c1,$t1,16
+	addc	$t2,$t2,$carry
+	adde	$t3,$t3,$c1
+	srwi	$carry,$t2,16
+	 insrwi	$t0,$t2,16,0		; 0..31 bits
+	insrwi	$carry,$t3,16,0
+	srwi	$c1,$t3,16
+	addc	$t4,$t4,$carry
+	adde	$t5,$t5,$c1
+	srwi	$carry,$t4,16
+	insrwi	$carry,$t5,16,0
+	srwi	$c1,$t5,16
+	addc	$t6,$t6,$carry
+	adde	$t7,$t7,$c1
+	srwi	$carry,$t6,16
+	 insrwi	$t4,$t6,16,0		; 32..63 bits
+	insrwi	$carry,$t7,16,0
+	srwi	$c1,$t7,16
+	 stw	$t0,12($tp)		; tp[j-1]
+	 stw	$t4,8($tp)
+
+	lwz	$t3,`$FRAME+32`($sp)	; permuted $t1
+	lwz	$t2,`$FRAME+36`($sp)	; permuted $t0
+	lwz	$t7,`$FRAME+40`($sp)	; permuted $t3
+	lwz	$t6,`$FRAME+44`($sp)	; permuted $t2
+	lwz	$t1,`$FRAME+48`($sp)	; permuted $t5
+	lwz	$t0,`$FRAME+52`($sp)	; permuted $t4
+	lwz	$t5,`$FRAME+56`($sp)	; permuted $t7
+	lwz	$t4,`$FRAME+60`($sp)	; permuted $t6
+
+	addc	$t2,$t2,$carry
+	adde	$t3,$t3,$c1
+	srwi	$carry,$t2,16
+	insrwi	$carry,$t3,16,0
+	srwi	$c1,$t3,16
+	addc	$t6,$t6,$carry
+	adde	$t7,$t7,$c1
+	srwi	$carry,$t6,16
+	 insrwi	$t2,$t6,16,0		; 64..95 bits
+	insrwi	$carry,$t7,16,0
+	srwi	$c1,$t7,16
+	addc	$t0,$t0,$carry
+	adde	$t1,$t1,$c1
+	srwi	$carry,$t0,16
+	insrwi	$carry,$t1,16,0
+	srwi	$c1,$t1,16
+	addc	$t4,$t4,$carry
+	adde	$t5,$t5,$c1
+	srwi	$carry,$t4,16
+	 insrwi	$t0,$t4,16,0		; 96..127 bits
+	insrwi	$carry,$t5,16,0
+	srwi	$c1,$t5,16
+	 stw	$t2,20($tp)		; tp[j]
+	 stwu	$t0,16($tp)
+
+	lwz	$t7,`$FRAME+64`($sp)
+	lwz	$t6,`$FRAME+68`($sp)
+	lwz	$t5,`$FRAME+72`($sp)
+	lwz	$t4,`$FRAME+76`($sp)
+
+	addc	$t6,$t6,$carry
+	adde	$t7,$t7,$c1
+	srwi	$carry,$t6,16
+	insrwi	$carry,$t7,16,0
+	srwi	$c1,$t7,16
+	addc	$t4,$t4,$carry
+	adde	$t5,$t5,$c1
+
+	insrwi	$t6,$t4,16,0
+	srwi	$t4,$t4,16
+	insrwi	$t4,$t5,16,0
+	srwi	$ovf,$t5,16
+	stw	$t6,12($tp)		; tp[num-1]
+	stw	$t4,8($tp)
+___
+}
+$code.=<<___;
 	slwi	$t7,$num,2
 	subf	$nap_d,$t7,$nap_d	; rewind pointer
 
 	li	$i,8			; i=1
 .align	5
 Louter:
-___
-$code.=<<___ if ($SIZE_T==8);
-	ldx	$t3,$bp,$i	; bp[i]
-___
-$code.=<<___ if ($SIZE_T==4);
-	add	$t0,$bp,$i
-	lwz	$t3,0($t0)		; bp[i,i+1]
-	lwz	$t0,4($t0)
-	insrdi	$t3,$t0,32,0
-___
-$code.=<<___;
-	ld	$t6,`$FRAME+$TRANSFER+8`($sp)	; tp[0]
-	mulld	$t7,$a0,$t3	; ap[0]*bp[i]
-
 	addi	$tp,$sp,`$FRAME+$TRANSFER`
-	add	$t7,$t7,$t6	; ap[0]*bp[i]+tp[0]
 	li	$carry,0
-	mulld	$t7,$t7,$n0	; tp[0]*n0
 	mtctr	$j
+___
+$code.=<<___ if ($SIZE_T==8);
+	ldx	$t3,$bp,$i		; bp[i]
 
+	ld	$t6,`$FRAME+$TRANSFER+8`($sp)	; tp[0]
+	mulld	$t7,$a0,$t3		; ap[0]*bp[i]
+	add	$t7,$t7,$t6		; ap[0]*bp[i]+tp[0]
 	; transfer bp[i] to FPU as 4x16-bit values
 	extrdi	$t0,$t3,16,48
 	extrdi	$t1,$t3,16,32
@@ -647,6 +904,8 @@ $code.=<<___;
 	std	$t1,`$FRAME+8`($sp)
 	std	$t2,`$FRAME+16`($sp)
 	std	$t3,`$FRAME+24`($sp)
+
+	mulld	$t7,$t7,$n0		; tp[0]*n0
 	; transfer (ap[0]*bp[i]+tp[0])*n0 to FPU as 4x16-bit values
 	extrdi	$t4,$t7,16,48
 	extrdi	$t5,$t7,16,32
@@ -656,7 +915,50 @@ $code.=<<___;
 	std	$t5,`$FRAME+40`($sp)
 	std	$t6,`$FRAME+48`($sp)
 	std	$t7,`$FRAME+56`($sp)
+___
+$code.=<<___ if ($SIZE_T==4);
+	add	$t0,$bp,$i
+	li	$c1,0
+	lwz	$t1,0($t0)		; bp[i,i+1]
+	lwz	$t3,4($t0)
+
+	mullw	$t4,$a0,$t1		; ap[0]*bp[i]
+	lwz	$t0,`$FRAME+$TRANSFER+8+4`($sp)	; tp[0]
+	mulhwu	$t5,$a0,$t1
+	lwz	$t2,`$FRAME+$TRANSFER+8`($sp)	; tp[0]
+	mullw	$t6,$a1,$t1
+	mullw	$t7,$a0,$t3
+	add	$t5,$t5,$t6
+	add	$t5,$t5,$t7
+	addc	$t4,$t4,$t0		; ap[0]*bp[i]+tp[0]
+	adde	$t5,$t5,$t2
+	; transfer bp[i] to FPU as 4x16-bit values
+	extrwi	$t0,$t1,16,16
+	extrwi	$t1,$t1,16,0
+	extrwi	$t2,$t3,16,16
+	extrwi	$t3,$t3,16,0
+	std	$t0,`$FRAME+0`($sp)	; yes, std in 32-bit build
+	std	$t1,`$FRAME+8`($sp)
+	std	$t2,`$FRAME+16`($sp)
+	std	$t3,`$FRAME+24`($sp)
 
+	mullw	$t0,$t4,$n0		; mulld tp[0]*n0
+	mulhwu	$t1,$t4,$n0
+	mullw	$t2,$t5,$n0
+	mullw	$t3,$t4,$n1
+	add	$t1,$t1,$t2
+	add	$t1,$t1,$t3
+	; transfer (ap[0]*bp[i]+tp[0])*n0 to FPU as 4x16-bit values
+	extrwi	$t4,$t0,16,16
+	extrwi	$t5,$t0,16,0
+	extrwi	$t6,$t1,16,16
+	extrwi	$t7,$t1,16,0
+	std	$t4,`$FRAME+32`($sp)	; yes, std in 32-bit build
+	std	$t5,`$FRAME+40`($sp)
+	std	$t6,`$FRAME+48`($sp)
+	std	$t7,`$FRAME+56`($sp)
+___
+$code.=<<___;
 	lfd	$A0,8($nap_d)		; load a[j] in double format
 	lfd	$A1,16($nap_d)
 	lfd	$A2,24($nap_d)		; load a[j+1] in double format
@@ -769,7 +1071,9 @@ Linner:
 	fmul	$dotb,$A3,$bd
 	 lfd	$A2,24($nap_d)		; load a[j+1] in double format
 	 lfd	$A3,32($nap_d)
-
+___
+if ($SIZE_T==8 or $flavour =~ /osx/) {
+$code.=<<___;
 	fmadd	$T1a,$N1,$na,$T1a
 	fmadd	$T1b,$N1,$nb,$T1b
 	 ld	$t0,`$FRAME+0`($sp)
@@ -856,10 +1160,131 @@ $code.=<<___;
 	 addze	$carry,$carry
 	 std	$t3,-16($tp)		; tp[j-1]
 	 std	$t5,-8($tp)		; tp[j]
+___
+} else {
+$code.=<<___;
+	fmadd	$T1a,$N1,$na,$T1a
+	fmadd	$T1b,$N1,$nb,$T1b
+	 lwz	$t1,`$FRAME+0`($sp)
+	 lwz	$t0,`$FRAME+4`($sp)
+	fmadd	$T2a,$N2,$na,$T2a
+	fmadd	$T2b,$N2,$nb,$T2b
+	 lwz	$t3,`$FRAME+8`($sp)
+	 lwz	$t2,`$FRAME+12`($sp)
+	fmadd	$T3a,$N3,$na,$T3a
+	fmadd	$T3b,$N3,$nb,$T3b
+	 lwz	$t5,`$FRAME+16`($sp)
+	 lwz	$t4,`$FRAME+20`($sp)
+	 addc	$t0,$t0,$carry
+	 adde	$t1,$t1,$c1
+	 srwi	$carry,$t0,16
+	fmadd	$T0a,$N0,$na,$T0a
+	fmadd	$T0b,$N0,$nb,$T0b
+	 lwz	$t7,`$FRAME+24`($sp)
+	 lwz	$t6,`$FRAME+28`($sp)
+	 srwi	$c1,$t1,16
+	 insrwi	$carry,$t1,16,0
+
+	fmadd	$T1a,$N0,$nc,$T1a
+	fmadd	$T1b,$N0,$nd,$T1b
+	 addc	$t2,$t2,$carry
+	 adde	$t3,$t3,$c1
+	 srwi	$carry,$t2,16
+	fmadd	$T2a,$N1,$nc,$T2a
+	fmadd	$T2b,$N1,$nd,$T2b
+	 insrwi	$t0,$t2,16,0		; 0..31 bits
+	 srwi	$c1,$t3,16
+	 insrwi	$carry,$t3,16,0
+	fmadd	$T3a,$N2,$nc,$T3a
+	fmadd	$T3b,$N2,$nd,$T3b
+	 lwz	$t2,12($tp)		; tp[j]
+	 lwz	$t3,8($tp)
+	 addc	$t4,$t4,$carry
+	 adde	$t5,$t5,$c1
+	 srwi	$carry,$t4,16
+	fmadd	$dota,$N3,$nc,$dota
+	fmadd	$dotb,$N3,$nd,$dotb
+	 srwi	$c1,$t5,16
+	 insrwi	$carry,$t5,16,0
+
+	fctid	$T0a,$T0a
+	 addc	$t6,$t6,$carry
+	 adde	$t7,$t7,$c1
+	 srwi	$carry,$t6,16
+	fctid	$T0b,$T0b
+	 insrwi	$t4,$t6,16,0		; 32..63 bits
+	 srwi	$c1,$t7,16
+	 insrwi	$carry,$t7,16,0
+	fctid	$T1a,$T1a
+	 addc	$t0,$t0,$t2
+	 adde	$t4,$t4,$t3
+	 lwz	$t3,`$FRAME+32`($sp)	; permuted $t1
+	 lwz	$t2,`$FRAME+36`($sp)	; permuted $t0
+	fctid	$T1b,$T1b
+	 addze	$carry,$carry
+	 addze	$c1,$c1
+	 stw	$t0,4($tp)		; tp[j-1]
+	 stw	$t4,0($tp)
+	fctid	$T2a,$T2a
+	 addc	$t2,$t2,$carry
+	 adde	$t3,$t3,$c1
+	 srwi	$carry,$t2,16
+	 lwz	$t7,`$FRAME+40`($sp)	; permuted $t3
+	 lwz	$t6,`$FRAME+44`($sp)	; permuted $t2
+	fctid	$T2b,$T2b
+	 srwi	$c1,$t3,16
+	 insrwi	$carry,$t3,16,0
+	 lwz	$t1,`$FRAME+48`($sp)	; permuted $t5
+	 lwz	$t0,`$FRAME+52`($sp)	; permuted $t4
+	fctid	$T3a,$T3a
+	 addc	$t6,$t6,$carry
+	 adde	$t7,$t7,$c1
+	 srwi	$carry,$t6,16
+	 lwz	$t5,`$FRAME+56`($sp)	; permuted $t7
+	 lwz	$t4,`$FRAME+60`($sp)	; permuted $t6
+	fctid	$T3b,$T3b
+
+	 insrwi	$t2,$t6,16,0		; 64..95 bits
+	insrwi	$carry,$t7,16,0
+	srwi	$c1,$t7,16
+	 lwz	$t6,20($tp)
+	 lwzu	$t7,16($tp)
+	addc	$t0,$t0,$carry
+	 stfd	$T0a,`$FRAME+0`($sp)
+	adde	$t1,$t1,$c1
+	srwi	$carry,$t0,16
+	 stfd	$T0b,`$FRAME+8`($sp)
+	insrwi	$carry,$t1,16,0
+	srwi	$c1,$t1,16
+	addc	$t4,$t4,$carry
+	 stfd	$T1a,`$FRAME+16`($sp)
+	adde	$t5,$t5,$c1
+	srwi	$carry,$t4,16
+	 insrwi	$t0,$t4,16,0		; 96..127 bits
+	 stfd	$T1b,`$FRAME+24`($sp)
+	insrwi	$carry,$t5,16,0
+	srwi	$c1,$t5,16
+
+	addc	$t2,$t2,$t6
+	 stfd	$T2a,`$FRAME+32`($sp)
+	adde	$t0,$t0,$t7
+	 stfd	$T2b,`$FRAME+40`($sp)
+	addze	$carry,$carry
+	 stfd	$T3a,`$FRAME+48`($sp)
+	addze	$c1,$c1
+	 stfd	$T3b,`$FRAME+56`($sp)
+	 stw	$t2,-4($tp)		; tp[j]
+	 stw	$t0,-8($tp)
+___
+}
+$code.=<<___;
 	bdnz-	Linner
 
 	fctid	$dota,$dota
 	fctid	$dotb,$dotb
+___
+if ($SIZE_T==8 or $flavour =~ /osx/) {
+$code.=<<___;
 	ld	$t0,`$FRAME+0`($sp)
 	ld	$t1,`$FRAME+8`($sp)
 	ld	$t2,`$FRAME+16`($sp)
@@ -926,7 +1351,116 @@ $code.=<<___;
 	insrdi	$t6,$t7,48,0
 	srdi	$ovf,$t7,48
 	std	$t6,0($tp)		; tp[num-1]
+___
+} else {
+$code.=<<___;
+	lwz	$t1,`$FRAME+0`($sp)
+	lwz	$t0,`$FRAME+4`($sp)
+	lwz	$t3,`$FRAME+8`($sp)
+	lwz	$t2,`$FRAME+12`($sp)
+	lwz	$t5,`$FRAME+16`($sp)
+	lwz	$t4,`$FRAME+20`($sp)
+	lwz	$t7,`$FRAME+24`($sp)
+	lwz	$t6,`$FRAME+28`($sp)
+	stfd	$dota,`$FRAME+64`($sp)
+	stfd	$dotb,`$FRAME+72`($sp)
 
+	addc	$t0,$t0,$carry
+	adde	$t1,$t1,$c1
+	srwi	$carry,$t0,16
+	insrwi	$carry,$t1,16,0
+	srwi	$c1,$t1,16
+	addc	$t2,$t2,$carry
+	adde	$t3,$t3,$c1
+	srwi	$carry,$t2,16
+	 insrwi	$t0,$t2,16,0		; 0..31 bits
+	 lwz	$t2,12($tp)		; tp[j]
+	insrwi	$carry,$t3,16,0
+	srwi	$c1,$t3,16
+	 lwz	$t3,8($tp)
+	addc	$t4,$t4,$carry
+	adde	$t5,$t5,$c1
+	srwi	$carry,$t4,16
+	insrwi	$carry,$t5,16,0
+	srwi	$c1,$t5,16
+	addc	$t6,$t6,$carry
+	adde	$t7,$t7,$c1
+	srwi	$carry,$t6,16
+	 insrwi	$t4,$t6,16,0		; 32..63 bits
+	insrwi	$carry,$t7,16,0
+	srwi	$c1,$t7,16
+
+	addc	$t0,$t0,$t2
+	adde	$t4,$t4,$t3
+	addze	$carry,$carry
+	addze	$c1,$c1
+	 stw	$t0,4($tp)		; tp[j-1]
+	 stw	$t4,0($tp)
+
+	lwz	$t3,`$FRAME+32`($sp)	; permuted $t1
+	lwz	$t2,`$FRAME+36`($sp)	; permuted $t0
+	lwz	$t7,`$FRAME+40`($sp)	; permuted $t3
+	lwz	$t6,`$FRAME+44`($sp)	; permuted $t2
+	lwz	$t1,`$FRAME+48`($sp)	; permuted $t5
+	lwz	$t0,`$FRAME+52`($sp)	; permuted $t4
+	lwz	$t5,`$FRAME+56`($sp)	; permuted $t7
+	lwz	$t4,`$FRAME+60`($sp)	; permuted $t6
+
+	addc	$t2,$t2,$carry
+	adde	$t3,$t3,$c1
+	srwi	$carry,$t2,16
+	insrwi	$carry,$t3,16,0
+	srwi	$c1,$t3,16
+	addc	$t6,$t6,$carry
+	adde	$t7,$t7,$c1
+	srwi	$carry,$t6,16
+	 insrwi	$t2,$t6,16,0		; 64..95 bits
+	 lwz	$t6,20($tp)
+	insrwi	$carry,$t7,16,0
+	srwi	$c1,$t7,16
+	 lwzu	$t7,16($tp)
+	addc	$t0,$t0,$carry
+	adde	$t1,$t1,$c1
+	srwi	$carry,$t0,16
+	insrwi	$carry,$t1,16,0
+	srwi	$c1,$t1,16
+	addc	$t4,$t4,$carry
+	adde	$t5,$t5,$c1
+	srwi	$carry,$t4,16
+	 insrwi	$t0,$t4,16,0		; 96..127 bits
+	insrwi	$carry,$t5,16,0
+	srwi	$c1,$t5,16
+
+	addc	$t2,$t2,$t6
+	adde	$t0,$t0,$t7
+	 lwz	$t7,`$FRAME+64`($sp)
+	 lwz	$t6,`$FRAME+68`($sp)
+	addze	$carry,$carry
+	addze	$c1,$c1
+	 lwz	$t5,`$FRAME+72`($sp)
+	 lwz	$t4,`$FRAME+76`($sp)
+
+	addc	$t6,$t6,$carry
+	adde	$t7,$t7,$c1
+	 stw	$t2,-4($tp)		; tp[j]
+	 stw	$t0,-8($tp)
+	addc	$t6,$t6,$ovf
+	addze	$t7,$t7
+	srwi	$carry,$t6,16
+	insrwi	$carry,$t7,16,0
+	srwi	$c1,$t7,16
+	addc	$t4,$t4,$carry
+	adde	$t5,$t5,$c1
+
+	insrwi	$t6,$t4,16,0
+	srwi	$t4,$t4,16
+	insrwi	$t4,$t5,16,0
+	srwi	$ovf,$t5,16
+	stw	$t6,4($tp)		; tp[num-1]
+	stw	$t4,0($tp)
+___
+}
+$code.=<<___;
 	slwi	$t7,$num,2
 	addi	$i,$i,8
 	subf	$nap_d,$t7,$nap_d	; rewind pointer
@@ -994,14 +1528,14 @@ $code.=<<___ if ($SIZE_T==4);
 	mtctr	$j
 
 .align	4
-Lsub:	ld	$t0,8($tp)	; load tp[j..j+3] in 64-bit word order
-	ldu	$t2,16($tp)
+Lsub:	lwz	$t0,12($tp)	; load tp[j..j+3] in 64-bit word order
+	lwz	$t1,8($tp)
+	lwz	$t2,20($tp)
+	lwzu	$t3,16($tp)
 	lwz	$t4,4($np)	; load np[j..j+3] in 32-bit word order
 	lwz	$t5,8($np)
 	lwz	$t6,12($np)
 	lwzu	$t7,16($np)
-	extrdi	$t1,$t0,32,0
-	extrdi	$t3,$t2,32,0
 	subfe	$t4,$t4,$t0	; tp[j]-np[j]
 	 stw	$t0,4($ap)	; save tp[j..j+3] in 32-bit word order
 	subfe	$t5,$t5,$t1	; tp[j+1]-np[j+1]
@@ -1052,6 +1586,9 @@ ___
 $code.=<<___;
 	$POP	$i,0($sp)
 	li	r3,1	; signal "handled"
+	$POP	r19,`-12*8-13*$SIZE_T`($i)
+	$POP	r20,`-12*8-12*$SIZE_T`($i)
+	$POP	r21,`-12*8-11*$SIZE_T`($i)
 	$POP	r22,`-12*8-10*$SIZE_T`($i)
 	$POP	r23,`-12*8-9*$SIZE_T`($i)
 	$POP	r24,`-12*8-8*$SIZE_T`($i)
@@ -1077,8 +1614,9 @@ $code.=<<___;
 	mr	$sp,$i
 	blr
 	.long	0
-	.byte	0,12,4,0,0x8c,10,6,0
+	.byte	0,12,4,0,0x8c,13,6,0
 	.long	0
+.size	.$fname,.-.$fname
 
 .asciz  "Montgomery Multiplication for PPC64, CRYPTOGAMS by <appro\@openssl.org>"
 ___
diff --git a/crypto/evp/e_aes.c b/crypto/evp/e_aes.c
index 1bfb5d9..51137fd 100644
--- a/crypto/evp/e_aes.c
+++ b/crypto/evp/e_aes.c
@@ -153,6 +153,20 @@ void AES_xts_decrypt(const char *inp,char *out,size_t len,
 			const unsigned char iv[16]);
 #endif
 
+#if	defined(OPENSSL_CPUID_OBJ) && (defined(__powerpc__) || defined(__ppc__) || defined(_ARCH_PPC))
+# include "ppc_arch.h"
+# ifdef VPAES_ASM
+#  define VPAES_CAPABLE	(OPENSSL_ppccap_P & PPC_ALTIVEC)
+# endif
+# define HWAES_CAPABLE	(OPENSSL_ppccap_P & PPC_CRYPTO207)
+# define HWAES_set_encrypt_key aes_p8_set_encrypt_key
+# define HWAES_set_decrypt_key aes_p8_set_decrypt_key
+# define HWAES_encrypt aes_p8_encrypt
+# define HWAES_decrypt aes_p8_decrypt
+# define HWAES_cbc_encrypt aes_p8_cbc_encrypt
+# define HWAES_ctr32_encrypt_blocks aes_p8_ctr32_encrypt_blocks
+#endif
+
 #if	defined(AES_ASM) && !defined(I386_ONLY) &&	(  \
 	((defined(__i386)	|| defined(__i386__)	|| \
 	  defined(_M_IX86)) && defined(OPENSSL_IA32_SSE2))|| \
diff --git a/crypto/modes/Makefile b/crypto/modes/Makefile
index c825b12..e684e02 100644
--- a/crypto/modes/Makefile
+++ b/crypto/modes/Makefile
@@ -56,6 +56,10 @@ ghash-alpha.s:	asm/ghash-alpha.pl
 	$(PERL) $< | $(CC) -E - | tee $@ > /dev/null
 ghash-parisc.s:	asm/ghash-parisc.pl
 	$(PERL) asm/ghash-parisc.pl $(PERLASM_SCHEME) $@
+ghashv8-armx.S:	asm/ghashv8-armx.pl
+	$(PERL) asm/ghashv8-armx.pl $(PERLASM_SCHEME) $@
+ghashp8-ppc.s:	asm/ghashp8-ppc.pl
+	$(PERL) asm/ghashp8-ppc.pl $(PERLASM_SCHEME) $@
 
 # GNU make "catch all"
 ghash-%.S:	asm/ghash-%.pl;	$(PERL) $< $(PERLASM_SCHEME) $@
diff --git a/crypto/modes/asm/ghashp8-ppc.pl b/crypto/modes/asm/ghashp8-ppc.pl
new file mode 100755
index 0000000..e76a58c
--- /dev/null
+++ b/crypto/modes/asm/ghashp8-ppc.pl
@@ -0,0 +1,234 @@
+#!/usr/bin/env perl
+#
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+#
+# GHASH for for PowerISA v2.07.
+#
+# July 2014
+#
+# Accurate performance measurements are problematic, because it's
+# always virtualized setup with possibly throttled processor.
+# Relative comparison is therefore more informative. This initial
+# version is ~2.1x slower than hardware-assisted AES-128-CTR, ~12x
+# faster than "4-bit" integer-only compiler-generated 64-bit code.
+# "Initial version" means that there is room for futher improvement.
+
+$flavour=shift;
+$output =shift;
+
+if ($flavour =~ /64/) {
+	$SIZE_T=8;
+	$LRSAVE=2*$SIZE_T;
+	$STU="stdu";
+	$POP="ld";
+	$PUSH="std";
+} elsif ($flavour =~ /32/) {
+	$SIZE_T=4;
+	$LRSAVE=$SIZE_T;
+	$STU="stwu";
+	$POP="lwz";
+	$PUSH="stw";
+} else { die "nonsense $flavour"; }
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or
+( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or
+die "can't locate ppc-xlate.pl";
+
+open STDOUT,"| $^X $xlate $flavour $output" || die "can't call $xlate: $!";
+
+my ($Xip,$Htbl,$inp,$len)=map("r$_",(3..6));	# argument block
+
+my ($Xl,$Xm,$Xh,$IN)=map("v$_",(0..3));
+my ($zero,$t0,$t1,$t2,$xC2,$H,$Hh,$Hl,$lemask)=map("v$_",(4..12));
+my $vrsave="r12";
+
+$code=<<___;
+.machine	"any"
+
+.text
+
+.globl	.gcm_init_p8
+.align	5
+.gcm_init_p8:
+	lis		r0,0xfff0
+	li		r8,0x10
+	mfspr		$vrsave,256
+	li		r9,0x20
+	mtspr		256,r0
+	li		r10,0x30
+	lvx_u		$H,0,r4			# load H
+
+	vspltisb	$xC2,-16		# 0xf0
+	vspltisb	$t0,1			# one
+	vaddubm		$xC2,$xC2,$xC2		# 0xe0
+	vxor		$zero,$zero,$zero
+	vor		$xC2,$xC2,$t0		# 0xe1
+	vsldoi		$xC2,$xC2,$zero,15	# 0xe1...
+	vsldoi		$t1,$zero,$t0,1		# ...1
+	vaddubm		$xC2,$xC2,$xC2		# 0xc2...
+	vspltisb	$t2,7
+	vor		$xC2,$xC2,$t1		# 0xc2....01
+	vspltb		$t1,$H,0		# most significant byte
+	vsl		$H,$H,$t0		# H<<=1
+	vsrab		$t1,$t1,$t2		# broadcast carry bit
+	vand		$t1,$t1,$xC2
+	vxor		$H,$H,$t1		# twisted H
+
+	vsldoi		$H,$H,$H,8		# twist even more ...
+	vsldoi		$xC2,$zero,$xC2,8	# 0xc2.0
+	vsldoi		$Hl,$zero,$H,8		# ... and split
+	vsldoi		$Hh,$H,$zero,8
+
+	stvx_u		$xC2,0,r3		# save pre-computed table
+	stvx_u		$Hl,r8,r3
+	stvx_u		$H, r9,r3
+	stvx_u		$Hh,r10,r3
+
+	mtspr		256,$vrsave
+	blr
+	.long		0
+	.byte		0,12,0x14,0,0,0,2,0
+	.long		0
+.size	.gcm_init_p8,.-.gcm_init_p8
+
+.globl	.gcm_gmult_p8
+.align	5
+.gcm_gmult_p8:
+	lis		r0,0xfff8
+	li		r8,0x10
+	mfspr		$vrsave,256
+	li		r9,0x20
+	mtspr		256,r0
+	li		r10,0x30
+	lvx_u		$IN,0,$Xip		# load Xi
+
+	lvx_u		$Hl,r8,$Htbl		# load pre-computed table
+	 le?lvsl	$lemask,r0,r0
+	lvx_u		$H, r9,$Htbl
+	 le?vspltisb	$t0,0x07
+	lvx_u		$Hh,r10,$Htbl
+	 le?vxor	$lemask,$lemask,$t0
+	lvx_u		$xC2,0,$Htbl
+	 le?vperm	$IN,$IN,$IN,$lemask
+	vxor		$zero,$zero,$zero
+
+	vpmsumd		$Xl,$IN,$Hl		# H.lo<6C>Xi.lo
+	vpmsumd		$Xm,$IN,$H		# H.hi<68>Xi.lo+H.lo<6C>Xi.hi
+	vpmsumd		$Xh,$IN,$Hh		# H.hi<68>Xi.hi
+
+	vpmsumd		$t2,$Xl,$xC2		# 1st phase
+
+	vsldoi		$t0,$Xm,$zero,8
+	vsldoi		$t1,$zero,$Xm,8
+	vxor		$Xl,$Xl,$t0
+	vxor		$Xh,$Xh,$t1
+
+	vsldoi		$Xl,$Xl,$Xl,8
+	vxor		$Xl,$Xl,$t2
+
+	vsldoi		$t1,$Xl,$Xl,8		# 2nd phase
+	vpmsumd		$Xl,$Xl,$xC2
+	vxor		$t1,$t1,$Xh
+	vxor		$Xl,$Xl,$t1
+
+	le?vperm	$Xl,$Xl,$Xl,$lemask
+	stvx_u		$Xl,0,$Xip		# write out Xi
+
+	mtspr		256,$vrsave
+	blr
+	.long		0
+	.byte		0,12,0x14,0,0,0,2,0
+	.long		0
+.size	.gcm_gmult_p8,.-.gcm_gmult_p8
+
+.globl	.gcm_ghash_p8
+.align	5
+.gcm_ghash_p8:
+	lis		r0,0xfff8
+	li		r8,0x10
+	mfspr		$vrsave,256
+	li		r9,0x20
+	mtspr		256,r0
+	li		r10,0x30
+	lvx_u		$Xl,0,$Xip		# load Xi
+
+	lvx_u		$Hl,r8,$Htbl		# load pre-computed table
+	 le?lvsl	$lemask,r0,r0
+	lvx_u		$H, r9,$Htbl
+	 le?vspltisb	$t0,0x07
+	lvx_u		$Hh,r10,$Htbl
+	 le?vxor	$lemask,$lemask,$t0
+	lvx_u		$xC2,0,$Htbl
+	 le?vperm	$Xl,$Xl,$Xl,$lemask
+	vxor		$zero,$zero,$zero
+
+	lvx_u		$IN,0,$inp
+	addi		$inp,$inp,16
+	subi		$len,$len,16
+	 le?vperm	$IN,$IN,$IN,$lemask
+	vxor		$IN,$IN,$Xl
+	b		Loop
+
+.align	5
+Loop:
+	 subic		$len,$len,16
+	vpmsumd		$Xl,$IN,$Hl		# H.lo<6C>Xi.lo
+	 subfe.		r0,r0,r0		# borrow?-1:0
+	vpmsumd		$Xm,$IN,$H		# H.hi<68>Xi.lo+H.lo<6C>Xi.hi
+	 and		r0,r0,$len
+	vpmsumd		$Xh,$IN,$Hh		# H.hi<68>Xi.hi
+	 add		$inp,$inp,r0
+
+	vpmsumd		$t2,$Xl,$xC2		# 1st phase
+
+	vsldoi		$t0,$Xm,$zero,8
+	vsldoi		$t1,$zero,$Xm,8
+	vxor		$Xl,$Xl,$t0
+	vxor		$Xh,$Xh,$t1
+
+	vsldoi		$Xl,$Xl,$Xl,8
+	vxor		$Xl,$Xl,$t2
+	 lvx_u		$IN,0,$inp
+	 addi		$inp,$inp,16
+
+	vsldoi		$t1,$Xl,$Xl,8		# 2nd phase
+	vpmsumd		$Xl,$Xl,$xC2
+	 le?vperm	$IN,$IN,$IN,$lemask
+	vxor		$t1,$t1,$Xh
+	vxor		$IN,$IN,$t1
+	vxor		$IN,$IN,$Xl
+	beq		Loop			# did $len-=16 borrow?
+
+	vxor		$Xl,$Xl,$t1
+	le?vperm	$Xl,$Xl,$Xl,$lemask
+	stvx_u		$Xl,0,$Xip		# write out Xi
+
+	mtspr		256,$vrsave
+	blr
+	.long		0
+	.byte		0,12,0x14,0,0,0,4,0
+	.long		0
+.size	.gcm_ghash_p8,.-.gcm_ghash_p8
+
+.asciz  "GHASH for PowerISA 2.07, CRYPTOGAMS by <appro\@openssl.org>"
+.align  2
+___
+
+foreach (split("\n",$code)) {
+	if ($flavour =~ /le$/o) {	# little-endian
+	    s/le\?//o		or
+	    s/be\?/#be#/o;
+	} else {
+	    s/le\?/#le#/o	or
+	    s/be\?//o;
+	}
+	print $_,"\n";
+}
+
+close STDOUT; # enforce flush
diff --git a/crypto/modes/gcm128.c b/crypto/modes/gcm128.c
index 0e6ff8b..6f8e7ee 100644
--- a/crypto/modes/gcm128.c
+++ b/crypto/modes/gcm128.c
@@ -671,6 +671,21 @@ void gcm_ghash_4bit_x86(u64 Xi[2],const u128 Htable[16],const u8 *inp,size_t len
 void gcm_gmult_neon(u64 Xi[2],const u128 Htable[16]);
 void gcm_ghash_neon(u64 Xi[2],const u128 Htable[16],const u8 *inp,size_t len);
 #  endif
+# elif defined(__sparc__) || defined(__sparc)
+#  include "sparc_arch.h"
+#  define GHASH_ASM_SPARC
+#  define GCM_FUNCREF_4BIT
+extern unsigned int OPENSSL_sparcv9cap_P[];
+void gcm_init_vis3(u128 Htable[16],const u64 Xi[2]);
+void gcm_gmult_vis3(u64 Xi[2],const u128 Htable[16]);
+void gcm_ghash_vis3(u64 Xi[2],const u128 Htable[16],const u8 *inp,size_t len);
+#elif defined(OPENSSL_CPUID_OBJ) && (defined(__powerpc__) || defined(__ppc__) || defined(_ARCH_PPC))
+#  include "ppc_arch.h"
+#  define GHASH_ASM_PPC
+#  define GCM_FUNCREF_4BIT
+void gcm_init_p8(u128 Htable[16],const u64 Xi[2]);
+void gcm_gmult_p8(u64 Xi[2],const u128 Htable[16]);
+void gcm_ghash_p8(u64 Xi[2],const u128 Htable[16],const u8 *inp,size_t len);
 # endif
 #endif
 
@@ -747,6 +762,16 @@ void CRYPTO_gcm128_init(GCM128_CONTEXT *ctx,void *key,block128_f block)
 		ctx->gmult = gcm_gmult_4bit;
 		ctx->ghash = gcm_ghash_4bit;
 	}
+# elif	defined(GHASH_ASM_PPC)
+	if (OPENSSL_ppccap_P & PPC_CRYPTO207) {
+		gcm_init_p8(ctx->Htable,ctx->H.u);
+		ctx->gmult = gcm_gmult_p8;
+		ctx->ghash = gcm_ghash_p8;
+	} else {
+		gcm_init_4bit(ctx->Htable,ctx->H.u);
+		ctx->gmult = gcm_gmult_4bit;
+		ctx->ghash = gcm_ghash_4bit;
+	}
 # else
 	gcm_init_4bit(ctx->Htable,ctx->H.u);
 # endif
diff --git a/crypto/perlasm/ppc-xlate.pl b/crypto/perlasm/ppc-xlate.pl
index a3edd98..f89e814 100755
--- a/crypto/perlasm/ppc-xlate.pl
+++ b/crypto/perlasm/ppc-xlate.pl
@@ -27,7 +27,8 @@ my $globl = sub {
 	/osx/		&& do { $name = "_$name";
 				last;
 			      };
-	/linux.*32/	&& do {	$ret .= ".globl	$name\n";
+	/linux.*(32|64le)/
+			&& do {	$ret .= ".globl	$name\n";
 				$ret .= ".type	$name,\@function";
 				last;
 			      };
@@ -37,7 +38,6 @@ my $globl = sub {
 				$ret .= ".align	3\n";
 				$ret .= "$name:\n";
 				$ret .= ".quad	.$name,.TOC.\@tocbase,0\n";
-				$ret .= ".size	$name,24\n";
 				$ret .= ".previous\n";
 
 				$name = ".$name";
@@ -50,7 +50,9 @@ my $globl = sub {
     $ret;
 };
 my $text = sub {
-    ($flavour =~ /aix/) ? ".csect" : ".text";
+    my $ret = ($flavour =~ /aix/) ? ".csect\t.text[PR],7" : ".text";
+    $ret = ".abiversion	2\n".$ret	if ($flavour =~ /linux.*64le/);
+    $ret;
 };
 my $machine = sub {
     my $junk = shift;
@@ -62,9 +64,12 @@ my $machine = sub {
     ".machine	$arch";
 };
 my $size = sub {
-    if ($flavour =~ /linux.*32/)
+    if ($flavour =~ /linux/)
     {	shift;
-	".size	" . join(",",@_);
+	my $name = shift; $name =~ s|^[\.\_]||;
+	my $ret  = ".size	$name,.-".($flavour=~/64$/?".":"").$name;
+	$ret .= "\n.size	.$name,.-.$name" if ($flavour=~/64$/);
+	$ret;
     }
     else
     {	"";	}
@@ -77,6 +82,25 @@ my $asciz = sub {
     else
     {	"";	}
 };
+my $quad = sub {
+    shift;
+    my @ret;
+    my ($hi,$lo);
+    for (@_) {
+	if (/^0x([0-9a-f]*?)([0-9a-f]{1,8})$/io)
+	{  $hi=$1?"0x$1":"0"; $lo="0x$2";  }
+	elsif (/^([0-9]+)$/o)
+	{  $hi=$1>>32; $lo=$1&0xffffffff;  } # error-prone with 32-bit perl
+	else
+	{  $hi=undef; $lo=$_; }
+
+	if (defined($hi))
+	{  push(@ret,$flavour=~/le$/o?".long\t$lo,$hi":".long\t$hi,$lo");  }
+	else
+	{  push(@ret,".quad	$lo");  }
+    }
+    join("\n",@ret);
+};
 
 ################################################################
 # simplified mnemonics not handled by at least one assembler
@@ -122,6 +146,46 @@ my $extrdi = sub {
     $b = ($b+$n)&63; $n = 64-$n;
     "	rldicl	$ra,$rs,$b,$n";
 };
+my $vmr = sub {
+    my ($f,$vx,$vy) = @_;
+    "	vor	$vx,$vy,$vy";
+};
+
+# PowerISA 2.06 stuff
+sub vsxmem_op {
+    my ($f, $vrt, $ra, $rb, $op) = @_;
+    "	.long	".sprintf "0x%X",(31<<26)|($vrt<<21)|($ra<<16)|($rb<<11)|($op*2+1);
+}
+# made-up unaligned memory reference AltiVec/VMX instructions
+my $lvx_u	= sub {	vsxmem_op(@_, 844); };	# lxvd2x
+my $stvx_u	= sub {	vsxmem_op(@_, 972); };	# stxvd2x
+my $lvdx_u	= sub {	vsxmem_op(@_, 588); };	# lxsdx
+my $stvdx_u	= sub {	vsxmem_op(@_, 716); };	# stxsdx
+my $lvx_4w	= sub { vsxmem_op(@_, 780); };	# lxvw4x
+my $stvx_4w	= sub { vsxmem_op(@_, 908); };	# stxvw4x
+
+# PowerISA 2.07 stuff
+sub vcrypto_op {
+    my ($f, $vrt, $vra, $vrb, $op) = @_;
+    "	.long	".sprintf "0x%X",(4<<26)|($vrt<<21)|($vra<<16)|($vrb<<11)|$op;
+}
+my $vcipher	= sub { vcrypto_op(@_, 1288); };
+my $vcipherlast	= sub { vcrypto_op(@_, 1289); };
+my $vncipher	= sub { vcrypto_op(@_, 1352); };
+my $vncipherlast= sub { vcrypto_op(@_, 1353); };
+my $vsbox	= sub { vcrypto_op(@_, 0, 1480); };
+my $vshasigmad	= sub { my ($st,$six)=splice(@_,-2); vcrypto_op(@_, $st<<4|$six, 1730); };
+my $vshasigmaw	= sub { my ($st,$six)=splice(@_,-2); vcrypto_op(@_, $st<<4|$six, 1666); };
+my $vpmsumb	= sub { vcrypto_op(@_, 1032); };
+my $vpmsumd	= sub { vcrypto_op(@_, 1224); };
+my $vpmsubh	= sub { vcrypto_op(@_, 1096); };
+my $vpmsumw	= sub { vcrypto_op(@_, 1160); };
+my $vaddudm	= sub { vcrypto_op(@_, 192);  };
+
+my $mtsle	= sub {
+    my ($f, $arg) = @_;
+    "	.long	".sprintf "0x%X",(31<<26)|($arg<<21)|(147*2);
+};
 
 while($line=<>) {
 
@@ -138,7 +202,10 @@ while($line=<>) {
     {
 	$line =~ s|(^[\.\w]+)\:\s*||;
 	my $label = $1;
-	printf "%s:",($GLOBALS{$label} or $label) if ($label);
+	if ($label) {
+	    printf "%s:",($GLOBALS{$label} or $label);
+	    printf "\n.localentry\t$GLOBALS{$label},0"	if ($GLOBALS{$label} && $flavour =~ /linux.*64le/);
+	}
     }
 
     {
@@ -147,7 +214,7 @@ while($line=<>) {
 	my $mnemonic = $2;
 	my $f = $3;
 	my $opcode = eval("\$$mnemonic");
-	$line =~ s|\bc?[rf]([0-9]+)\b|$1|g if ($c ne "." and $flavour !~ /osx/);
+	$line =~ s/\b(c?[rf]|v|vs)([0-9]+)\b/$2/g if ($c ne "." and $flavour !~ /osx/);
 	if (ref($opcode) eq 'CODE') { $line = &$opcode($f,split(',',$line)); }
 	elsif ($mnemonic)           { $line = $c.$mnemonic.$f."\t".$line; }
     }
diff --git a/crypto/ppc_arch.h b/crypto/ppc_arch.h
new file mode 100644
index 0000000..1192edf
--- /dev/null
+++ b/crypto/ppc_arch.h
@@ -0,0 +1,10 @@
+#ifndef __PPC_ARCH_H__
+#define __PPC_ARCH_H__
+
+extern unsigned int OPENSSL_ppccap_P;
+
+#define PPC_FPU64	(1<<0)
+#define PPC_ALTIVEC	(1<<1)
+#define PPC_CRYPTO207	(1<<2)
+
+#endif
diff --git a/crypto/ppccap.c b/crypto/ppccap.c
index f71ba66..13c2ca5 100644
--- a/crypto/ppccap.c
+++ b/crypto/ppccap.c
@@ -4,13 +4,15 @@
 #include <setjmp.h>
 #include <signal.h>
 #include <unistd.h>
+#if defined(__linux) || defined(_AIX)
+#include <sys/utsname.h>
+#endif
 #include <crypto.h>
 #include <openssl/bn.h>
 
-#define PPC_FPU64	(1<<0)
-#define PPC_ALTIVEC	(1<<1)
+#include "ppc_arch.h"
 
-static int OPENSSL_ppccap_P = 0;
+unsigned int OPENSSL_ppccap_P = 0;
 
 static sigset_t all_masked;
 
@@ -22,7 +24,7 @@ int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp, const BN_U
 
 	if (sizeof(size_t)==4)
 		{
-#if (defined(__APPLE__) && defined(__MACH__))
+#if 1 || (defined(__APPLE__) && defined(__MACH__))
 		if (num>=8 && (num&3)==0 && (OPENSSL_ppccap_P&PPC_FPU64))
 			return bn_mul_mont_fpu64(rp,ap,bp,np,n0,num);
 #else
@@ -50,11 +52,28 @@ int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp, const BN_U
 	}
 #endif
 
+void sha256_block_p8(void *ctx,const void *inp,size_t len);
+void sha256_block_ppc(void *ctx,const void *inp,size_t len);
+void sha256_block_data_order(void *ctx,const void *inp,size_t len)
+	{
+	OPENSSL_ppccap_P&PPC_CRYPTO207?	sha256_block_p8(ctx,inp,len):
+					sha256_block_ppc(ctx,inp,len);
+	}
+
+void sha512_block_p8(void *ctx,const void *inp,size_t len);
+void sha512_block_ppc(void *ctx,const void *inp,size_t len);
+void sha512_block_data_order(void *ctx,const void *inp,size_t len)
+	{
+	OPENSSL_ppccap_P&PPC_CRYPTO207?	sha512_block_p8(ctx,inp,len):
+					sha512_block_ppc(ctx,inp,len);
+	}
+
 static sigjmp_buf ill_jmp;
 static void ill_handler (int sig) { siglongjmp(ill_jmp,sig); }
 
 void OPENSSL_ppc64_probe(void);
 void OPENSSL_altivec_probe(void);
+void OPENSSL_crypto207_probe(void);
 
 void OPENSSL_cpuid_setup(void)
 	{
@@ -85,12 +104,14 @@ void OPENSSL_cpuid_setup(void)
 	OPENSSL_ppccap_P = 0;
 
 #if defined(_AIX)
-	if (sizeof(size_t)==4
+	if (sizeof(size_t)==4)
+		{
+		struct utsname uts;
 # if defined(_SC_AIX_KERNEL_BITMODE)
-	    && sysconf(_SC_AIX_KERNEL_BITMODE)!=64
+		if (sysconf(_SC_AIX_KERNEL_BITMODE)!=64)	return;
 # endif
-	   )
-		return;
+		if (uname(&uts)!=0 || atoi(uts.version)<6)	return;
+		}
 #endif
 
 	memset(&ill_act,0,sizeof(ill_act));
@@ -102,6 +123,10 @@ void OPENSSL_cpuid_setup(void)
 
 	if (sizeof(size_t)==4)
 		{
+#ifdef __linux
+		struct utsname uts;
+		if (uname(&uts)==0 && strcmp(uts.machine,"ppc64")==0)
+#endif
 		if (sigsetjmp(ill_jmp,1) == 0)
 			{
 			OPENSSL_ppc64_probe();
@@ -119,6 +144,11 @@ void OPENSSL_cpuid_setup(void)
 		{
 		OPENSSL_altivec_probe();
 		OPENSSL_ppccap_P |= PPC_ALTIVEC;
+		if (sigsetjmp(ill_jmp,1) == 0)
+			{
+			OPENSSL_crypto207_probe();
+			OPENSSL_ppccap_P |= PPC_CRYPTO207;
+			}
 		}
 
 	sigaction (SIGILL,&ill_oact,NULL);
diff --git a/crypto/ppccpuid.pl b/crypto/ppccpuid.pl
index 4ba736a..56cc851 100755
--- a/crypto/ppccpuid.pl
+++ b/crypto/ppccpuid.pl
@@ -31,6 +31,7 @@ $code=<<___;
 	blr
 	.long	0
 	.byte	0,12,0x14,0,0,0,0,0
+.size	.OPENSSL_ppc64_probe,.-.OPENSSL_ppc64_probe
 
 .globl	.OPENSSL_altivec_probe
 .align	4
@@ -39,6 +40,17 @@ $code=<<___;
 	blr
 	.long	0
 	.byte	0,12,0x14,0,0,0,0,0
+.size	.OPENSSL_altivec_probe,.-..OPENSSL_altivec_probe
+
+.globl	.OPENSSL_crypto207_probe
+.align	4
+.OPENSSL_crypto207_probe:
+	lvx_u	v0,0,r1
+	vcipher	v0,v0,v0
+	blr
+	.long	0
+	.byte	0,12,0x14,0,0,0,0,0
+.size	.OPENSSL_crypto207_probe,.-.OPENSSL_crypto207_probe
 
 .globl	.OPENSSL_wipe_cpu
 .align	4
@@ -71,6 +83,7 @@ $code=<<___;
 	blr
 	.long	0
 	.byte	0,12,0x14,0,0,0,0,0
+.size	.OPENSSL_wipe_cpu,.-.OPENSSL_wipe_cpu
 
 .globl	.OPENSSL_atomic_add
 .align	4
@@ -84,6 +97,7 @@ Ladd:	lwarx	r5,0,r3
 	.long	0
 	.byte	0,12,0x14,0,0,0,2,0
 	.long	0
+.size	.OPENSSL_atomic_add,.-.OPENSSL_atomic_add
 
 .globl	.OPENSSL_rdtsc
 .align	4
@@ -93,6 +107,7 @@ Ladd:	lwarx	r5,0,r3
 	blr
 	.long	0
 	.byte	0,12,0x14,0,0,0,0,0
+.size	.OPENSSL_rdtsc,.-.OPENSSL_rdtsc
 
 .globl	.OPENSSL_cleanse
 .align	4
@@ -125,7 +140,99 @@ Laligned:
 	.long	0
 	.byte	0,12,0x14,0,0,0,2,0
 	.long	0
+.size	.OPENSSL_cleanse,.-.OPENSSL_cleanse
+___
+{
+my ($out,$cnt,$max)=("r3","r4","r5");
+my ($tick,$lasttick)=("r6","r7");
+my ($diff,$lastdiff)=("r8","r9");
+
+$code.=<<___;
+.globl	.OPENSSL_instrument_bus
+.align	4
+.OPENSSL_instrument_bus:
+	mtctr	$cnt
+
+	mftb	$lasttick		# collect 1st tick
+	li	$diff,0
+
+	dcbf	0,$out			# flush cache line
+	lwarx	$tick,0,$out		# load and lock
+	add	$tick,$tick,$diff
+	stwcx.	$tick,0,$out
+	stwx	$tick,0,$out
+
+Loop:	mftb	$tick
+	sub	$diff,$tick,$lasttick
+	mr	$lasttick,$tick
+	dcbf	0,$out			# flush cache line
+	lwarx	$tick,0,$out		# load and lock
+	add	$tick,$tick,$diff
+	stwcx.	$tick,0,$out
+	stwx	$tick,0,$out
+	addi	$out,$out,4		# ++$out
+	bdnz	Loop
+
+	mr	r3,$cnt
+	blr
+	.long	0
+	.byte	0,12,0x14,0,0,0,2,0
+	.long	0
+.size	.OPENSSL_instrument_bus,.-.OPENSSL_instrument_bus
+
+.globl	.OPENSSL_instrument_bus2
+.align	4
+.OPENSSL_instrument_bus2:
+	mr	r0,$cnt
+	slwi	$cnt,$cnt,2
+
+	mftb	$lasttick		# collect 1st tick
+	li	$diff,0
+
+	dcbf	0,$out			# flush cache line
+	lwarx	$tick,0,$out		# load and lock
+	add	$tick,$tick,$diff
+	stwcx.	$tick,0,$out
+	stwx	$tick,0,$out
+
+	mftb	$tick			# collect 1st diff
+	sub	$diff,$tick,$lasttick
+	mr	$lasttick,$tick
+	mr	$lastdiff,$diff
+Loop2:
+	dcbf	0,$out			# flush cache line
+	lwarx	$tick,0,$out		# load and lock
+	add	$tick,$tick,$diff
+	stwcx.	$tick,0,$out
+	stwx	$tick,0,$out
+
+	addic.	$max,$max,-1
+	beq	Ldone2
+
+	mftb	$tick
+	sub	$diff,$tick,$lasttick
+	mr	$lasttick,$tick
+	cmplw	7,$diff,$lastdiff
+	mr	$lastdiff,$diff
+
+	mfcr	$tick			# pull cr
+	not	$tick,$tick		# flip bits
+	rlwinm	$tick,$tick,1,29,29	# isolate flipped eq bit and scale
+
+	sub.	$cnt,$cnt,$tick		# conditional --$cnt
+	add	$out,$out,$tick		# conditional ++$out
+	bne	Loop2
+
+Ldone2:
+	srwi	$cnt,$cnt,2
+	sub	r3,r0,$cnt
+	blr
+	.long	0
+	.byte	0,12,0x14,0,0,0,3,0
+	.long	0
+.size	.OPENSSL_instrument_bus2,.-.OPENSSL_instrument_bus2
 ___
+}
 
 $code =~ s/\`([^\`]*)\`/eval $1/gem;
 print $code;
diff --git a/crypto/sha/Makefile b/crypto/sha/Makefile
index 6d191d3..58c6705 100644
--- a/crypto/sha/Makefile
+++ b/crypto/sha/Makefile
@@ -73,6 +73,8 @@ sha512-sparcv9.s:asm/sha512-sparcv9.pl;	$(PERL) asm/sha512-sparcv9.pl $@ $(CFLAG
 sha1-ppc.s:	asm/sha1-ppc.pl;	$(PERL) asm/sha1-ppc.pl $(PERLASM_SCHEME) $@
 sha256-ppc.s:	asm/sha512-ppc.pl;	$(PERL) asm/sha512-ppc.pl $(PERLASM_SCHEME) $@
 sha512-ppc.s:	asm/sha512-ppc.pl;	$(PERL) asm/sha512-ppc.pl $(PERLASM_SCHEME) $@
+sha256p8-ppc.s:	asm/sha512p8-ppc.pl;	$(PERL) asm/sha512p8-ppc.pl $(PERLASM_SCHEME) $@
+sha512p8-ppc.s:	asm/sha512p8-ppc.pl;	$(PERL) asm/sha512p8-ppc.pl $(PERLASM_SCHEME) $@
 
 sha1-parisc.s:	asm/sha1-parisc.pl;	$(PERL) asm/sha1-parisc.pl $(PERLASM_SCHEME) $@
 sha256-parisc.s:asm/sha512-parisc.pl;	$(PERL) asm/sha512-parisc.pl $(PERLASM_SCHEME) $@
diff --git a/crypto/sha/asm/sha1-ppc.pl b/crypto/sha/asm/sha1-ppc.pl
index 2140dd2..df59896 100755
--- a/crypto/sha/asm/sha1-ppc.pl
+++ b/crypto/sha/asm/sha1-ppc.pl
@@ -9,8 +9,7 @@
 
 # I let hardware handle unaligned input(*), except on page boundaries
 # (see below for details). Otherwise straightforward implementation
-# with X vector in register bank. The module is big-endian [which is
-# not big deal as there're no little-endian targets left around].
+# with X vector in register bank.
 #
 # (*) this means that this module is inappropriate for PPC403? Does
 #     anybody know if pre-POWER3 can sustain unaligned load?
@@ -38,6 +37,10 @@ if ($flavour =~ /64/) {
 	$PUSH	="stw";
 } else { die "nonsense $flavour"; }
 
+# Define endianess based on flavour
+# i.e.: linux64le
+$LITTLE_ENDIAN = ($flavour=~/le$/) ? $SIZE_T : 0;
+
 $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
 ( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or
 ( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or
@@ -68,14 +71,28 @@ $T  ="r12";
 @X=("r16","r17","r18","r19","r20","r21","r22","r23",
     "r24","r25","r26","r27","r28","r29","r30","r31");
 
+sub loadbe {
+my ($dst, $src, $temp_reg) = @_;
+$code.=<<___ if (!$LITTLE_ENDIAN);
+	lwz	$dst,$src
+___
+$code.=<<___ if ($LITTLE_ENDIAN);
+	lwz	$temp_reg,$src
+	rotlwi	$dst,$temp_reg,8
+	rlwimi	$dst,$temp_reg,24,0,7
+	rlwimi	$dst,$temp_reg,24,16,23
+___
+}
+
 sub BODY_00_19 {
 my ($i,$a,$b,$c,$d,$e,$f)=@_;
 my $j=$i+1;
-$code.=<<___ if ($i==0);
-	lwz	@X[$i],`$i*4`($inp)
-___
+
+	# Since the last value of $f is discarded, we can use
+	# it as a temp reg to swap byte-order when needed.
+	loadbe("@X[$i]","`$i*4`($inp)",$f) if ($i==0);
+	loadbe("@X[$j]","`$j*4`($inp)",$f) if ($i<15);
 $code.=<<___ if ($i<15);
-	lwz	@X[$j],`$j*4`($inp)
 	add	$f,$K,$e
 	rotlwi	$e,$a,5
 	add	$f,$f,@X[$i]
@@ -108,31 +125,31 @@ my ($i,$a,$b,$c,$d,$e,$f)=@_;
 my $j=$i+1;
 $code.=<<___ if ($i<79);
 	add	$f,$K,$e
+	xor	$t0,$b,$d
 	rotlwi	$e,$a,5
 	xor	@X[$j%16],@X[$j%16],@X[($j+2)%16]
 	add	$f,$f,@X[$i%16]
-	xor	$t0,$b,$c
+	xor	$t0,$t0,$c
 	xor	@X[$j%16],@X[$j%16],@X[($j+8)%16]
-	add	$f,$f,$e
+	add	$f,$f,$t0
 	rotlwi	$b,$b,30
-	xor	$t0,$t0,$d
 	xor	@X[$j%16],@X[$j%16],@X[($j+13)%16]
-	add	$f,$f,$t0
+	add	$f,$f,$e
 	rotlwi	@X[$j%16],@X[$j%16],1
 ___
 $code.=<<___ if ($i==79);
 	add	$f,$K,$e
+	xor	$t0,$b,$d
 	rotlwi	$e,$a,5
 	lwz	r16,0($ctx)
 	add	$f,$f,@X[$i%16]
-	xor	$t0,$b,$c
+	xor	$t0,$t0,$c
 	lwz	r17,4($ctx)
-	add	$f,$f,$e
+	add	$f,$f,$t0
 	rotlwi	$b,$b,30
 	lwz	r18,8($ctx)
-	xor	$t0,$t0,$d
 	lwz	r19,12($ctx)
-	add	$f,$f,$t0
+	add	$f,$f,$e
 	lwz	r20,16($ctx)
 ___
 }
@@ -316,6 +333,7 @@ $code.=<<___;
 	blr
 	.long	0
 	.byte	0,12,0x14,0,0,0,0,0
+.size	.sha1_block_data_order,.-.sha1_block_data_order
 ___
 $code.=<<___;
 .asciz	"SHA1 block transform for PPC, CRYPTOGAMS by <appro\@fy.chalmers.se>"
diff --git a/crypto/sha/asm/sha512-ppc.pl b/crypto/sha/asm/sha512-ppc.pl
index 6b44a68..734f3c1 100755
--- a/crypto/sha/asm/sha512-ppc.pl
+++ b/crypto/sha/asm/sha512-ppc.pl
@@ -1,7 +1,7 @@
 #!/usr/bin/env perl
 
 # ====================================================================
-# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
 # project. The module is, however, dual licensed under OpenSSL and
 # CRYPTOGAMS licenses depending on where you obtain it. For further
 # details see http://www.openssl.org/~appro/cryptogams/.
@@ -9,8 +9,7 @@
 
 # I let hardware handle unaligned input, except on page boundaries
 # (see below for details). Otherwise straightforward implementation
-# with X vector in register bank. The module is big-endian [which is
-# not big deal as there're no little-endian targets left around].
+# with X vector in register bank.
 
 #			sha256		|	sha512
 # 			-m64	-m32	|	-m64	-m32
@@ -56,6 +55,8 @@ if ($flavour =~ /64/) {
 	$PUSH="stw";
 } else { die "nonsense $flavour"; }
 
+$LITTLE_ENDIAN = ($flavour=~/le$/) ? $SIZE_T : 0;
+
 $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
 ( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or
 ( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or
@@ -64,7 +65,7 @@ die "can't locate ppc-xlate.pl";
 open STDOUT,"| $^X $xlate $flavour $output" || die "can't call $xlate: $!";
 
 if ($output =~ /512/) {
-	$func="sha512_block_data_order";
+	$func="sha512_block_ppc";
 	$SZ=8;
 	@Sigma0=(28,34,39);
 	@Sigma1=(14,18,41);
@@ -76,7 +77,7 @@ if ($output =~ /512/) {
 	$ROR="rotrdi";
 	$SHR="srdi";
 } else {
-	$func="sha256_block_data_order";
+	$func="sha256_block_ppc";
 	$SZ=4;
 	@Sigma0=( 2,13,22);
 	@Sigma1=( 6,11,25);
@@ -110,7 +111,7 @@ $B  ="r9";
 $C  ="r10";
 $D  ="r11";
 $E  ="r12";
-$F  ="r13";	$F="r2" if ($SIZE_T==8);# reassigned to exempt TLS pointer
+$F  =$t1;	$t1 = "r0";	# stay away from "r13";
 $G  ="r14";
 $H  ="r15";
 
@@ -118,24 +119,23 @@ $H  ="r15";
 @X=("r16","r17","r18","r19","r20","r21","r22","r23",
     "r24","r25","r26","r27","r28","r29","r30","r31");
 
-$inp="r31";	# reassigned $inp! aliases with @X[15]
+$inp="r31" if($SZ==4 || $SIZE_T==8);	# reassigned $inp! aliases with @X[15]
 
 sub ROUND_00_15 {
 my ($i,$a,$b,$c,$d,$e,$f,$g,$h)=@_;
 $code.=<<___;
-	$LD	$T,`$i*$SZ`($Tbl)
 	$ROR	$a0,$e,$Sigma1[0]
 	$ROR	$a1,$e,$Sigma1[1]
 	and	$t0,$f,$e
-	andc	$t1,$g,$e
-	add	$T,$T,$h
 	xor	$a0,$a0,$a1
+	add	$h,$h,$t1
+	andc	$t1,$g,$e
 	$ROR	$a1,$a1,`$Sigma1[2]-$Sigma1[1]`
 	or	$t0,$t0,$t1		; Ch(e,f,g)
-	add	$T,$T,@X[$i]
+	add	$h,$h,@X[$i%16]
 	xor	$a0,$a0,$a1		; Sigma1(e)
-	add	$T,$T,$t0
-	add	$T,$T,$a0
+	add	$h,$h,$t0
+	add	$h,$h,$a0
 
 	$ROR	$a0,$a,$Sigma0[0]
 	$ROR	$a1,$a,$Sigma0[1]
@@ -146,9 +146,14 @@ $code.=<<___;
 	xor	$t0,$t0,$t1
 	and	$t1,$b,$c
 	xor	$a0,$a0,$a1		; Sigma0(a)
-	add	$d,$d,$T
+	add	$d,$d,$h
 	xor	$t0,$t0,$t1		; Maj(a,b,c)
-	add	$h,$T,$a0
+___
+$code.=<<___ if ($i<15);
+	$LD	$t1,`($i+1)*$SZ`($Tbl)
+___
+$code.=<<___;
+	add	$h,$h,$a0
 	add	$h,$h,$t0
 
 ___
@@ -169,10 +174,11 @@ $code.=<<___;
 	add	@X[$i],@X[$i],@X[($i+9)%16]
 	xor	$a0,$a0,$a1		; sigma0(X[(i+1)&0x0f])
 	xor	$t0,$t0,$t1		; sigma1(X[(i+14)&0x0f])
+	$LD	$t1,`$i*$SZ`($Tbl)
 	add	@X[$i],@X[$i],$a0
 	add	@X[$i],@X[$i],$t0
 ___
-&ROUND_00_15($i,$a,$b,$c,$d,$e,$f,$g,$h);
+&ROUND_00_15($i+16,$a,$b,$c,$d,$e,$f,$g,$h);
 }
 
 $code=<<___;
@@ -188,8 +194,6 @@ $func:
 
 	$PUSH	$ctx,`$FRAME-$SIZE_T*22`($sp)
 
-	$PUSH	$toc,`$FRAME-$SIZE_T*20`($sp)
-	$PUSH	r13,`$FRAME-$SIZE_T*19`($sp)
 	$PUSH	r14,`$FRAME-$SIZE_T*18`($sp)
 	$PUSH	r15,`$FRAME-$SIZE_T*17`($sp)
 	$PUSH	r16,`$FRAME-$SIZE_T*16`($sp)
@@ -209,7 +213,10 @@ $func:
 	$PUSH	r30,`$FRAME-$SIZE_T*2`($sp)
 	$PUSH	r31,`$FRAME-$SIZE_T*1`($sp)
 	$PUSH	r0,`$FRAME+$LRSAVE`($sp)
+___
 
+if ($SZ==4 || $SIZE_T==8) {
+$code.=<<___;
 	$LD	$A,`0*$SZ`($ctx)
 	mr	$inp,r4				; incarnate $inp
 	$LD	$B,`1*$SZ`($ctx)
@@ -219,7 +226,16 @@ $func:
 	$LD	$F,`5*$SZ`($ctx)
 	$LD	$G,`6*$SZ`($ctx)
 	$LD	$H,`7*$SZ`($ctx)
+___
+} else {
+  for ($i=16;$i<32;$i++) {
+    $code.=<<___;
+	lwz	r$i,`$LITTLE_ENDIAN^(4*($i-16))`($ctx)
+___
+  }
+}
 
+$code.=<<___;
 	bl	LPICmeup
 LPICedup:
 	andi.	r0,$inp,3
@@ -255,6 +271,9 @@ Lunaligned:
 Lcross_page:
 	li	$t1,`16*$SZ/4`
 	mtctr	$t1
+___
+if ($SZ==4 || $SIZE_T==8) {
+$code.=<<___;
 	addi	r20,$sp,$LOCALS			; aligned spot below the frame
 Lmemcpy:
 	lbz	r16,0($inp)
@@ -268,7 +287,26 @@ Lmemcpy:
 	stb	r19,3(r20)
 	addi	r20,r20,4
 	bdnz	Lmemcpy
+___
+} else {
+$code.=<<___;
+	addi	r12,$sp,$LOCALS			; aligned spot below the frame
+Lmemcpy:
+	lbz	r8,0($inp)
+	lbz	r9,1($inp)
+	lbz	r10,2($inp)
+	lbz	r11,3($inp)
+	addi	$inp,$inp,4
+	stb	r8,0(r12)
+	stb	r9,1(r12)
+	stb	r10,2(r12)
+	stb	r11,3(r12)
+	addi	r12,r12,4
+	bdnz	Lmemcpy
+___
+}
 
+$code.=<<___;
 	$PUSH	$inp,`$FRAME-$SIZE_T*26`($sp)	; save real inp
 	addi	$t1,$sp,`$LOCALS+16*$SZ`	; fictitious end pointer
 	addi	$inp,$sp,$LOCALS		; fictitious inp pointer
@@ -283,8 +321,6 @@ Lmemcpy:
 
 Ldone:
 	$POP	r0,`$FRAME+$LRSAVE`($sp)
-	$POP	$toc,`$FRAME-$SIZE_T*20`($sp)
-	$POP	r13,`$FRAME-$SIZE_T*19`($sp)
 	$POP	r14,`$FRAME-$SIZE_T*18`($sp)
 	$POP	r15,`$FRAME-$SIZE_T*17`($sp)
 	$POP	r16,`$FRAME-$SIZE_T*16`($sp)
@@ -309,27 +345,48 @@ Ldone:
 	.long	0
 	.byte	0,12,4,1,0x80,18,3,0
 	.long	0
+___
 
+if ($SZ==4 || $SIZE_T==8) {
+$code.=<<___;
 .align	4
 Lsha2_block_private:
+	$LD	$t1,0($Tbl)
 ___
 for($i=0;$i<16;$i++) {
-$code.=<<___ if ($SZ==4);
+$code.=<<___ if ($SZ==4 && !$LITTLE_ENDIAN);
 	lwz	@X[$i],`$i*$SZ`($inp)
 ___
+$code.=<<___ if ($SZ==4 && $LITTLE_ENDIAN);
+	lwz	$a0,`$i*$SZ`($inp)
+	rotlwi	@X[$i],$a0,8
+	rlwimi	@X[$i],$a0,24,0,7
+	rlwimi	@X[$i],$a0,24,16,23
+___
 # 64-bit loads are split to 2x32-bit ones, as CPU can't handle
 # unaligned 64-bit loads, only 32-bit ones...
-$code.=<<___ if ($SZ==8);
+$code.=<<___ if ($SZ==8 && !$LITTLE_ENDIAN);
 	lwz	$t0,`$i*$SZ`($inp)
 	lwz	@X[$i],`$i*$SZ+4`($inp)
 	insrdi	@X[$i],$t0,32,0
 ___
+$code.=<<___ if ($SZ==8 && $LITTLE_ENDIAN);
+	lwz	$a0,`$i*$SZ`($inp)
+	 lwz	$a1,`$i*$SZ+4`($inp)
+	rotlwi	$t0,$a0,8
+	 rotlwi	@X[$i],$a1,8
+	rlwimi	$t0,$a0,24,0,7
+	 rlwimi	@X[$i],$a1,24,0,7
+	rlwimi	$t0,$a0,24,16,23
+	 rlwimi	@X[$i],$a1,24,16,23
+	insrdi	@X[$i],$t0,32,0
+___
 	&ROUND_00_15($i,@V);
 	unshift(@V,pop(@V));
 }
 $code.=<<___;
-	li	$T,`$rounds/16-1`
-	mtctr	$T
+	li	$t0,`$rounds/16-1`
+	mtctr	$t0
 .align	4
 Lrounds:
 	addi	$Tbl,$Tbl,`16*$SZ`
@@ -377,7 +434,282 @@ $code.=<<___;
 	blr
 	.long	0
 	.byte	0,12,0x14,0,0,0,0,0
+.size	$func,.-$func
+___
+} else {
+########################################################################
+# SHA512 for PPC32, X vector is off-loaded to stack...
+#
+#			|	sha512
+#			|	-m32
+# ----------------------+-----------------------
+# PPC74x0,gcc-4.0.1	|	+48%
+# POWER6,gcc-4.4.6	|	+124%(*)
+# POWER7,gcc-4.4.6	|	+79%(*)
+# e300,gcc-4.1.0	|	+167%
+#
+# (*)	~1/3 of -m64 result [and ~20% better than -m32 code generated
+#	by xlc-12.1]
+
+my $XOFF=$LOCALS;
+
+my @V=map("r$_",(16..31));	# A..H
+
+my ($s0,$s1,$t0,$t1,$t2,$t3,$a0,$a1,$a2,$a3)=map("r$_",(0,5,6,8..12,14,15));
+my ($x0,$x1)=("r3","r4");	# zaps $ctx and $inp
+
+sub ROUND_00_15_ppc32 {
+my ($i,	$ahi,$alo,$bhi,$blo,$chi,$clo,$dhi,$dlo,
+	$ehi,$elo,$fhi,$flo,$ghi,$glo,$hhi,$hlo)=@_;
+
+$code.=<<___;
+	lwz	$t2,`$SZ*($i%16)+($LITTLE_ENDIAN^4)`($Tbl)
+	 xor	$a0,$flo,$glo
+	lwz	$t3,`$SZ*($i%16)+($LITTLE_ENDIAN^0)`($Tbl)
+	 xor	$a1,$fhi,$ghi
+	addc	$hlo,$hlo,$t0			; h+=x[i]
+	stw	$t0,`$XOFF+0+$SZ*($i%16)`($sp)	; save x[i]
+
+	srwi	$s0,$elo,$Sigma1[0]
+	srwi	$s1,$ehi,$Sigma1[0]
+	 and	$a0,$a0,$elo
+	adde	$hhi,$hhi,$t1
+	 and	$a1,$a1,$ehi
+	stw	$t1,`$XOFF+4+$SZ*($i%16)`($sp)
+	srwi	$t0,$elo,$Sigma1[1]
+	srwi	$t1,$ehi,$Sigma1[1]
+	 addc	$hlo,$hlo,$t2			; h+=K512[i]
+	insrwi	$s0,$ehi,$Sigma1[0],0
+	insrwi	$s1,$elo,$Sigma1[0],0
+	 xor	$a0,$a0,$glo			; Ch(e,f,g)
+	 adde	$hhi,$hhi,$t3
+	 xor	$a1,$a1,$ghi
+	insrwi	$t0,$ehi,$Sigma1[1],0
+	insrwi	$t1,$elo,$Sigma1[1],0
+	 addc	$hlo,$hlo,$a0			; h+=Ch(e,f,g)
+	srwi	$t2,$ehi,$Sigma1[2]-32
+	srwi	$t3,$elo,$Sigma1[2]-32
+	xor	$s0,$s0,$t0
+	xor	$s1,$s1,$t1
+	insrwi	$t2,$elo,$Sigma1[2]-32,0
+	insrwi	$t3,$ehi,$Sigma1[2]-32,0
+	 xor	$a0,$alo,$blo			; a^b, b^c in next round
+	 adde	$hhi,$hhi,$a1
+	 xor	$a1,$ahi,$bhi
+	xor	$s0,$s0,$t2			; Sigma1(e)
+	xor	$s1,$s1,$t3
+
+	srwi	$t0,$alo,$Sigma0[0]
+	 and	$a2,$a2,$a0
+	 addc	$hlo,$hlo,$s0			; h+=Sigma1(e)
+	 and	$a3,$a3,$a1
+	srwi	$t1,$ahi,$Sigma0[0]
+	srwi	$s0,$ahi,$Sigma0[1]-32
+	 adde	$hhi,$hhi,$s1
+	srwi	$s1,$alo,$Sigma0[1]-32
+	insrwi	$t0,$ahi,$Sigma0[0],0
+	insrwi	$t1,$alo,$Sigma0[0],0
+	 xor	$a2,$a2,$blo			; Maj(a,b,c)
+	 addc	$dlo,$dlo,$hlo			; d+=h
+	 xor	$a3,$a3,$bhi
+	insrwi	$s0,$alo,$Sigma0[1]-32,0
+	insrwi	$s1,$ahi,$Sigma0[1]-32,0
+	 adde	$dhi,$dhi,$hhi
+	srwi	$t2,$ahi,$Sigma0[2]-32
+	srwi	$t3,$alo,$Sigma0[2]-32
+	xor	$s0,$s0,$t0
+	 addc	$hlo,$hlo,$a2			; h+=Maj(a,b,c)
+	xor	$s1,$s1,$t1
+	insrwi	$t2,$alo,$Sigma0[2]-32,0
+	insrwi	$t3,$ahi,$Sigma0[2]-32,0
+	 adde	$hhi,$hhi,$a3
+___
+$code.=<<___ if ($i>=15);
+	lwz	$t0,`$XOFF+0+$SZ*(($i+2)%16)`($sp)
+	lwz	$t1,`$XOFF+4+$SZ*(($i+2)%16)`($sp)
+___
+$code.=<<___ if ($i<15 && !$LITTLE_ENDIAN);
+	lwz	$t1,`$SZ*($i+1)+0`($inp)
+	lwz	$t0,`$SZ*($i+1)+4`($inp)
 ___
+$code.=<<___ if ($i<15 && $LITTLE_ENDIAN);
+	lwz	$a2,`$SZ*($i+1)+0`($inp)
+	 lwz	$a3,`$SZ*($i+1)+4`($inp)
+	rotlwi	$t1,$a2,8
+	 rotlwi	$t0,$a3,8
+	rlwimi	$t1,$a2,24,0,7
+	 rlwimi	$t0,$a3,24,0,7
+	rlwimi	$t1,$a2,24,16,23
+	 rlwimi	$t0,$a3,24,16,23
+___
+$code.=<<___;
+	xor	$s0,$s0,$t2			; Sigma0(a)
+	xor	$s1,$s1,$t3
+	addc	$hlo,$hlo,$s0			; h+=Sigma0(a)
+	adde	$hhi,$hhi,$s1
+___
+$code.=<<___ if ($i==15);
+	lwz	$x0,`$XOFF+0+$SZ*(($i+1)%16)`($sp)
+	lwz	$x1,`$XOFF+4+$SZ*(($i+1)%16)`($sp)
+___
+}
+sub ROUND_16_xx_ppc32 {
+my ($i,	$ahi,$alo,$bhi,$blo,$chi,$clo,$dhi,$dlo,
+	$ehi,$elo,$fhi,$flo,$ghi,$glo,$hhi,$hlo)=@_;
+
+$code.=<<___;
+	srwi	$s0,$t0,$sigma0[0]
+	srwi	$s1,$t1,$sigma0[0]
+	srwi	$t2,$t0,$sigma0[1]
+	srwi	$t3,$t1,$sigma0[1]
+	insrwi	$s0,$t1,$sigma0[0],0
+	insrwi	$s1,$t0,$sigma0[0],0
+	srwi	$a0,$t0,$sigma0[2]
+	insrwi	$t2,$t1,$sigma0[1],0
+	insrwi	$t3,$t0,$sigma0[1],0
+	insrwi	$a0,$t1,$sigma0[2],0
+	xor	$s0,$s0,$t2
+	 lwz	$t2,`$XOFF+0+$SZ*(($i+14)%16)`($sp)
+	srwi	$a1,$t1,$sigma0[2]
+	xor	$s1,$s1,$t3
+	 lwz	$t3,`$XOFF+4+$SZ*(($i+14)%16)`($sp)
+	xor	$a0,$a0,$s0
+	 srwi	$s0,$t2,$sigma1[0]
+	xor	$a1,$a1,$s1
+	 srwi	$s1,$t3,$sigma1[0]
+	addc	$x0,$x0,$a0			; x[i]+=sigma0(x[i+1])
+	 srwi	$a0,$t3,$sigma1[1]-32
+	insrwi	$s0,$t3,$sigma1[0],0
+	insrwi	$s1,$t2,$sigma1[0],0
+	adde	$x1,$x1,$a1
+	 srwi	$a1,$t2,$sigma1[1]-32
+
+	insrwi	$a0,$t2,$sigma1[1]-32,0
+	srwi	$t2,$t2,$sigma1[2]
+	insrwi	$a1,$t3,$sigma1[1]-32,0
+	insrwi	$t2,$t3,$sigma1[2],0
+	xor	$s0,$s0,$a0
+	 lwz	$a0,`$XOFF+0+$SZ*(($i+9)%16)`($sp)
+	srwi	$t3,$t3,$sigma1[2]
+	xor	$s1,$s1,$a1
+	 lwz	$a1,`$XOFF+4+$SZ*(($i+9)%16)`($sp)
+	xor	$s0,$s0,$t2
+	 addc	$x0,$x0,$a0			; x[i]+=x[i+9]
+	xor	$s1,$s1,$t3
+	 adde	$x1,$x1,$a1
+	addc	$x0,$x0,$s0			; x[i]+=sigma1(x[i+14])
+	adde	$x1,$x1,$s1
+___
+	($t0,$t1,$x0,$x1) = ($x0,$x1,$t0,$t1);
+	&ROUND_00_15_ppc32(@_);
+}
+
+$code.=<<___;
+.align	4
+Lsha2_block_private:
+___
+$code.=<<___ if (!$LITTLE_ENDIAN);
+	lwz	$t1,0($inp)
+	xor	$a2,@V[3],@V[5]		; B^C, magic seed
+	lwz	$t0,4($inp)
+	xor	$a3,@V[2],@V[4]
+___
+$code.=<<___ if ($LITTLE_ENDIAN);
+	lwz	$a1,0($inp)
+	xor	$a2,@V[3],@V[5]		; B^C, magic seed
+	lwz	$a0,4($inp)
+	xor	$a3,@V[2],@V[4]
+	rotlwi	$t1,$a1,8
+	 rotlwi	$t0,$a0,8
+	rlwimi	$t1,$a1,24,0,7
+	 rlwimi	$t0,$a0,24,0,7
+	rlwimi	$t1,$a1,24,16,23
+	 rlwimi	$t0,$a0,24,16,23
+___
+for($i=0;$i<16;$i++) {
+	&ROUND_00_15_ppc32($i,@V);
+	unshift(@V,pop(@V));	unshift(@V,pop(@V));
+	($a0,$a1,$a2,$a3) = ($a2,$a3,$a0,$a1);
+}
+$code.=<<___;
+	li	$a0,`$rounds/16-1`
+	mtctr	$a0
+.align	4
+Lrounds:
+	addi	$Tbl,$Tbl,`16*$SZ`
+___
+for(;$i<32;$i++) {
+	&ROUND_16_xx_ppc32($i,@V);
+	unshift(@V,pop(@V));	unshift(@V,pop(@V));
+	($a0,$a1,$a2,$a3) = ($a2,$a3,$a0,$a1);
+}
+$code.=<<___;
+	bdnz-	Lrounds
+
+	$POP	$ctx,`$FRAME-$SIZE_T*22`($sp)
+	$POP	$inp,`$FRAME-$SIZE_T*23`($sp)	; inp pointer
+	$POP	$num,`$FRAME-$SIZE_T*24`($sp)	; end pointer
+	subi	$Tbl,$Tbl,`($rounds-16)*$SZ`	; rewind Tbl
+
+	lwz	$t0,`$LITTLE_ENDIAN^0`($ctx)
+	lwz	$t1,`$LITTLE_ENDIAN^4`($ctx)
+	lwz	$t2,`$LITTLE_ENDIAN^8`($ctx)
+	lwz	$t3,`$LITTLE_ENDIAN^12`($ctx)
+	lwz	$a0,`$LITTLE_ENDIAN^16`($ctx)
+	lwz	$a1,`$LITTLE_ENDIAN^20`($ctx)
+	lwz	$a2,`$LITTLE_ENDIAN^24`($ctx)
+	addc	@V[1],@V[1],$t1
+	lwz	$a3,`$LITTLE_ENDIAN^28`($ctx)
+	adde	@V[0],@V[0],$t0
+	lwz	$t0,`$LITTLE_ENDIAN^32`($ctx)
+	addc	@V[3],@V[3],$t3
+	lwz	$t1,`$LITTLE_ENDIAN^36`($ctx)
+	adde	@V[2],@V[2],$t2
+	lwz	$t2,`$LITTLE_ENDIAN^40`($ctx)
+	addc	@V[5],@V[5],$a1
+	lwz	$t3,`$LITTLE_ENDIAN^44`($ctx)
+	adde	@V[4],@V[4],$a0
+	lwz	$a0,`$LITTLE_ENDIAN^48`($ctx)
+	addc	@V[7],@V[7],$a3
+	lwz	$a1,`$LITTLE_ENDIAN^52`($ctx)
+	adde	@V[6],@V[6],$a2
+	lwz	$a2,`$LITTLE_ENDIAN^56`($ctx)
+	addc	@V[9],@V[9],$t1
+	lwz	$a3,`$LITTLE_ENDIAN^60`($ctx)
+	adde	@V[8],@V[8],$t0
+	stw	@V[0],`$LITTLE_ENDIAN^0`($ctx)
+	stw	@V[1],`$LITTLE_ENDIAN^4`($ctx)
+	addc	@V[11],@V[11],$t3
+	stw	@V[2],`$LITTLE_ENDIAN^8`($ctx)
+	stw	@V[3],`$LITTLE_ENDIAN^12`($ctx)
+	adde	@V[10],@V[10],$t2
+	stw	@V[4],`$LITTLE_ENDIAN^16`($ctx)
+	stw	@V[5],`$LITTLE_ENDIAN^20`($ctx)
+	addc	@V[13],@V[13],$a1
+	stw	@V[6],`$LITTLE_ENDIAN^24`($ctx)
+	stw	@V[7],`$LITTLE_ENDIAN^28`($ctx)
+	adde	@V[12],@V[12],$a0
+	stw	@V[8],`$LITTLE_ENDIAN^32`($ctx)
+	stw	@V[9],`$LITTLE_ENDIAN^36`($ctx)
+	addc	@V[15],@V[15],$a3
+	stw	@V[10],`$LITTLE_ENDIAN^40`($ctx)
+	stw	@V[11],`$LITTLE_ENDIAN^44`($ctx)
+	adde	@V[14],@V[14],$a2
+	stw	@V[12],`$LITTLE_ENDIAN^48`($ctx)
+	stw	@V[13],`$LITTLE_ENDIAN^52`($ctx)
+	stw	@V[14],`$LITTLE_ENDIAN^56`($ctx)
+	stw	@V[15],`$LITTLE_ENDIAN^60`($ctx)
+
+	addi	$inp,$inp,`16*$SZ`		; advance inp
+	$PUSH	$inp,`$FRAME-$SIZE_T*23`($sp)
+	$UCMP	$inp,$num
+	bne	Lsha2_block_private
+	blr
+	.long	0
+	.byte	0,12,0x14,0,0,0,0,0
+.size	$func,.-$func
+___
+}
 
 # Ugly hack here, because PPC assembler syntax seem to vary too
 # much from platforms to platform...
@@ -395,46 +727,46 @@ LPICmeup:
 	.space	`64-9*4`
 ___
 $code.=<<___ if ($SZ==8);
-	.long	0x428a2f98,0xd728ae22,0x71374491,0x23ef65cd
-	.long	0xb5c0fbcf,0xec4d3b2f,0xe9b5dba5,0x8189dbbc
-	.long	0x3956c25b,0xf348b538,0x59f111f1,0xb605d019
-	.long	0x923f82a4,0xaf194f9b,0xab1c5ed5,0xda6d8118
-	.long	0xd807aa98,0xa3030242,0x12835b01,0x45706fbe
-	.long	0x243185be,0x4ee4b28c,0x550c7dc3,0xd5ffb4e2
-	.long	0x72be5d74,0xf27b896f,0x80deb1fe,0x3b1696b1
-	.long	0x9bdc06a7,0x25c71235,0xc19bf174,0xcf692694
-	.long	0xe49b69c1,0x9ef14ad2,0xefbe4786,0x384f25e3
-	.long	0x0fc19dc6,0x8b8cd5b5,0x240ca1cc,0x77ac9c65
-	.long	0x2de92c6f,0x592b0275,0x4a7484aa,0x6ea6e483
-	.long	0x5cb0a9dc,0xbd41fbd4,0x76f988da,0x831153b5
-	.long	0x983e5152,0xee66dfab,0xa831c66d,0x2db43210
-	.long	0xb00327c8,0x98fb213f,0xbf597fc7,0xbeef0ee4
-	.long	0xc6e00bf3,0x3da88fc2,0xd5a79147,0x930aa725
-	.long	0x06ca6351,0xe003826f,0x14292967,0x0a0e6e70
-	.long	0x27b70a85,0x46d22ffc,0x2e1b2138,0x5c26c926
-	.long	0x4d2c6dfc,0x5ac42aed,0x53380d13,0x9d95b3df
-	.long	0x650a7354,0x8baf63de,0x766a0abb,0x3c77b2a8
-	.long	0x81c2c92e,0x47edaee6,0x92722c85,0x1482353b
-	.long	0xa2bfe8a1,0x4cf10364,0xa81a664b,0xbc423001
-	.long	0xc24b8b70,0xd0f89791,0xc76c51a3,0x0654be30
-	.long	0xd192e819,0xd6ef5218,0xd6990624,0x5565a910
-	.long	0xf40e3585,0x5771202a,0x106aa070,0x32bbd1b8
-	.long	0x19a4c116,0xb8d2d0c8,0x1e376c08,0x5141ab53
-	.long	0x2748774c,0xdf8eeb99,0x34b0bcb5,0xe19b48a8
-	.long	0x391c0cb3,0xc5c95a63,0x4ed8aa4a,0xe3418acb
-	.long	0x5b9cca4f,0x7763e373,0x682e6ff3,0xd6b2b8a3
-	.long	0x748f82ee,0x5defb2fc,0x78a5636f,0x43172f60
-	.long	0x84c87814,0xa1f0ab72,0x8cc70208,0x1a6439ec
-	.long	0x90befffa,0x23631e28,0xa4506ceb,0xde82bde9
-	.long	0xbef9a3f7,0xb2c67915,0xc67178f2,0xe372532b
-	.long	0xca273ece,0xea26619c,0xd186b8c7,0x21c0c207
-	.long	0xeada7dd6,0xcde0eb1e,0xf57d4f7f,0xee6ed178
-	.long	0x06f067aa,0x72176fba,0x0a637dc5,0xa2c898a6
-	.long	0x113f9804,0xbef90dae,0x1b710b35,0x131c471b
-	.long	0x28db77f5,0x23047d84,0x32caab7b,0x40c72493
-	.long	0x3c9ebe0a,0x15c9bebc,0x431d67c4,0x9c100d4c
-	.long	0x4cc5d4be,0xcb3e42b6,0x597f299c,0xfc657e2a
-	.long	0x5fcb6fab,0x3ad6faec,0x6c44198c,0x4a475817
+	.quad	0x428a2f98d728ae22,0x7137449123ef65cd
+	.quad	0xb5c0fbcfec4d3b2f,0xe9b5dba58189dbbc
+	.quad	0x3956c25bf348b538,0x59f111f1b605d019
+	.quad	0x923f82a4af194f9b,0xab1c5ed5da6d8118
+	.quad	0xd807aa98a3030242,0x12835b0145706fbe
+	.quad	0x243185be4ee4b28c,0x550c7dc3d5ffb4e2
+	.quad	0x72be5d74f27b896f,0x80deb1fe3b1696b1
+	.quad	0x9bdc06a725c71235,0xc19bf174cf692694
+	.quad	0xe49b69c19ef14ad2,0xefbe4786384f25e3
+	.quad	0x0fc19dc68b8cd5b5,0x240ca1cc77ac9c65
+	.quad	0x2de92c6f592b0275,0x4a7484aa6ea6e483
+	.quad	0x5cb0a9dcbd41fbd4,0x76f988da831153b5
+	.quad	0x983e5152ee66dfab,0xa831c66d2db43210
+	.quad	0xb00327c898fb213f,0xbf597fc7beef0ee4
+	.quad	0xc6e00bf33da88fc2,0xd5a79147930aa725
+	.quad	0x06ca6351e003826f,0x142929670a0e6e70
+	.quad	0x27b70a8546d22ffc,0x2e1b21385c26c926
+	.quad	0x4d2c6dfc5ac42aed,0x53380d139d95b3df
+	.quad	0x650a73548baf63de,0x766a0abb3c77b2a8
+	.quad	0x81c2c92e47edaee6,0x92722c851482353b
+	.quad	0xa2bfe8a14cf10364,0xa81a664bbc423001
+	.quad	0xc24b8b70d0f89791,0xc76c51a30654be30
+	.quad	0xd192e819d6ef5218,0xd69906245565a910
+	.quad	0xf40e35855771202a,0x106aa07032bbd1b8
+	.quad	0x19a4c116b8d2d0c8,0x1e376c085141ab53
+	.quad	0x2748774cdf8eeb99,0x34b0bcb5e19b48a8
+	.quad	0x391c0cb3c5c95a63,0x4ed8aa4ae3418acb
+	.quad	0x5b9cca4f7763e373,0x682e6ff3d6b2b8a3
+	.quad	0x748f82ee5defb2fc,0x78a5636f43172f60
+	.quad	0x84c87814a1f0ab72,0x8cc702081a6439ec
+	.quad	0x90befffa23631e28,0xa4506cebde82bde9
+	.quad	0xbef9a3f7b2c67915,0xc67178f2e372532b
+	.quad	0xca273eceea26619c,0xd186b8c721c0c207
+	.quad	0xeada7dd6cde0eb1e,0xf57d4f7fee6ed178
+	.quad	0x06f067aa72176fba,0x0a637dc5a2c898a6
+	.quad	0x113f9804bef90dae,0x1b710b35131c471b
+	.quad	0x28db77f523047d84,0x32caab7b40c72493
+	.quad	0x3c9ebe0a15c9bebc,0x431d67c49c100d4c
+	.quad	0x4cc5d4becb3e42b6,0x597f299cfc657e2a
+	.quad	0x5fcb6fab3ad6faec,0x6c44198c4a475817
 ___
 $code.=<<___ if ($SZ==4);
 	.long	0x428a2f98,0x71374491,0xb5c0fbcf,0xe9b5dba5
diff --git a/crypto/sha/asm/sha512p8-ppc.pl b/crypto/sha/asm/sha512p8-ppc.pl
new file mode 100755
index 0000000..a316b31
--- /dev/null
+++ b/crypto/sha/asm/sha512p8-ppc.pl
@@ -0,0 +1,423 @@
+#!/usr/bin/env perl
+
+# ====================================================================
+# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+
+# SHA256/512 for PowerISA v2.07.
+#
+# Accurate performance measurements are problematic, because it's
+# always virtualized setup with possibly throttled processor.
+# Relative comparison is therefore more informative. This module is
+# ~60% faster than integer-only sha512-ppc.pl. To anchor to something
+# else, SHA256 is 24% slower than sha1-ppc.pl and 2.5x slower than
+# hardware-assisted aes-128-cbc encrypt. SHA512 is 20% faster than
+# sha1-ppc.pl and 1.6x slower than aes-128-cbc. Another interesting
+# result is degree of computational resources' utilization. POWER8 is
+# "massively multi-threaded chip" and difference between single- and
+# maximum multi-process benchmark results tells that utlization is
+# whooping 94%. For sha512-ppc.pl we get [not unimpressive] 84% and
+# for sha1-ppc.pl - 73%. 100% means that multi-process result equals
+# to single-process one, given that all threads end up on the same
+# physical core.
+
+$flavour=shift;
+$output =shift;
+
+if ($flavour =~ /64/) {
+	$SIZE_T=8;
+	$LRSAVE=2*$SIZE_T;
+	$STU="stdu";
+	$POP="ld";
+	$PUSH="std";
+} elsif ($flavour =~ /32/) {
+	$SIZE_T=4;
+	$LRSAVE=$SIZE_T;
+	$STU="stwu";
+	$POP="lwz";
+	$PUSH="stw";
+} else { die "nonsense $flavour"; }
+
+$LENDIAN=($flavour=~/le/);
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or
+( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or
+die "can't locate ppc-xlate.pl";
+
+open STDOUT,"| $^X $xlate $flavour $output" || die "can't call $xlate: $!";
+
+if ($output =~ /512/) {
+	$bits=512;
+	$SZ=8;
+	$sz="d";
+	$rounds=80;
+} else {
+	$bits=256;
+	$SZ=4;
+	$sz="w";
+	$rounds=64;
+}
+
+$func="sha${bits}_block_p8";
+$FRAME=8*$SIZE_T;
+
+$sp ="r1";
+$toc="r2";
+$ctx="r3";
+$inp="r4";
+$num="r5";
+$Tbl="r6";
+$idx="r7";
+$lrsave="r8";
+$offload="r11";
+$vrsave="r12";
+($x00,$x10,$x20,$x30,$x40,$x50,$x60,$x70)=map("r$_",(0,10,26..31));
+
+@V=($A,$B,$C,$D,$E,$F,$G,$H)=map("v$_",(0..7));
+@X=map("v$_",(8..23));
+($Ki,$Func,$S0,$S1,$s0,$s1,$lemask)=map("v$_",(24..31));
+
+sub ROUND {
+my ($i,$a,$b,$c,$d,$e,$f,$g,$h)=@_;
+my $j=($i+1)%16;
+
+$code.=<<___		if ($i<15 && ($i%(16/$SZ))==(16/$SZ-1));
+	lvx_u		@X[$i+1],0,$inp		; load X[i] in advance
+	addi		$inp,$inp,16
+___
+$code.=<<___		if ($i<16 && ($i%(16/$SZ)));
+	vsldoi		@X[$i],@X[$i-1],@X[$i-1],$SZ
+___
+$code.=<<___		if ($LENDIAN && $i<16 && ($i%(16/$SZ))==0);
+	vperm		@X[$i],@X[$i],@X[$i],$lemask
+___
+$code.=<<___;
+	`"vshasigma${sz}	$s0,@X[($j+1)%16],0,0"		if ($i>=15)`
+	vsel		$Func,$g,$f,$e		; Ch(e,f,g)
+	vshasigma${sz}	$S1,$e,1,15		; Sigma1(e)
+	vaddu${sz}m	$h,$h,@X[$i%16]		; h+=X[i]
+	vshasigma${sz}	$S0,$a,1,0		; Sigma0(a)
+	`"vshasigma${sz}	$s1,@X[($j+14)%16],0,15"	if ($i>=15)`
+	vaddu${sz}m	$h,$h,$Func		; h+=Ch(e,f,g)
+	vxor		$Func,$a,$b
+	`"vaddu${sz}m		@X[$j],@X[$j],@X[($j+9)%16]"	if ($i>=15)`
+	vaddu${sz}m	$h,$h,$S1		; h+=Sigma1(e)
+	vsel		$Func,$b,$c,$Func	; Maj(a,b,c)
+	vaddu${sz}m	$g,$g,$Ki		; future h+=K[i]
+	vaddu${sz}m	$d,$d,$h		; d+=h
+	vaddu${sz}m	$S0,$S0,$Func		; Sigma0(a)+Maj(a,b,c)
+	`"vaddu${sz}m		@X[$j],@X[$j],$s0"		if ($i>=15)`
+	lvx		$Ki,$idx,$Tbl		; load next K[i]
+	addi		$idx,$idx,16
+	vaddu${sz}m	$h,$h,$S0		; h+=Sigma0(a)+Maj(a,b,c)
+	`"vaddu${sz}m		@X[$j],@X[$j],$s1"		if ($i>=15)`
+___
+}
+
+$code=<<___;
+.machine	"any"
+.text
+
+.globl	$func
+.align	6
+$func:
+	$STU		$sp,-`($FRAME+21*16+6*$SIZE_T)`($sp)
+	mflr		$lrsave
+	li		r10,`$FRAME+8*16+15`
+	li		r11,`$FRAME+8*16+31`
+	stvx		v20,r10,$sp		# ABI says so
+	addi		r10,r10,32
+	mfspr		$vrsave,256
+	stvx		v21,r11,$sp
+	addi		r11,r11,32
+	stvx		v22,r10,$sp
+	addi		r10,r10,32
+	stvx		v23,r11,$sp
+	addi		r11,r11,32
+	stvx		v24,r10,$sp
+	addi		r10,r10,32
+	stvx		v25,r11,$sp
+	addi		r11,r11,32
+	stvx		v26,r10,$sp
+	addi		r10,r10,32
+	stvx		v27,r11,$sp
+	addi		r11,r11,32
+	stvx		v28,r10,$sp
+	addi		r10,r10,32
+	stvx		v29,r11,$sp
+	addi		r11,r11,32
+	stvx		v30,r10,$sp
+	stvx		v31,r11,$sp
+	li		r11,-1
+	stw		$vrsave,`$FRAME+21*16-4`($sp)	# save vrsave
+	li		$x10,0x10
+	$PUSH		r26,`$FRAME+21*16+0*$SIZE_T`($sp)
+	li		$x20,0x20
+	$PUSH		r27,`$FRAME+21*16+1*$SIZE_T`($sp)
+	li		$x30,0x30
+	$PUSH		r28,`$FRAME+21*16+2*$SIZE_T`($sp)
+	li		$x40,0x40
+	$PUSH		r29,`$FRAME+21*16+3*$SIZE_T`($sp)
+	li		$x50,0x50
+	$PUSH		r30,`$FRAME+21*16+4*$SIZE_T`($sp)
+	li		$x60,0x60
+	$PUSH		r31,`$FRAME+21*16+5*$SIZE_T`($sp)
+	li		$x70,0x70
+	$PUSH		$lrsave,`$FRAME+21*16+6*$SIZE_T+$LRSAVE`($sp)
+	mtspr		256,r11
+
+	bl		LPICmeup
+	addi		$offload,$sp,$FRAME+15
+___
+$code.=<<___		if ($LENDIAN);
+	li		$idx,8
+	lvsl		$lemask,0,$idx
+	vspltisb	$Ki,0x0f
+	vxor		$lemask,$lemask,$Ki
+___
+$code.=<<___		if ($SZ==4);
+	lvx_4w		$A,$x00,$ctx
+	lvx_4w		$E,$x10,$ctx
+	vsldoi		$B,$A,$A,4		# unpack
+	vsldoi		$C,$A,$A,8
+	vsldoi		$D,$A,$A,12
+	vsldoi		$F,$E,$E,4
+	vsldoi		$G,$E,$E,8
+	vsldoi		$H,$E,$E,12
+___
+$code.=<<___		if ($SZ==8);
+	lvx_u		$A,$x00,$ctx
+	lvx_u		$C,$x10,$ctx
+	lvx_u		$E,$x20,$ctx
+	vsldoi		$B,$A,$A,8		# unpack
+	lvx_u		$G,$x30,$ctx
+	vsldoi		$D,$C,$C,8
+	vsldoi		$F,$E,$E,8
+	vsldoi		$H,$G,$G,8
+___
+$code.=<<___;
+	li		r0,`($rounds-16)/16`	# inner loop counter
+	b		Loop
+.align	5
+Loop:
+	lvx		$Ki,$x00,$Tbl
+	li		$idx,16
+	lvx_u		@X[0],0,$inp
+	addi		$inp,$inp,16
+	stvx		$A,$x00,$offload	# offload $A-$H
+	stvx		$B,$x10,$offload
+	stvx		$C,$x20,$offload
+	stvx		$D,$x30,$offload
+	stvx		$E,$x40,$offload
+	stvx		$F,$x50,$offload
+	stvx		$G,$x60,$offload
+	stvx		$H,$x70,$offload
+	vaddu${sz}m	$H,$H,$Ki		# h+K[i]
+	lvx		$Ki,$idx,$Tbl
+	addi		$idx,$idx,16
+___
+for ($i=0;$i<16;$i++)	{ &ROUND($i,@V); unshift(@V,pop(@V)); }
+$code.=<<___;
+	mtctr		r0
+	b		L16_xx
+.align	5
+L16_xx:
+___
+for (;$i<32;$i++)	{ &ROUND($i,@V); unshift(@V,pop(@V)); }
+$code.=<<___;
+	bdnz		L16_xx
+
+	lvx		@X[2],$x00,$offload
+	subic.		$num,$num,1
+	lvx		@X[3],$x10,$offload
+	vaddu${sz}m	$A,$A,@X[2]
+	lvx		@X[4],$x20,$offload
+	vaddu${sz}m	$B,$B,@X[3]
+	lvx		@X[5],$x30,$offload
+	vaddu${sz}m	$C,$C,@X[4]
+	lvx		@X[6],$x40,$offload
+	vaddu${sz}m	$D,$D,@X[5]
+	lvx		@X[7],$x50,$offload
+	vaddu${sz}m	$E,$E,@X[6]
+	lvx		@X[8],$x60,$offload
+	vaddu${sz}m	$F,$F,@X[7]
+	lvx		@X[9],$x70,$offload
+	vaddu${sz}m	$G,$G,@X[8]
+	vaddu${sz}m	$H,$H,@X[9]
+	bne		Loop
+___
+$code.=<<___		if ($SZ==4);
+	lvx		@X[0],$idx,$Tbl
+	addi		$idx,$idx,16
+	vperm		$A,$A,$B,$Ki		# pack the answer
+	lvx		@X[1],$idx,$Tbl
+	vperm		$E,$E,$F,$Ki
+	vperm		$A,$A,$C,@X[0]
+	vperm		$E,$E,$G,@X[0]
+	vperm		$A,$A,$D,@X[1]
+	vperm		$E,$E,$H,@X[1]
+	stvx_4w		$A,$x00,$ctx
+	stvx_4w		$E,$x10,$ctx
+___
+$code.=<<___		if ($SZ==8);
+	vperm		$A,$A,$B,$Ki		# pack the answer
+	vperm		$C,$C,$D,$Ki
+	vperm		$E,$E,$F,$Ki
+	vperm		$G,$G,$H,$Ki
+	stvx_u		$A,$x00,$ctx
+	stvx_u		$C,$x10,$ctx
+	stvx_u		$E,$x20,$ctx
+	stvx_u		$G,$x30,$ctx
+___
+$code.=<<___;
+	li		r10,`$FRAME+8*16+15`
+	mtlr		$lrsave
+	li		r11,`$FRAME+8*16+31`
+	mtspr		256,$vrsave
+	lvx		v20,r10,$sp		# ABI says so
+	addi		r10,r10,32
+	lvx		v21,r11,$sp
+	addi		r11,r11,32
+	lvx		v22,r10,$sp
+	addi		r10,r10,32
+	lvx		v23,r11,$sp
+	addi		r11,r11,32
+	lvx		v24,r10,$sp
+	addi		r10,r10,32
+	lvx		v25,r11,$sp
+	addi		r11,r11,32
+	lvx		v26,r10,$sp
+	addi		r10,r10,32
+	lvx		v27,r11,$sp
+	addi		r11,r11,32
+	lvx		v28,r10,$sp
+	addi		r10,r10,32
+	lvx		v29,r11,$sp
+	addi		r11,r11,32
+	lvx		v30,r10,$sp
+	lvx		v31,r11,$sp
+	$POP		r26,`$FRAME+21*16+0*$SIZE_T`($sp)
+	$POP		r27,`$FRAME+21*16+1*$SIZE_T`($sp)
+	$POP		r28,`$FRAME+21*16+2*$SIZE_T`($sp)
+	$POP		r29,`$FRAME+21*16+3*$SIZE_T`($sp)
+	$POP		r30,`$FRAME+21*16+4*$SIZE_T`($sp)
+	$POP		r31,`$FRAME+21*16+5*$SIZE_T`($sp)
+	addi		$sp,$sp,`$FRAME+21*16+6*$SIZE_T`
+	blr
+	.long		0
+	.byte		0,12,4,1,0x80,6,3,0
+	.long		0
+.size	$func,.-$func
+___
+
+# Ugly hack here, because PPC assembler syntax seem to vary too
+# much from platforms to platform...
+$code.=<<___;
+.align	6
+LPICmeup:
+	mflr	r0
+	bcl	20,31,\$+4
+	mflr	$Tbl	; vvvvvv "distance" between . and 1st data entry
+	addi	$Tbl,$Tbl,`64-8`
+	mtlr	r0
+	blr
+	.long	0
+	.byte	0,12,0x14,0,0,0,0,0
+	.space	`64-9*4`
+___
+
+if ($SZ==8) {
+    local *table = sub {
+	foreach(@_) { $code.=".quad	$_,$_\n"; }
+    };
+    table(
+	"0x428a2f98d728ae22","0x7137449123ef65cd",
+	"0xb5c0fbcfec4d3b2f","0xe9b5dba58189dbbc",
+	"0x3956c25bf348b538","0x59f111f1b605d019",
+	"0x923f82a4af194f9b","0xab1c5ed5da6d8118",
+	"0xd807aa98a3030242","0x12835b0145706fbe",
+	"0x243185be4ee4b28c","0x550c7dc3d5ffb4e2",
+	"0x72be5d74f27b896f","0x80deb1fe3b1696b1",
+	"0x9bdc06a725c71235","0xc19bf174cf692694",
+	"0xe49b69c19ef14ad2","0xefbe4786384f25e3",
+	"0x0fc19dc68b8cd5b5","0x240ca1cc77ac9c65",
+	"0x2de92c6f592b0275","0x4a7484aa6ea6e483",
+	"0x5cb0a9dcbd41fbd4","0x76f988da831153b5",
+	"0x983e5152ee66dfab","0xa831c66d2db43210",
+	"0xb00327c898fb213f","0xbf597fc7beef0ee4",
+	"0xc6e00bf33da88fc2","0xd5a79147930aa725",
+	"0x06ca6351e003826f","0x142929670a0e6e70",
+	"0x27b70a8546d22ffc","0x2e1b21385c26c926",
+	"0x4d2c6dfc5ac42aed","0x53380d139d95b3df",
+	"0x650a73548baf63de","0x766a0abb3c77b2a8",
+	"0x81c2c92e47edaee6","0x92722c851482353b",
+	"0xa2bfe8a14cf10364","0xa81a664bbc423001",
+	"0xc24b8b70d0f89791","0xc76c51a30654be30",
+	"0xd192e819d6ef5218","0xd69906245565a910",
+	"0xf40e35855771202a","0x106aa07032bbd1b8",
+	"0x19a4c116b8d2d0c8","0x1e376c085141ab53",
+	"0x2748774cdf8eeb99","0x34b0bcb5e19b48a8",
+	"0x391c0cb3c5c95a63","0x4ed8aa4ae3418acb",
+	"0x5b9cca4f7763e373","0x682e6ff3d6b2b8a3",
+	"0x748f82ee5defb2fc","0x78a5636f43172f60",
+	"0x84c87814a1f0ab72","0x8cc702081a6439ec",
+	"0x90befffa23631e28","0xa4506cebde82bde9",
+	"0xbef9a3f7b2c67915","0xc67178f2e372532b",
+	"0xca273eceea26619c","0xd186b8c721c0c207",
+	"0xeada7dd6cde0eb1e","0xf57d4f7fee6ed178",
+	"0x06f067aa72176fba","0x0a637dc5a2c898a6",
+	"0x113f9804bef90dae","0x1b710b35131c471b",
+	"0x28db77f523047d84","0x32caab7b40c72493",
+	"0x3c9ebe0a15c9bebc","0x431d67c49c100d4c",
+	"0x4cc5d4becb3e42b6","0x597f299cfc657e2a",
+	"0x5fcb6fab3ad6faec","0x6c44198c4a475817","0");
+$code.=<<___	if (!$LENDIAN);
+.quad	0x0001020304050607,0x1011121314151617
+___
+$code.=<<___	if ($LENDIAN);	# quad-swapped
+.quad	0x1011121314151617,0x0001020304050607
+___
+} else {
+    local *table = sub {
+	foreach(@_) { $code.=".long	$_,$_,$_,$_\n"; }
+    };
+    table(
+	"0x428a2f98","0x71374491","0xb5c0fbcf","0xe9b5dba5",
+	"0x3956c25b","0x59f111f1","0x923f82a4","0xab1c5ed5",
+	"0xd807aa98","0x12835b01","0x243185be","0x550c7dc3",
+	"0x72be5d74","0x80deb1fe","0x9bdc06a7","0xc19bf174",
+	"0xe49b69c1","0xefbe4786","0x0fc19dc6","0x240ca1cc",
+	"0x2de92c6f","0x4a7484aa","0x5cb0a9dc","0x76f988da",
+	"0x983e5152","0xa831c66d","0xb00327c8","0xbf597fc7",
+	"0xc6e00bf3","0xd5a79147","0x06ca6351","0x14292967",
+	"0x27b70a85","0x2e1b2138","0x4d2c6dfc","0x53380d13",
+	"0x650a7354","0x766a0abb","0x81c2c92e","0x92722c85",
+	"0xa2bfe8a1","0xa81a664b","0xc24b8b70","0xc76c51a3",
+	"0xd192e819","0xd6990624","0xf40e3585","0x106aa070",
+	"0x19a4c116","0x1e376c08","0x2748774c","0x34b0bcb5",
+	"0x391c0cb3","0x4ed8aa4a","0x5b9cca4f","0x682e6ff3",
+	"0x748f82ee","0x78a5636f","0x84c87814","0x8cc70208",
+	"0x90befffa","0xa4506ceb","0xbef9a3f7","0xc67178f2","0");
+$code.=<<___	if (!$LENDIAN);
+.long	0x00010203,0x10111213,0x10111213,0x10111213
+.long	0x00010203,0x04050607,0x10111213,0x10111213
+.long	0x00010203,0x04050607,0x08090a0b,0x10111213
+___
+$code.=<<___	if ($LENDIAN);	# word-swapped
+.long	0x10111213,0x10111213,0x10111213,0x00010203
+.long	0x10111213,0x10111213,0x04050607,0x00010203
+.long	0x10111213,0x08090a0b,0x04050607,0x00010203
+___
+}
+$code.=<<___;
+.asciz	"SHA${bits} for PowerISA 2.07, CRYPTOGAMS by <appro\@openssl.org>"
+.align	2
+___
+
+$code =~ s/\`([^\`]*)\`/eval $1/gem;
+print $code;
+close STDOUT;