import of OpenSSL 0.9.8k

2 files changed, 282 insertions, 641 deletions

diff --git a/lib/libcrypto/aes/asm/aes-ppc.pl b/lib/libcrypto/aes/asm/aes-ppc.pl
index 7c52cbe5f9f..ce427655ef7 100644
--- a/lib/libcrypto/aes/asm/aes-ppc.pl
+++ b/lib/libcrypto/aes/asm/aes-ppc.pl
@@ -7,7 +7,7 @@
 # details see http://www.openssl.org/~appro/cryptogams/.
 # ====================================================================
 
-# Needs more work: key setup, CBC routine...
+# Needs more work: key setup, page boundaries, CBC routine...
 #
 # ppc_AES_[en|de]crypt perform at 18 cycles per byte processed with
 # 128-bit key, which is ~40% better than 64-bit code generated by gcc
@@ -16,30 +16,15 @@
 # at 1/2 of ppc_AES_encrypt speed, while ppc_AES_decrypt_compact -
 # at 1/3 of ppc_AES_decrypt.
 
-# February 2010
-#
-# Rescheduling instructions to favour Power6 pipeline gave 10%
-# performance improvement on the platfrom in question (and marginal
-# improvement even on others). It should be noted that Power6 fails
-# to process byte in 18 cycles, only in 23, because it fails to issue
-# 4 load instructions in two cycles, only in 3. As result non-compact
-# block subroutines are 25% slower than one would expect. Compact
-# functions scale better, because they have pure computational part,
-# which scales perfectly with clock frequency. To be specific
-# ppc_AES_encrypt_compact operates at 42 cycles per byte, while
-# ppc_AES_decrypt_compact - at 55 (in 64-bit build).
-
 $flavour = 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";
@@ -118,19 +103,15 @@ LAES_Te:
 	addi	$Tbl0,$Tbl0,`128-8`
 	mtlr	r0
 	blr
-	.long	0
-	.byte	0,12,0x14,0,0,0,0,0
-	.space	`64-9*4`
+	.space	`32-24`
 LAES_Td:
 	mflr	r0
 	bcl	20,31,\$+4
 	mflr	$Tbl0	;    vvvvvvvv "distance" between . and 1st data entry
-	addi	$Tbl0,$Tbl0,`128-64-8+2048+256`
+	addi	$Tbl0,$Tbl0,`128-8-32+2048+256`
 	mtlr	r0
 	blr
-	.long	0
-	.byte	0,12,0x14,0,0,0,0,0
-	.space	`128-64-9*4`
+	.space	`128-32-24`
 ___
 &_data_word(
 	0xc66363a5, 0xf87c7c84, 0xee777799, 0xf67b7b8d,
@@ -334,9 +315,10 @@ $code.=<<___;
 .globl	.AES_encrypt
 .align	7
 .AES_encrypt:
-	$STU	$sp,-$FRAME($sp)
 	mflr	r0
+	$STU	$sp,-$FRAME($sp)
 
+	$PUSH	r0,`$FRAME-$SIZE_T*21`($sp)
 	$PUSH	$toc,`$FRAME-$SIZE_T*20`($sp)
 	$PUSH	r13,`$FRAME-$SIZE_T*19`($sp)
 	$PUSH	r14,`$FRAME-$SIZE_T*18`($sp)
@@ -357,14 +339,7 @@ $code.=<<___;
 	$PUSH	r29,`$FRAME-$SIZE_T*3`($sp)
 	$PUSH	r30,`$FRAME-$SIZE_T*2`($sp)
 	$PUSH	r31,`$FRAME-$SIZE_T*1`($sp)
-	$PUSH	r0,`$FRAME+$LRSAVE`($sp)
-
-	andi.	$t0,$inp,3
-	andi.	$t1,$out,3
-	or.	$t0,$t0,$t1
-	bne	Lenc_unaligned
 
-Lenc_unaligned_ok:
 	lwz	$s0,0($inp)
 	lwz	$s1,4($inp)
 	lwz	$s2,8($inp)
@@ -375,80 +350,8 @@ Lenc_unaligned_ok:
 	stw	$s1,4($out)
 	stw	$s2,8($out)
 	stw	$s3,12($out)
-	b	Lenc_done
-
-Lenc_unaligned:
-	subfic	$t0,$inp,4096
-	subfic	$t1,$out,4096
-	andi.	$t0,$t0,4096-16
-	beq	Lenc_xpage
-	andi.	$t1,$t1,4096-16
-	bne	Lenc_unaligned_ok
 
-Lenc_xpage:
-	lbz	$acc00,0($inp)
-	lbz	$acc01,1($inp)
-	lbz	$acc02,2($inp)
-	lbz	$s0,3($inp)
-	lbz	$acc04,4($inp)
-	lbz	$acc05,5($inp)
-	lbz	$acc06,6($inp)
-	lbz	$s1,7($inp)
-	lbz	$acc08,8($inp)
-	lbz	$acc09,9($inp)
-	lbz	$acc10,10($inp)
-	insrwi	$s0,$acc00,8,0
-	lbz	$s2,11($inp)
-	insrwi	$s1,$acc04,8,0
-	lbz	$acc12,12($inp)
-	insrwi	$s0,$acc01,8,8
-	lbz	$acc13,13($inp)
-	insrwi	$s1,$acc05,8,8
-	lbz	$acc14,14($inp)
-	insrwi	$s0,$acc02,8,16
-	lbz	$s3,15($inp)
-	insrwi	$s1,$acc06,8,16
-	insrwi	$s2,$acc08,8,0
-	insrwi	$s3,$acc12,8,0
-	insrwi	$s2,$acc09,8,8
-	insrwi	$s3,$acc13,8,8
-	insrwi	$s2,$acc10,8,16
-	insrwi	$s3,$acc14,8,16
-
-	bl	LAES_Te
-	bl	Lppc_AES_encrypt_compact
-
-	extrwi	$acc00,$s0,8,0
-	extrwi	$acc01,$s0,8,8
-	stb	$acc00,0($out)
-	extrwi	$acc02,$s0,8,16
-	stb	$acc01,1($out)
-	stb	$acc02,2($out)
-	extrwi	$acc04,$s1,8,0
-	stb	$s0,3($out)
-	extrwi	$acc05,$s1,8,8
-	stb	$acc04,4($out)
-	extrwi	$acc06,$s1,8,16
-	stb	$acc05,5($out)
-	stb	$acc06,6($out)
-	extrwi	$acc08,$s2,8,0
-	stb	$s1,7($out)
-	extrwi	$acc09,$s2,8,8
-	stb	$acc08,8($out)
-	extrwi	$acc10,$s2,8,16
-	stb	$acc09,9($out)
-	stb	$acc10,10($out)
-	extrwi	$acc12,$s3,8,0
-	stb	$s2,11($out)
-	extrwi	$acc13,$s3,8,8
-	stb	$acc12,12($out)
-	extrwi	$acc14,$s3,8,16
-	stb	$acc13,13($out)
-	stb	$acc14,14($out)
-	stb	$s3,15($out)
-
-Lenc_done:
-	$POP	r0,`$FRAME+$LRSAVE`($sp)
+	$POP	r0,`$FRAME-$SIZE_T*21`($sp)
 	$POP	$toc,`$FRAME-$SIZE_T*20`($sp)
 	$POP	r13,`$FRAME-$SIZE_T*19`($sp)
 	$POP	r14,`$FRAME-$SIZE_T*18`($sp)
@@ -472,21 +375,18 @@ Lenc_done:
 	mtlr	r0
 	addi	$sp,$sp,$FRAME
 	blr
-	.long	0
-	.byte	0,12,4,1,0x80,18,3,0
-	.long	0
 
-.align	5
+.align	4
 Lppc_AES_encrypt:
 	lwz	$acc00,240($key)
-	addi	$Tbl1,$Tbl0,3
 	lwz	$t0,0($key)
-	addi	$Tbl2,$Tbl0,2
 	lwz	$t1,4($key)
-	addi	$Tbl3,$Tbl0,1
 	lwz	$t2,8($key)
-	addi	$acc00,$acc00,-1
 	lwz	$t3,12($key)
+	addi	$Tbl1,$Tbl0,3
+	addi	$Tbl2,$Tbl0,2
+	addi	$Tbl3,$Tbl0,1
+	addi	$acc00,$acc00,-1
 	addi	$key,$key,16
 	xor	$s0,$s0,$t0
 	xor	$s1,$s1,$t1
@@ -497,47 +397,47 @@ Lppc_AES_encrypt:
 Lenc_loop:
 	rlwinm	$acc00,$s0,`32-24+3`,21,28
 	rlwinm	$acc01,$s1,`32-24+3`,21,28
+	lwz	$t0,0($key)
+	lwz	$t1,4($key)
 	rlwinm	$acc02,$s2,`32-24+3`,21,28
 	rlwinm	$acc03,$s3,`32-24+3`,21,28
-	lwz	$t0,0($key)
+	lwz	$t2,8($key)
+	lwz	$t3,12($key)
 	rlwinm	$acc04,$s1,`32-16+3`,21,28
-	lwz	$t1,4($key)
 	rlwinm	$acc05,$s2,`32-16+3`,21,28
-	lwz	$t2,8($key)
+	lwzx	$acc00,$Tbl0,$acc00
+	lwzx	$acc01,$Tbl0,$acc01
 	rlwinm	$acc06,$s3,`32-16+3`,21,28
-	lwz	$t3,12($key)
 	rlwinm	$acc07,$s0,`32-16+3`,21,28
-	lwzx	$acc00,$Tbl0,$acc00
+	lwzx	$acc02,$Tbl0,$acc02
+	lwzx	$acc03,$Tbl0,$acc03
 	rlwinm	$acc08,$s2,`32-8+3`,21,28
-	lwzx	$acc01,$Tbl0,$acc01
 	rlwinm	$acc09,$s3,`32-8+3`,21,28
-	lwzx	$acc02,$Tbl0,$acc02
+	lwzx	$acc04,$Tbl1,$acc04
+	lwzx	$acc05,$Tbl1,$acc05
 	rlwinm	$acc10,$s0,`32-8+3`,21,28
-	lwzx	$acc03,$Tbl0,$acc03
 	rlwinm	$acc11,$s1,`32-8+3`,21,28
-	lwzx	$acc04,$Tbl1,$acc04
+	lwzx	$acc06,$Tbl1,$acc06
+	lwzx	$acc07,$Tbl1,$acc07
 	rlwinm	$acc12,$s3,`0+3`,21,28
-	lwzx	$acc05,$Tbl1,$acc05
 	rlwinm	$acc13,$s0,`0+3`,21,28
-	lwzx	$acc06,$Tbl1,$acc06
+	lwzx	$acc08,$Tbl2,$acc08
+	lwzx	$acc09,$Tbl2,$acc09
 	rlwinm	$acc14,$s1,`0+3`,21,28
-	lwzx	$acc07,$Tbl1,$acc07
 	rlwinm	$acc15,$s2,`0+3`,21,28
-	lwzx	$acc08,$Tbl2,$acc08
+	lwzx	$acc10,$Tbl2,$acc10
+	lwzx	$acc11,$Tbl2,$acc11
 	xor	$t0,$t0,$acc00
-	lwzx	$acc09,$Tbl2,$acc09
 	xor	$t1,$t1,$acc01
-	lwzx	$acc10,$Tbl2,$acc10
+	lwzx	$acc12,$Tbl3,$acc12
+	lwzx	$acc13,$Tbl3,$acc13
 	xor	$t2,$t2,$acc02
-	lwzx	$acc11,$Tbl2,$acc11
 	xor	$t3,$t3,$acc03
-	lwzx	$acc12,$Tbl3,$acc12
+	lwzx	$acc14,$Tbl3,$acc14
+	lwzx	$acc15,$Tbl3,$acc15
 	xor	$t0,$t0,$acc04
-	lwzx	$acc13,$Tbl3,$acc13
 	xor	$t1,$t1,$acc05
-	lwzx	$acc14,$Tbl3,$acc14
 	xor	$t2,$t2,$acc06
-	lwzx	$acc15,$Tbl3,$acc15
 	xor	$t3,$t3,$acc07
 	xor	$t0,$t0,$acc08
 	xor	$t1,$t1,$acc09
@@ -552,61 +452,61 @@ Lenc_loop:
 
 	addi	$Tbl2,$Tbl0,2048
 	nop
-	lwz	$t0,0($key)
+	lwz	$acc08,`2048+0`($Tbl0)	! prefetch Te4
+	lwz	$acc09,`2048+32`($Tbl0)
+	lwz	$acc10,`2048+64`($Tbl0)
+	lwz	$acc11,`2048+96`($Tbl0)
+	lwz	$acc08,`2048+128`($Tbl0)
+	lwz	$acc09,`2048+160`($Tbl0)
+	lwz	$acc10,`2048+192`($Tbl0)
+	lwz	$acc11,`2048+224`($Tbl0)
 	rlwinm	$acc00,$s0,`32-24`,24,31
-	lwz	$t1,4($key)
 	rlwinm	$acc01,$s1,`32-24`,24,31
-	lwz	$t2,8($key)
+	lwz	$t0,0($key)
+	lwz	$t1,4($key)
 	rlwinm	$acc02,$s2,`32-24`,24,31
-	lwz	$t3,12($key)
 	rlwinm	$acc03,$s3,`32-24`,24,31
-	lwz	$acc08,`2048+0`($Tbl0)	! prefetch Te4
+	lwz	$t2,8($key)
+	lwz	$t3,12($key)
 	rlwinm	$acc04,$s1,`32-16`,24,31
-	lwz	$acc09,`2048+32`($Tbl0)
 	rlwinm	$acc05,$s2,`32-16`,24,31
-	lwz	$acc10,`2048+64`($Tbl0)
+	lbzx	$acc00,$Tbl2,$acc00
+	lbzx	$acc01,$Tbl2,$acc01
 	rlwinm	$acc06,$s3,`32-16`,24,31
-	lwz	$acc11,`2048+96`($Tbl0)
 	rlwinm	$acc07,$s0,`32-16`,24,31
-	lwz	$acc12,`2048+128`($Tbl0)
+	lbzx	$acc02,$Tbl2,$acc02
+	lbzx	$acc03,$Tbl2,$acc03
 	rlwinm	$acc08,$s2,`32-8`,24,31
-	lwz	$acc13,`2048+160`($Tbl0)
 	rlwinm	$acc09,$s3,`32-8`,24,31
-	lwz	$acc14,`2048+192`($Tbl0)
+	lbzx	$acc04,$Tbl2,$acc04
+	lbzx	$acc05,$Tbl2,$acc05
 	rlwinm	$acc10,$s0,`32-8`,24,31
-	lwz	$acc15,`2048+224`($Tbl0)
 	rlwinm	$acc11,$s1,`32-8`,24,31
-	lbzx	$acc00,$Tbl2,$acc00
+	lbzx	$acc06,$Tbl2,$acc06
+	lbzx	$acc07,$Tbl2,$acc07
 	rlwinm	$acc12,$s3,`0`,24,31
-	lbzx	$acc01,$Tbl2,$acc01
 	rlwinm	$acc13,$s0,`0`,24,31
-	lbzx	$acc02,$Tbl2,$acc02
+	lbzx	$acc08,$Tbl2,$acc08
+	lbzx	$acc09,$Tbl2,$acc09
 	rlwinm	$acc14,$s1,`0`,24,31
-	lbzx	$acc03,$Tbl2,$acc03
 	rlwinm	$acc15,$s2,`0`,24,31
-	lbzx	$acc04,$Tbl2,$acc04
+	lbzx	$acc10,$Tbl2,$acc10
+	lbzx	$acc11,$Tbl2,$acc11
 	rlwinm	$s0,$acc00,24,0,7
-	lbzx	$acc05,$Tbl2,$acc05
 	rlwinm	$s1,$acc01,24,0,7
-	lbzx	$acc06,$Tbl2,$acc06
+	lbzx	$acc12,$Tbl2,$acc12
+	lbzx	$acc13,$Tbl2,$acc13
 	rlwinm	$s2,$acc02,24,0,7
-	lbzx	$acc07,$Tbl2,$acc07
 	rlwinm	$s3,$acc03,24,0,7
-	lbzx	$acc08,$Tbl2,$acc08
+	lbzx	$acc14,$Tbl2,$acc14
+	lbzx	$acc15,$Tbl2,$acc15
 	rlwimi	$s0,$acc04,16,8,15
-	lbzx	$acc09,$Tbl2,$acc09
 	rlwimi	$s1,$acc05,16,8,15
-	lbzx	$acc10,$Tbl2,$acc10
 	rlwimi	$s2,$acc06,16,8,15
-	lbzx	$acc11,$Tbl2,$acc11
 	rlwimi	$s3,$acc07,16,8,15
-	lbzx	$acc12,$Tbl2,$acc12
 	rlwimi	$s0,$acc08,8,16,23
-	lbzx	$acc13,$Tbl2,$acc13
 	rlwimi	$s1,$acc09,8,16,23
-	lbzx	$acc14,$Tbl2,$acc14
 	rlwimi	$s2,$acc10,8,16,23
-	lbzx	$acc15,$Tbl2,$acc15
 	rlwimi	$s3,$acc11,8,16,23
 	or	$s0,$s0,$acc12
 	or	$s1,$s1,$acc13
@@ -617,80 +517,78 @@ Lenc_loop:
 	xor	$s2,$s2,$t2
 	xor	$s3,$s3,$t3
 	blr
-	.long	0
-	.byte	0,12,0x14,0,0,0,0,0
 
 .align	4
 Lppc_AES_encrypt_compact:
 	lwz	$acc00,240($key)
-	addi	$Tbl1,$Tbl0,2048
 	lwz	$t0,0($key)
-	lis	$mask80,0x8080
 	lwz	$t1,4($key)
-	lis	$mask1b,0x1b1b
 	lwz	$t2,8($key)
-	ori	$mask80,$mask80,0x8080
 	lwz	$t3,12($key)
-	ori	$mask1b,$mask1b,0x1b1b
+	addi	$Tbl1,$Tbl0,2048
+	lis	$mask80,0x8080
+	lis	$mask1b,0x1b1b
 	addi	$key,$key,16
+	ori	$mask80,$mask80,0x8080
+	ori	$mask1b,$mask1b,0x1b1b
 	mtctr	$acc00
 .align	4
 Lenc_compact_loop:
 	xor	$s0,$s0,$t0
 	xor	$s1,$s1,$t1
-	rlwinm	$acc00,$s0,`32-24`,24,31
 	xor	$s2,$s2,$t2
-	rlwinm	$acc01,$s1,`32-24`,24,31
 	xor	$s3,$s3,$t3
+	rlwinm	$acc00,$s0,`32-24`,24,31
+	rlwinm	$acc01,$s1,`32-24`,24,31
 	rlwinm	$acc02,$s2,`32-24`,24,31
 	rlwinm	$acc03,$s3,`32-24`,24,31
+	lbzx	$acc00,$Tbl1,$acc00
+	lbzx	$acc01,$Tbl1,$acc01
 	rlwinm	$acc04,$s1,`32-16`,24,31
 	rlwinm	$acc05,$s2,`32-16`,24,31
+	lbzx	$acc02,$Tbl1,$acc02
+	lbzx	$acc03,$Tbl1,$acc03
 	rlwinm	$acc06,$s3,`32-16`,24,31
 	rlwinm	$acc07,$s0,`32-16`,24,31
-	lbzx	$acc00,$Tbl1,$acc00
+	lbzx	$acc04,$Tbl1,$acc04
+	lbzx	$acc05,$Tbl1,$acc05
 	rlwinm	$acc08,$s2,`32-8`,24,31
-	lbzx	$acc01,$Tbl1,$acc01
 	rlwinm	$acc09,$s3,`32-8`,24,31
-	lbzx	$acc02,$Tbl1,$acc02
+	lbzx	$acc06,$Tbl1,$acc06
+	lbzx	$acc07,$Tbl1,$acc07
 	rlwinm	$acc10,$s0,`32-8`,24,31
-	lbzx	$acc03,$Tbl1,$acc03
 	rlwinm	$acc11,$s1,`32-8`,24,31
-	lbzx	$acc04,$Tbl1,$acc04
+	lbzx	$acc08,$Tbl1,$acc08
+	lbzx	$acc09,$Tbl1,$acc09
 	rlwinm	$acc12,$s3,`0`,24,31
-	lbzx	$acc05,$Tbl1,$acc05
 	rlwinm	$acc13,$s0,`0`,24,31
-	lbzx	$acc06,$Tbl1,$acc06
+	lbzx	$acc10,$Tbl1,$acc10
+	lbzx	$acc11,$Tbl1,$acc11
 	rlwinm	$acc14,$s1,`0`,24,31
-	lbzx	$acc07,$Tbl1,$acc07
 	rlwinm	$acc15,$s2,`0`,24,31
-	lbzx	$acc08,$Tbl1,$acc08
+	lbzx	$acc12,$Tbl1,$acc12
+	lbzx	$acc13,$Tbl1,$acc13
 	rlwinm	$s0,$acc00,24,0,7
-	lbzx	$acc09,$Tbl1,$acc09
 	rlwinm	$s1,$acc01,24,0,7
-	lbzx	$acc10,$Tbl1,$acc10
+	lbzx	$acc14,$Tbl1,$acc14
+	lbzx	$acc15,$Tbl1,$acc15
 	rlwinm	$s2,$acc02,24,0,7
-	lbzx	$acc11,$Tbl1,$acc11
 	rlwinm	$s3,$acc03,24,0,7
-	lbzx	$acc12,$Tbl1,$acc12
 	rlwimi	$s0,$acc04,16,8,15
-	lbzx	$acc13,$Tbl1,$acc13
 	rlwimi	$s1,$acc05,16,8,15
-	lbzx	$acc14,$Tbl1,$acc14
 	rlwimi	$s2,$acc06,16,8,15
-	lbzx	$acc15,$Tbl1,$acc15
 	rlwimi	$s3,$acc07,16,8,15
 	rlwimi	$s0,$acc08,8,16,23
 	rlwimi	$s1,$acc09,8,16,23
 	rlwimi	$s2,$acc10,8,16,23
 	rlwimi	$s3,$acc11,8,16,23
 	lwz	$t0,0($key)
-	or	$s0,$s0,$acc12
 	lwz	$t1,4($key)
+	or	$s0,$s0,$acc12
 	or	$s1,$s1,$acc13
 	lwz	$t2,8($key)
-	or	$s2,$s2,$acc14
 	lwz	$t3,12($key)
+	or	$s2,$s2,$acc14
 	or	$s3,$s3,$acc15
 
 	addi	$key,$key,16
@@ -701,12 +599,12 @@ Lenc_compact_loop:
 	and	$acc02,$s2,$mask80
 	and	$acc03,$s3,$mask80
 	srwi	$acc04,$acc00,7		# r1>>7
-	andc	$acc08,$s0,$mask80	# r0&0x7f7f7f7f
 	srwi	$acc05,$acc01,7
-	andc	$acc09,$s1,$mask80
 	srwi	$acc06,$acc02,7
-	andc	$acc10,$s2,$mask80
 	srwi	$acc07,$acc03,7
+	andc	$acc08,$s0,$mask80	# r0&0x7f7f7f7f
+	andc	$acc09,$s1,$mask80
+	andc	$acc10,$s2,$mask80
 	andc	$acc11,$s3,$mask80
 	sub	$acc00,$acc00,$acc04	# r1-(r1>>7)
 	sub	$acc01,$acc01,$acc05
@@ -722,32 +620,32 @@ Lenc_compact_loop:
 	and	$acc03,$acc03,$mask1b
 	xor	$acc00,$acc00,$acc08	# r2
 	xor	$acc01,$acc01,$acc09
-	 rotlwi	$acc12,$s0,16		# ROTATE(r0,16)
 	xor	$acc02,$acc02,$acc10
-	 rotlwi	$acc13,$s1,16
 	xor	$acc03,$acc03,$acc11
-	 rotlwi	$acc14,$s2,16
 
-	xor	$s0,$s0,$acc00		# r0^r2
+	rotlwi	$acc12,$s0,16		# ROTATE(r0,16)
+	rotlwi	$acc13,$s1,16
+	rotlwi	$acc14,$s2,16
 	rotlwi	$acc15,$s3,16
+	xor	$s0,$s0,$acc00		# r0^r2
 	xor	$s1,$s1,$acc01
-	rotrwi	$s0,$s0,24		# ROTATE(r2^r0,24)
 	xor	$s2,$s2,$acc02
-	rotrwi	$s1,$s1,24
 	xor	$s3,$s3,$acc03
+	rotrwi	$s0,$s0,24		# ROTATE(r2^r0,24)
+	rotrwi	$s1,$s1,24
 	rotrwi	$s2,$s2,24
-	xor	$s0,$s0,$acc00		# ROTATE(r2^r0,24)^r2
 	rotrwi	$s3,$s3,24
+	xor	$s0,$s0,$acc00		# ROTATE(r2^r0,24)^r2
 	xor	$s1,$s1,$acc01
 	xor	$s2,$s2,$acc02
 	xor	$s3,$s3,$acc03
 	rotlwi	$acc08,$acc12,8		# ROTATE(r0,24)
-	xor	$s0,$s0,$acc12		#
 	rotlwi	$acc09,$acc13,8
-	xor	$s1,$s1,$acc13
 	rotlwi	$acc10,$acc14,8
-	xor	$s2,$s2,$acc14
 	rotlwi	$acc11,$acc15,8
+	xor	$s0,$s0,$acc12		#
+	xor	$s1,$s1,$acc13
+	xor	$s2,$s2,$acc14
 	xor	$s3,$s3,$acc15
 	xor	$s0,$s0,$acc08		#
 	xor	$s1,$s1,$acc09
@@ -762,15 +660,14 @@ Lenc_compact_done:
 	xor	$s2,$s2,$t2
 	xor	$s3,$s3,$t3
 	blr
-	.long	0
-	.byte	0,12,0x14,0,0,0,0,0
 
 .globl	.AES_decrypt
 .align	7
 .AES_decrypt:
-	$STU	$sp,-$FRAME($sp)
 	mflr	r0
+	$STU	$sp,-$FRAME($sp)
 
+	$PUSH	r0,`$FRAME-$SIZE_T*21`($sp)
 	$PUSH	$toc,`$FRAME-$SIZE_T*20`($sp)
 	$PUSH	r13,`$FRAME-$SIZE_T*19`($sp)
 	$PUSH	r14,`$FRAME-$SIZE_T*18`($sp)
@@ -791,14 +688,7 @@ Lenc_compact_done:
 	$PUSH	r29,`$FRAME-$SIZE_T*3`($sp)
 	$PUSH	r30,`$FRAME-$SIZE_T*2`($sp)
 	$PUSH	r31,`$FRAME-$SIZE_T*1`($sp)
-	$PUSH	r0,`$FRAME+$LRSAVE`($sp)
-
-	andi.	$t0,$inp,3
-	andi.	$t1,$out,3
-	or.	$t0,$t0,$t1
-	bne	Ldec_unaligned
 
-Ldec_unaligned_ok:
 	lwz	$s0,0($inp)
 	lwz	$s1,4($inp)
 	lwz	$s2,8($inp)
@@ -809,80 +699,8 @@ Ldec_unaligned_ok:
 	stw	$s1,4($out)
 	stw	$s2,8($out)
 	stw	$s3,12($out)
-	b	Ldec_done
-
-Ldec_unaligned:
-	subfic	$t0,$inp,4096
-	subfic	$t1,$out,4096
-	andi.	$t0,$t0,4096-16
-	beq	Ldec_xpage
-	andi.	$t1,$t1,4096-16
-	bne	Ldec_unaligned_ok
-
-Ldec_xpage:
-	lbz	$acc00,0($inp)
-	lbz	$acc01,1($inp)
-	lbz	$acc02,2($inp)
-	lbz	$s0,3($inp)
-	lbz	$acc04,4($inp)
-	lbz	$acc05,5($inp)
-	lbz	$acc06,6($inp)
-	lbz	$s1,7($inp)
-	lbz	$acc08,8($inp)
-	lbz	$acc09,9($inp)
-	lbz	$acc10,10($inp)
-	insrwi	$s0,$acc00,8,0
-	lbz	$s2,11($inp)
-	insrwi	$s1,$acc04,8,0
-	lbz	$acc12,12($inp)
-	insrwi	$s0,$acc01,8,8
-	lbz	$acc13,13($inp)
-	insrwi	$s1,$acc05,8,8
-	lbz	$acc14,14($inp)
-	insrwi	$s0,$acc02,8,16
-	lbz	$s3,15($inp)
-	insrwi	$s1,$acc06,8,16
-	insrwi	$s2,$acc08,8,0
-	insrwi	$s3,$acc12,8,0
-	insrwi	$s2,$acc09,8,8
-	insrwi	$s3,$acc13,8,8
-	insrwi	$s2,$acc10,8,16
-	insrwi	$s3,$acc14,8,16
-
-	bl	LAES_Td
-	bl	Lppc_AES_decrypt_compact
 
-	extrwi	$acc00,$s0,8,0
-	extrwi	$acc01,$s0,8,8
-	stb	$acc00,0($out)
-	extrwi	$acc02,$s0,8,16
-	stb	$acc01,1($out)
-	stb	$acc02,2($out)
-	extrwi	$acc04,$s1,8,0
-	stb	$s0,3($out)
-	extrwi	$acc05,$s1,8,8
-	stb	$acc04,4($out)
-	extrwi	$acc06,$s1,8,16
-	stb	$acc05,5($out)
-	stb	$acc06,6($out)
-	extrwi	$acc08,$s2,8,0
-	stb	$s1,7($out)
-	extrwi	$acc09,$s2,8,8
-	stb	$acc08,8($out)
-	extrwi	$acc10,$s2,8,16
-	stb	$acc09,9($out)
-	stb	$acc10,10($out)
-	extrwi	$acc12,$s3,8,0
-	stb	$s2,11($out)
-	extrwi	$acc13,$s3,8,8
-	stb	$acc12,12($out)
-	extrwi	$acc14,$s3,8,16
-	stb	$acc13,13($out)
-	stb	$acc14,14($out)
-	stb	$s3,15($out)
-
-Ldec_done:
-	$POP	r0,`$FRAME+$LRSAVE`($sp)
+	$POP	r0,`$FRAME-$SIZE_T*21`($sp)
 	$POP	$toc,`$FRAME-$SIZE_T*20`($sp)
 	$POP	r13,`$FRAME-$SIZE_T*19`($sp)
 	$POP	r14,`$FRAME-$SIZE_T*18`($sp)
@@ -906,21 +724,18 @@ Ldec_done:
 	mtlr	r0
 	addi	$sp,$sp,$FRAME
 	blr
-	.long	0
-	.byte	0,12,4,1,0x80,18,3,0
-	.long	0
 
-.align	5
+.align	4
 Lppc_AES_decrypt:
 	lwz	$acc00,240($key)
-	addi	$Tbl1,$Tbl0,3
 	lwz	$t0,0($key)
-	addi	$Tbl2,$Tbl0,2
 	lwz	$t1,4($key)
-	addi	$Tbl3,$Tbl0,1
 	lwz	$t2,8($key)
-	addi	$acc00,$acc00,-1
 	lwz	$t3,12($key)
+	addi	$Tbl1,$Tbl0,3
+	addi	$Tbl2,$Tbl0,2
+	addi	$Tbl3,$Tbl0,1
+	addi	$acc00,$acc00,-1
 	addi	$key,$key,16
 	xor	$s0,$s0,$t0
 	xor	$s1,$s1,$t1
@@ -931,47 +746,47 @@ Lppc_AES_decrypt:
 Ldec_loop:
 	rlwinm	$acc00,$s0,`32-24+3`,21,28
 	rlwinm	$acc01,$s1,`32-24+3`,21,28
+	lwz	$t0,0($key)
+	lwz	$t1,4($key)
 	rlwinm	$acc02,$s2,`32-24+3`,21,28
 	rlwinm	$acc03,$s3,`32-24+3`,21,28
-	lwz	$t0,0($key)
+	lwz	$t2,8($key)
+	lwz	$t3,12($key)
 	rlwinm	$acc04,$s3,`32-16+3`,21,28
-	lwz	$t1,4($key)
 	rlwinm	$acc05,$s0,`32-16+3`,21,28
-	lwz	$t2,8($key)
+	lwzx	$acc00,$Tbl0,$acc00
+	lwzx	$acc01,$Tbl0,$acc01
 	rlwinm	$acc06,$s1,`32-16+3`,21,28
-	lwz	$t3,12($key)
 	rlwinm	$acc07,$s2,`32-16+3`,21,28
-	lwzx	$acc00,$Tbl0,$acc00
+	lwzx	$acc02,$Tbl0,$acc02
+	lwzx	$acc03,$Tbl0,$acc03
 	rlwinm	$acc08,$s2,`32-8+3`,21,28
-	lwzx	$acc01,$Tbl0,$acc01
 	rlwinm	$acc09,$s3,`32-8+3`,21,28
-	lwzx	$acc02,$Tbl0,$acc02
+	lwzx	$acc04,$Tbl1,$acc04
+	lwzx	$acc05,$Tbl1,$acc05
 	rlwinm	$acc10,$s0,`32-8+3`,21,28
-	lwzx	$acc03,$Tbl0,$acc03
 	rlwinm	$acc11,$s1,`32-8+3`,21,28
-	lwzx	$acc04,$Tbl1,$acc04
+	lwzx	$acc06,$Tbl1,$acc06
+	lwzx	$acc07,$Tbl1,$acc07
 	rlwinm	$acc12,$s1,`0+3`,21,28
-	lwzx	$acc05,$Tbl1,$acc05
 	rlwinm	$acc13,$s2,`0+3`,21,28
-	lwzx	$acc06,$Tbl1,$acc06
+	lwzx	$acc08,$Tbl2,$acc08
+	lwzx	$acc09,$Tbl2,$acc09
 	rlwinm	$acc14,$s3,`0+3`,21,28
-	lwzx	$acc07,$Tbl1,$acc07
 	rlwinm	$acc15,$s0,`0+3`,21,28
-	lwzx	$acc08,$Tbl2,$acc08
+	lwzx	$acc10,$Tbl2,$acc10
+	lwzx	$acc11,$Tbl2,$acc11
 	xor	$t0,$t0,$acc00
-	lwzx	$acc09,$Tbl2,$acc09
 	xor	$t1,$t1,$acc01
-	lwzx	$acc10,$Tbl2,$acc10
+	lwzx	$acc12,$Tbl3,$acc12
+	lwzx	$acc13,$Tbl3,$acc13
 	xor	$t2,$t2,$acc02
-	lwzx	$acc11,$Tbl2,$acc11
 	xor	$t3,$t3,$acc03
-	lwzx	$acc12,$Tbl3,$acc12
+	lwzx	$acc14,$Tbl3,$acc14
+	lwzx	$acc15,$Tbl3,$acc15
 	xor	$t0,$t0,$acc04
-	lwzx	$acc13,$Tbl3,$acc13
 	xor	$t1,$t1,$acc05
-	lwzx	$acc14,$Tbl3,$acc14
 	xor	$t2,$t2,$acc06
-	lwzx	$acc15,$Tbl3,$acc15
 	xor	$t3,$t3,$acc07
 	xor	$t0,$t0,$acc08
 	xor	$t1,$t1,$acc09
@@ -986,57 +801,57 @@ Ldec_loop:
 
 	addi	$Tbl2,$Tbl0,2048
 	nop
-	lwz	$t0,0($key)
+	lwz	$acc08,`2048+0`($Tbl0)	! prefetch Td4
+	lwz	$acc09,`2048+32`($Tbl0)
+	lwz	$acc10,`2048+64`($Tbl0)
+	lwz	$acc11,`2048+96`($Tbl0)
+	lwz	$acc08,`2048+128`($Tbl0)
+	lwz	$acc09,`2048+160`($Tbl0)
+	lwz	$acc10,`2048+192`($Tbl0)
+	lwz	$acc11,`2048+224`($Tbl0)
 	rlwinm	$acc00,$s0,`32-24`,24,31
-	lwz	$t1,4($key)
 	rlwinm	$acc01,$s1,`32-24`,24,31
-	lwz	$t2,8($key)
+	lwz	$t0,0($key)
+	lwz	$t1,4($key)
 	rlwinm	$acc02,$s2,`32-24`,24,31
-	lwz	$t3,12($key)
 	rlwinm	$acc03,$s3,`32-24`,24,31
-	lwz	$acc08,`2048+0`($Tbl0)	! prefetch Td4
+	lwz	$t2,8($key)
+	lwz	$t3,12($key)
 	rlwinm	$acc04,$s3,`32-16`,24,31
-	lwz	$acc09,`2048+32`($Tbl0)
 	rlwinm	$acc05,$s0,`32-16`,24,31
-	lwz	$acc10,`2048+64`($Tbl0)
 	lbzx	$acc00,$Tbl2,$acc00
-	lwz	$acc11,`2048+96`($Tbl0)
 	lbzx	$acc01,$Tbl2,$acc01
-	lwz	$acc12,`2048+128`($Tbl0)
 	rlwinm	$acc06,$s1,`32-16`,24,31
-	lwz	$acc13,`2048+160`($Tbl0)
 	rlwinm	$acc07,$s2,`32-16`,24,31
-	lwz	$acc14,`2048+192`($Tbl0)
+	lbzx	$acc02,$Tbl2,$acc02
+	lbzx	$acc03,$Tbl2,$acc03
 	rlwinm	$acc08,$s2,`32-8`,24,31
-	lwz	$acc15,`2048+224`($Tbl0)
 	rlwinm	$acc09,$s3,`32-8`,24,31
-	lbzx	$acc02,$Tbl2,$acc02
+	lbzx	$acc04,$Tbl2,$acc04
+	lbzx	$acc05,$Tbl2,$acc05
 	rlwinm	$acc10,$s0,`32-8`,24,31
-	lbzx	$acc03,$Tbl2,$acc03
 	rlwinm	$acc11,$s1,`32-8`,24,31
-	lbzx	$acc04,$Tbl2,$acc04
+	lbzx	$acc06,$Tbl2,$acc06
+	lbzx	$acc07,$Tbl2,$acc07
 	rlwinm	$acc12,$s1,`0`,24,31
-	lbzx	$acc05,$Tbl2,$acc05
 	rlwinm	$acc13,$s2,`0`,24,31
-	lbzx	$acc06,$Tbl2,$acc06
+	lbzx	$acc08,$Tbl2,$acc08
+	lbzx	$acc09,$Tbl2,$acc09
 	rlwinm	$acc14,$s3,`0`,24,31
-	lbzx	$acc07,$Tbl2,$acc07
 	rlwinm	$acc15,$s0,`0`,24,31
-	lbzx	$acc08,$Tbl2,$acc08
+	lbzx	$acc10,$Tbl2,$acc10
+	lbzx	$acc11,$Tbl2,$acc11
 	rlwinm	$s0,$acc00,24,0,7
-	lbzx	$acc09,$Tbl2,$acc09
 	rlwinm	$s1,$acc01,24,0,7
-	lbzx	$acc10,$Tbl2,$acc10
+	lbzx	$acc12,$Tbl2,$acc12
+	lbzx	$acc13,$Tbl2,$acc13
 	rlwinm	$s2,$acc02,24,0,7
-	lbzx	$acc11,$Tbl2,$acc11
 	rlwinm	$s3,$acc03,24,0,7
-	lbzx	$acc12,$Tbl2,$acc12
+	lbzx	$acc14,$Tbl2,$acc14
+	lbzx	$acc15,$Tbl2,$acc15
 	rlwimi	$s0,$acc04,16,8,15
-	lbzx	$acc13,$Tbl2,$acc13
 	rlwimi	$s1,$acc05,16,8,15
-	lbzx	$acc14,$Tbl2,$acc14
 	rlwimi	$s2,$acc06,16,8,15
-	lbzx	$acc15,$Tbl2,$acc15
 	rlwimi	$s3,$acc07,16,8,15
 	rlwimi	$s0,$acc08,8,16,23
 	rlwimi	$s1,$acc09,8,16,23
@@ -1051,22 +866,20 @@ Ldec_loop:
 	xor	$s2,$s2,$t2
 	xor	$s3,$s3,$t3
 	blr
-	.long	0
-	.byte	0,12,0x14,0,0,0,0,0
 
 .align	4
 Lppc_AES_decrypt_compact:
 	lwz	$acc00,240($key)
-	addi	$Tbl1,$Tbl0,2048
 	lwz	$t0,0($key)
-	lis	$mask80,0x8080
 	lwz	$t1,4($key)
-	lis	$mask1b,0x1b1b
 	lwz	$t2,8($key)
-	ori	$mask80,$mask80,0x8080
 	lwz	$t3,12($key)
-	ori	$mask1b,$mask1b,0x1b1b
+	addi	$Tbl1,$Tbl0,2048
+	lis	$mask80,0x8080
+	lis	$mask1b,0x1b1b
 	addi	$key,$key,16
+	ori	$mask80,$mask80,0x8080
+	ori	$mask1b,$mask1b,0x1b1b
 ___
 $code.=<<___ if ($SIZE_T==8);
 	insrdi	$mask80,$mask80,32,0
@@ -1078,59 +891,59 @@ $code.=<<___;
 Ldec_compact_loop:
 	xor	$s0,$s0,$t0
 	xor	$s1,$s1,$t1
-	rlwinm	$acc00,$s0,`32-24`,24,31
 	xor	$s2,$s2,$t2
-	rlwinm	$acc01,$s1,`32-24`,24,31
 	xor	$s3,$s3,$t3
+	rlwinm	$acc00,$s0,`32-24`,24,31
+	rlwinm	$acc01,$s1,`32-24`,24,31
 	rlwinm	$acc02,$s2,`32-24`,24,31
 	rlwinm	$acc03,$s3,`32-24`,24,31
+	lbzx	$acc00,$Tbl1,$acc00
+	lbzx	$acc01,$Tbl1,$acc01
 	rlwinm	$acc04,$s3,`32-16`,24,31
 	rlwinm	$acc05,$s0,`32-16`,24,31
+	lbzx	$acc02,$Tbl1,$acc02
+	lbzx	$acc03,$Tbl1,$acc03
 	rlwinm	$acc06,$s1,`32-16`,24,31
 	rlwinm	$acc07,$s2,`32-16`,24,31
-	lbzx	$acc00,$Tbl1,$acc00
+	lbzx	$acc04,$Tbl1,$acc04
+	lbzx	$acc05,$Tbl1,$acc05
 	rlwinm	$acc08,$s2,`32-8`,24,31
-	lbzx	$acc01,$Tbl1,$acc01
 	rlwinm	$acc09,$s3,`32-8`,24,31
-	lbzx	$acc02,$Tbl1,$acc02
+	lbzx	$acc06,$Tbl1,$acc06
+	lbzx	$acc07,$Tbl1,$acc07
 	rlwinm	$acc10,$s0,`32-8`,24,31
-	lbzx	$acc03,$Tbl1,$acc03
 	rlwinm	$acc11,$s1,`32-8`,24,31
-	lbzx	$acc04,$Tbl1,$acc04
+	lbzx	$acc08,$Tbl1,$acc08
+	lbzx	$acc09,$Tbl1,$acc09
 	rlwinm	$acc12,$s1,`0`,24,31
-	lbzx	$acc05,$Tbl1,$acc05
 	rlwinm	$acc13,$s2,`0`,24,31
-	lbzx	$acc06,$Tbl1,$acc06
+	lbzx	$acc10,$Tbl1,$acc10
+	lbzx	$acc11,$Tbl1,$acc11
 	rlwinm	$acc14,$s3,`0`,24,31
-	lbzx	$acc07,$Tbl1,$acc07
 	rlwinm	$acc15,$s0,`0`,24,31
-	lbzx	$acc08,$Tbl1,$acc08
+	lbzx	$acc12,$Tbl1,$acc12
+	lbzx	$acc13,$Tbl1,$acc13
 	rlwinm	$s0,$acc00,24,0,7
-	lbzx	$acc09,$Tbl1,$acc09
 	rlwinm	$s1,$acc01,24,0,7
-	lbzx	$acc10,$Tbl1,$acc10
+	lbzx	$acc14,$Tbl1,$acc14
+	lbzx	$acc15,$Tbl1,$acc15
 	rlwinm	$s2,$acc02,24,0,7
-	lbzx	$acc11,$Tbl1,$acc11
 	rlwinm	$s3,$acc03,24,0,7
-	lbzx	$acc12,$Tbl1,$acc12
 	rlwimi	$s0,$acc04,16,8,15
-	lbzx	$acc13,$Tbl1,$acc13
 	rlwimi	$s1,$acc05,16,8,15
-	lbzx	$acc14,$Tbl1,$acc14
 	rlwimi	$s2,$acc06,16,8,15
-	lbzx	$acc15,$Tbl1,$acc15
 	rlwimi	$s3,$acc07,16,8,15
 	rlwimi	$s0,$acc08,8,16,23
 	rlwimi	$s1,$acc09,8,16,23
 	rlwimi	$s2,$acc10,8,16,23
 	rlwimi	$s3,$acc11,8,16,23
 	lwz	$t0,0($key)
-	or	$s0,$s0,$acc12
 	lwz	$t1,4($key)
+	or	$s0,$s0,$acc12
 	or	$s1,$s1,$acc13
 	lwz	$t2,8($key)
-	or	$s2,$s2,$acc14
 	lwz	$t3,12($key)
+	or	$s2,$s2,$acc14
 	or	$s3,$s3,$acc15
 
 	addi	$key,$key,16
@@ -1204,12 +1017,12 @@ $code.=<<___ if ($SIZE_T==4);
 	and	$acc02,$s2,$mask80
 	and	$acc03,$s3,$mask80
 	srwi	$acc04,$acc00,7		# r1>>7
-	andc	$acc08,$s0,$mask80	# r0&0x7f7f7f7f
 	srwi	$acc05,$acc01,7
-	andc	$acc09,$s1,$mask80
 	srwi	$acc06,$acc02,7
-	andc	$acc10,$s2,$mask80
 	srwi	$acc07,$acc03,7
+	andc	$acc08,$s0,$mask80	# r0&0x7f7f7f7f
+	andc	$acc09,$s1,$mask80
+	andc	$acc10,$s2,$mask80
 	andc	$acc11,$s3,$mask80
 	sub	$acc00,$acc00,$acc04	# r1-(r1>>7)
 	sub	$acc01,$acc01,$acc05
@@ -1233,12 +1046,12 @@ $code.=<<___ if ($SIZE_T==4);
 	and	$acc06,$acc02,$mask80
 	and	$acc07,$acc03,$mask80
 	srwi	$acc08,$acc04,7		# r1>>7
-	andc	$acc12,$acc00,$mask80	# r2&0x7f7f7f7f
 	srwi	$acc09,$acc05,7
-	andc	$acc13,$acc01,$mask80
 	srwi	$acc10,$acc06,7
-	andc	$acc14,$acc02,$mask80
 	srwi	$acc11,$acc07,7
+	andc	$acc12,$acc00,$mask80	# r2&0x7f7f7f7f
+	andc	$acc13,$acc01,$mask80
+	andc	$acc14,$acc02,$mask80
 	andc	$acc15,$acc03,$mask80
 	sub	$acc04,$acc04,$acc08	# r1-(r1>>7)
 	sub	$acc05,$acc05,$acc09
@@ -1259,13 +1072,13 @@ $code.=<<___ if ($SIZE_T==4);
 
 	and	$acc08,$acc04,$mask80	# r1=r4&0x80808080
 	and	$acc09,$acc05,$mask80
-	srwi	$acc12,$acc08,7		# r1>>7
 	and	$acc10,$acc06,$mask80
-	srwi	$acc13,$acc09,7
 	and	$acc11,$acc07,$mask80
+	srwi	$acc12,$acc08,7		# r1>>7
+	srwi	$acc13,$acc09,7
 	srwi	$acc14,$acc10,7
-	sub	$acc08,$acc08,$acc12	# r1-(r1>>7)
 	srwi	$acc15,$acc11,7
+	sub	$acc08,$acc08,$acc12	# r1-(r1>>7)
 	sub	$acc09,$acc09,$acc13
 	sub	$acc10,$acc10,$acc14
 	sub	$acc11,$acc11,$acc15
@@ -1298,10 +1111,10 @@ ___
 $code.=<<___;
 	rotrwi	$s0,$s0,8		# = ROTATE(r0,8)
 	rotrwi	$s1,$s1,8
-	xor	$s0,$s0,$acc00		# ^= r2^r0
 	rotrwi	$s2,$s2,8
-	xor	$s1,$s1,$acc01
 	rotrwi	$s3,$s3,8
+	xor	$s0,$s0,$acc00		# ^= r2^r0
+	xor	$s1,$s1,$acc01
 	xor	$s2,$s2,$acc02
 	xor	$s3,$s3,$acc03
 	xor	$acc00,$acc00,$acc08
@@ -1309,32 +1122,32 @@ $code.=<<___;
 	xor	$acc02,$acc02,$acc10
 	xor	$acc03,$acc03,$acc11
 	xor	$s0,$s0,$acc04		# ^= r4^r0
-	rotrwi	$acc00,$acc00,24
 	xor	$s1,$s1,$acc05
-	rotrwi	$acc01,$acc01,24
 	xor	$s2,$s2,$acc06
-	rotrwi	$acc02,$acc02,24
 	xor	$s3,$s3,$acc07
+	rotrwi	$acc00,$acc00,24
+	rotrwi	$acc01,$acc01,24
+	rotrwi	$acc02,$acc02,24
 	rotrwi	$acc03,$acc03,24
 	xor	$acc04,$acc04,$acc08
 	xor	$acc05,$acc05,$acc09
 	xor	$acc06,$acc06,$acc10
 	xor	$acc07,$acc07,$acc11
 	xor	$s0,$s0,$acc08		# ^= r8 [^((r4^r0)^(r2^r0)=r4^r2)]
-	rotrwi	$acc04,$acc04,16
 	xor	$s1,$s1,$acc09
-	rotrwi	$acc05,$acc05,16
 	xor	$s2,$s2,$acc10
-	rotrwi	$acc06,$acc06,16
 	xor	$s3,$s3,$acc11
+	rotrwi	$acc04,$acc04,16
+	rotrwi	$acc05,$acc05,16
+	rotrwi	$acc06,$acc06,16
 	rotrwi	$acc07,$acc07,16
 	xor	$s0,$s0,$acc00		# ^= ROTATE(r8^r2^r0,24)
-	rotrwi	$acc08,$acc08,8
 	xor	$s1,$s1,$acc01
-	rotrwi	$acc09,$acc09,8
 	xor	$s2,$s2,$acc02
-	rotrwi	$acc10,$acc10,8
 	xor	$s3,$s3,$acc03
+	rotrwi	$acc08,$acc08,8
+	rotrwi	$acc09,$acc09,8
+	rotrwi	$acc10,$acc10,8
 	rotrwi	$acc11,$acc11,8
 	xor	$s0,$s0,$acc04		# ^= ROTATE(r8^r4^r0,16)
 	xor	$s1,$s1,$acc05
@@ -1353,9 +1166,7 @@ Ldec_compact_done:
 	xor	$s2,$s2,$t2
 	xor	$s3,$s3,$t3
 	blr
-	.long	0
-	.byte	0,12,0x14,0,0,0,0,0
-
+.long	0
 .asciz	"AES for PPC, CRYPTOGAMS by <appro\@openssl.org>"
 .align	7
 ___
diff --git a/lib/libcrypto/bn/asm/ppc64-mont.pl b/lib/libcrypto/bn/asm/ppc64-mont.pl
index a14e769ad05..3449b35855d 100644
--- a/lib/libcrypto/bn/asm/ppc64-mont.pl
+++ b/lib/libcrypto/bn/asm/ppc64-mont.pl
@@ -45,40 +45,23 @@
 # on 1.8GHz PPC970, it's only 5-55% faster. Still far from impressive
 # in absolute terms, but it's apparently the way Power 6 is...
 
-# December 2009
-
-# Adapted for 32-bit build this module delivers 25-120%, yes, more
-# than *twice* for longer keys, performance improvement over 32-bit
-# ppc-mont.pl on 1.8GHz PPC970. However! This implementation utilizes
-# even 64-bit integer operations and the trouble is that most PPC
-# operating systems don't preserve upper halves of general purpose
-# registers upon 32-bit signal delivery. They do preserve them upon
-# context switch, but not signalling:-( This means that asynchronous
-# signals have to be blocked upon entry to this subroutine. Signal
-# masking (and of course complementary unmasking) has quite an impact
-# on performance, naturally larger for shorter keys. It's so severe
-# that 512-bit key performance can be as low as 1/3 of expected one.
-# This is why this routine can be engaged for longer key operations
-# only on these OSes, see crypto/ppccap.c for further details. MacOS X
-# is an exception from this and doesn't require signal masking, and
-# that's where above improvement coefficients were collected. For
-# others alternative would be to break dependence on upper halves of
-# GPRs by sticking to 32-bit integer operations...
-
 $flavour = shift;
 
 if ($flavour =~ /32/) {
 	$SIZE_T=4;
 	$RZONE=	224;
-	$fname=	"bn_mul_mont_fpu64";
+	$FRAME=	$SIZE_T*12+8*12;
+	$fname=	"bn_mul_mont_ppc64";
 
 	$STUX=	"stwux";	# store indexed and update
 	$PUSH=	"stw";
 	$POP=	"lwz";
+	die "not implemented yet";
 } elsif ($flavour =~ /64/) {
 	$SIZE_T=8;
 	$RZONE=	288;
-	$fname=	"bn_mul_mont_fpu64";
+	$FRAME=	$SIZE_T*12+8*12;
+	$fname=	"bn_mul_mont";
 
 	# same as above, but 64-bit mnemonics...
 	$STUX=	"stdux";	# store indexed and update
@@ -93,7 +76,7 @@ die "can't locate ppc-xlate.pl";
 
 open STDOUT,"| $^X $xlate $flavour ".shift || die "can't call $xlate: $!";
 
-$FRAME=64;	# padded frame header
+$FRAME=($FRAME+63)&~63;
 $TRANSFER=16*8;
 
 $carry="r0";
@@ -110,16 +93,16 @@ $tp="r10";
 $j="r11";
 $i="r12";
 # non-volatile registers
-$nap_d="r22";	# interleaved ap and np in double format
-$a0="r23";	# ap[0]
-$t0="r24";	# temporary registers
-$t1="r25";
-$t2="r26";
-$t3="r27";
-$t4="r28";
-$t5="r29";
-$t6="r30";
-$t7="r31";
+$nap_d="r14";	# interleaved ap and np in double format
+$a0="r15";	# ap[0]
+$t0="r16";	# temporary registers
+$t1="r17";
+$t2="r18";
+$t3="r19";
+$t4="r20";
+$t5="r21";
+$t6="r22";
+$t7="r23";
 
 # PPC offers enough register bank capacity to unroll inner loops twice
 #
@@ -149,17 +132,28 @@ $ba="f0";	$bb="f1";	$bc="f2";	$bd="f3";
 $na="f4";	$nb="f5";	$nc="f6";	$nd="f7";
 $dota="f8";	$dotb="f9";
 $A0="f10";	$A1="f11";	$A2="f12";	$A3="f13";
-$N0="f20";	$N1="f21";	$N2="f22";	$N3="f23";
-$T0a="f24";	$T0b="f25";
-$T1a="f26";	$T1b="f27";
-$T2a="f28";	$T2b="f29";
-$T3a="f30";	$T3b="f31";
+$N0="f14";	$N1="f15";	$N2="f16";	$N3="f17";
+$T0a="f18";	$T0b="f19";
+$T1a="f20";	$T1b="f21";
+$T2a="f22";	$T2b="f23";
+$T3a="f24";	$T3b="f25";
 
 # sp----------->+-------------------------------+
 #		| saved sp			|
 #		+-------------------------------+
+#		|				|
+#		+-------------------------------+
+#		| 10 saved gpr, r14-r23		|
+#		.				.
+#		.				.
+#   +12*size_t	+-------------------------------+
+#		| 12 saved fpr, f14-f25		|
 #		.				.
-#   +64		+-------------------------------+
+#		.				.
+#   +12*8	+-------------------------------+
+#		| padding to 64 byte boundary	|
+#		.				.
+#   +X		+-------------------------------+
 #		| 16 gpr<->fpr transfer zone	|
 #		.				.
 #		.				.
@@ -179,16 +173,6 @@ $T3a="f30";	$T3b="f31";
 #		.				.
 #		.				.
 #		+-------------------------------+
-#		.				.
-#   -12*size_t	+-------------------------------+
-#		| 10 saved gpr, r22-r31		|
-#		.				.
-#		.				.
-#   -12*8	+-------------------------------+
-#		| 12 saved fpr, f20-f31		|
-#		.				.
-#		.				.
-#		+-------------------------------+
 
 $code=<<___;
 .machine "any"
@@ -197,14 +181,14 @@ $code=<<___;
 .globl	.$fname
 .align	5
 .$fname:
-	cmpwi	$num,`3*8/$SIZE_T`
+	cmpwi	$num,4
 	mr	$rp,r3		; $rp is reassigned
 	li	r3,0		; possible "not handled" return code
 	bltlr-
-	andi.	r0,$num,`16/$SIZE_T-1`		; $num has to be "even"
+	andi.	r0,$num,1	; $num has to be even
 	bnelr-
 
-	slwi	$num,$num,`log($SIZE_T)/log(2)`	; num*=sizeof(BN_LONG)
+	slwi	$num,$num,3	; num*=8
 	li	$i,-4096
 	slwi	$tp,$num,2	; place for {an}p_{lh}[num], i.e. 4*num
 	add	$tp,$tp,$num	; place for tp[num+1]
@@ -212,50 +196,35 @@ $code=<<___;
 	subf	$tp,$tp,$sp	; $sp-$tp
 	and	$tp,$tp,$i	; minimize TLB usage
 	subf	$tp,$sp,$tp	; $tp-$sp
-	mr	$i,$sp
 	$STUX	$sp,$sp,$tp	; alloca
 
-	$PUSH	r22,`-12*8-10*$SIZE_T`($i)
-	$PUSH	r23,`-12*8-9*$SIZE_T`($i)
-	$PUSH	r24,`-12*8-8*$SIZE_T`($i)
-	$PUSH	r25,`-12*8-7*$SIZE_T`($i)
-	$PUSH	r26,`-12*8-6*$SIZE_T`($i)
-	$PUSH	r27,`-12*8-5*$SIZE_T`($i)
-	$PUSH	r28,`-12*8-4*$SIZE_T`($i)
-	$PUSH	r29,`-12*8-3*$SIZE_T`($i)
-	$PUSH	r30,`-12*8-2*$SIZE_T`($i)
-	$PUSH	r31,`-12*8-1*$SIZE_T`($i)
-	stfd	f20,`-12*8`($i)
-	stfd	f21,`-11*8`($i)
-	stfd	f22,`-10*8`($i)
-	stfd	f23,`-9*8`($i)
-	stfd	f24,`-8*8`($i)
-	stfd	f25,`-7*8`($i)
-	stfd	f26,`-6*8`($i)
-	stfd	f27,`-5*8`($i)
-	stfd	f28,`-4*8`($i)
-	stfd	f29,`-3*8`($i)
-	stfd	f30,`-2*8`($i)
-	stfd	f31,`-1*8`($i)
-___
-$code.=<<___ if ($SIZE_T==8);
+	$PUSH	r14,`2*$SIZE_T`($sp)
+	$PUSH	r15,`3*$SIZE_T`($sp)
+	$PUSH	r16,`4*$SIZE_T`($sp)
+	$PUSH	r17,`5*$SIZE_T`($sp)
+	$PUSH	r18,`6*$SIZE_T`($sp)
+	$PUSH	r19,`7*$SIZE_T`($sp)
+	$PUSH	r20,`8*$SIZE_T`($sp)
+	$PUSH	r21,`9*$SIZE_T`($sp)
+	$PUSH	r22,`10*$SIZE_T`($sp)
+	$PUSH	r23,`11*$SIZE_T`($sp)
+	stfd	f14,`12*$SIZE_T+0`($sp)
+	stfd	f15,`12*$SIZE_T+8`($sp)
+	stfd	f16,`12*$SIZE_T+16`($sp)
+	stfd	f17,`12*$SIZE_T+24`($sp)
+	stfd	f18,`12*$SIZE_T+32`($sp)
+	stfd	f19,`12*$SIZE_T+40`($sp)
+	stfd	f20,`12*$SIZE_T+48`($sp)
+	stfd	f21,`12*$SIZE_T+56`($sp)
+	stfd	f22,`12*$SIZE_T+64`($sp)
+	stfd	f23,`12*$SIZE_T+72`($sp)
+	stfd	f24,`12*$SIZE_T+80`($sp)
+	stfd	f25,`12*$SIZE_T+88`($sp)
+
 	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
@@ -289,8 +258,6 @@ $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
@@ -299,18 +266,6 @@ $code.=<<___ if ($SIZE_T==8);
 	lwz	$t5,0($np)
 	lwz	$t6,12($np)		; load n[j+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	$t3,12($ap)
-	lwz	$t4,0($np)		; load n[j..j+3] as 32-bit word pairs
-	lwz	$t5,4($np)
-	lwz	$t6,8($np)
-	lwz	$t7,12($np)
-___
-$code.=<<___;
 	lfd	$ba,`$FRAME+0`($sp)
 	lfd	$bb,`$FRAME+8`($sp)
 	lfd	$bc,`$FRAME+16`($sp)
@@ -419,8 +374,6 @@ $code.=<<___;
 
 .align	5
 L1st:
-___
-$code.=<<___ if ($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
@@ -429,18 +382,6 @@ $code.=<<___ if ($SIZE_T==8);
 	lwz	$t5,0($np)
 	lwz	$t6,12($np)		; load n[j+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	$t3,12($ap)
-	lwz	$t4,0($np)		; load n[j..j+3] as 32-bit word pairs
-	lwz	$t5,4($np)
-	lwz	$t6,8($np)
-	lwz	$t7,12($np)
-___
-$code.=<<___;
 	std	$t0,`$FRAME+64`($sp)
 	std	$t1,`$FRAME+72`($sp)
 	std	$t2,`$FRAME+80`($sp)
@@ -618,17 +559,7 @@ $code.=<<___;
 	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]
 
@@ -830,13 +761,6 @@ Linner:
 	stfd	$T0b,`$FRAME+8`($sp)
 	 add	$t7,$t7,$carry
 	 addc	$t3,$t0,$t1
-___
-$code.=<<___ if ($SIZE_T==4);		# adjust XER[CA]
-	extrdi	$t0,$t0,32,0
-	extrdi	$t1,$t1,32,0
-	adde	$t0,$t0,$t1
-___
-$code.=<<___;
 	stfd	$T1a,`$FRAME+16`($sp)
 	stfd	$T1b,`$FRAME+24`($sp)
 	 insrdi	$t4,$t7,16,0		; 64..127 bits
@@ -844,13 +768,6 @@ $code.=<<___;
 	stfd	$T2a,`$FRAME+32`($sp)
 	stfd	$T2b,`$FRAME+40`($sp)
 	 adde	$t5,$t4,$t2
-___
-$code.=<<___ if ($SIZE_T==4);		# adjust XER[CA]
-	extrdi	$t4,$t4,32,0
-	extrdi	$t2,$t2,32,0
-	adde	$t4,$t4,$t2
-___
-$code.=<<___;
 	stfd	$T3a,`$FRAME+48`($sp)
 	stfd	$T3b,`$FRAME+56`($sp)
 	 addze	$carry,$carry
@@ -899,21 +816,7 @@ $code.=<<___;
 	ld	$t7,`$FRAME+72`($sp)
 
 	addc	$t3,$t0,$t1
-___
-$code.=<<___ if ($SIZE_T==4);		# adjust XER[CA]
-	extrdi	$t0,$t0,32,0
-	extrdi	$t1,$t1,32,0
-	adde	$t0,$t0,$t1
-___
-$code.=<<___;
 	adde	$t5,$t4,$t2
-___
-$code.=<<___ if ($SIZE_T==4);		# adjust XER[CA]
-	extrdi	$t4,$t4,32,0
-	extrdi	$t2,$t2,32,0
-	adde	$t4,$t4,$t2
-___
-$code.=<<___;
 	addze	$carry,$carry
 
 	std	$t3,-16($tp)		; tp[j-1]
@@ -932,9 +835,7 @@ $code.=<<___;
 	subf	$nap_d,$t7,$nap_d	; rewind pointer
 	cmpw	$i,$num
 	blt-	Louter
-___
 
-$code.=<<___ if ($SIZE_T==8);
 	subf	$np,$num,$np	; rewind np
 	addi	$j,$j,1		; restore counter
 	subfc	$i,$i,$i	; j=0 and "clear" XER[CA]
@@ -982,105 +883,34 @@ Lcopy:				; copy or in-place refresh
 	stdx	$i,$t4,$i
 	addi	$i,$i,16
 	bdnz-	Lcopy
-___
-$code.=<<___ if ($SIZE_T==4);
-	subf	$np,$num,$np	; rewind np
-	addi	$j,$j,1		; restore counter
-	subfc	$i,$i,$i	; j=0 and "clear" XER[CA]
-	addi	$tp,$sp,`$FRAME+$TRANSFER`
-	addi	$np,$np,-4
-	addi	$rp,$rp,-4
-	addi	$ap,$sp,`$FRAME+$TRANSFER+4`
-	mtctr	$j
-
-.align	4
-Lsub:	ld	$t0,8($tp)	; load tp[j..j+3] in 64-bit word order
-	ldu	$t2,16($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]
-	 stw	$t1,8($ap)
-	subfe	$t6,$t6,$t2	; tp[j+2]-np[j+2]
-	 stw	$t2,12($ap)
-	subfe	$t7,$t7,$t3	; tp[j+3]-np[j+3]
-	 stwu	$t3,16($ap)
-	stw	$t4,4($rp)
-	stw	$t5,8($rp)
-	stw	$t6,12($rp)
-	stwu	$t7,16($rp)
-	bdnz-	Lsub
-
-	li	$i,0
-	subfe	$ovf,$i,$ovf	; handle upmost overflow bit
-	addi	$tp,$sp,`$FRAME+$TRANSFER+4`
-	subf	$rp,$num,$rp	; rewind rp
-	and	$ap,$tp,$ovf
-	andc	$np,$rp,$ovf
-	or	$ap,$ap,$np	; ap=borrow?tp:rp
-	addi	$tp,$sp,`$FRAME+$TRANSFER`
-	mtctr	$j
-
-.align	4
-Lcopy:				; copy or in-place refresh
-	lwz	$t0,4($ap)
-	lwz	$t1,8($ap)
-	lwz	$t2,12($ap)
-	lwzu	$t3,16($ap)
-	std	$i,8($nap_d)	; zap nap_d
-	std	$i,16($nap_d)
-	std	$i,24($nap_d)
-	std	$i,32($nap_d)
-	std	$i,40($nap_d)
-	std	$i,48($nap_d)
-	std	$i,56($nap_d)
-	stdu	$i,64($nap_d)
-	stw	$t0,4($rp)
-	stw	$t1,8($rp)
-	stw	$t2,12($rp)
-	stwu	$t3,16($rp)
-	std	$i,8($tp)	; zap tp at once
-	stdu	$i,16($tp)
-	bdnz-	Lcopy
-___
 
-$code.=<<___;
-	$POP	$i,0($sp)
+	$POP	r14,`2*$SIZE_T`($sp)
+	$POP	r15,`3*$SIZE_T`($sp)
+	$POP	r16,`4*$SIZE_T`($sp)
+	$POP	r17,`5*$SIZE_T`($sp)
+	$POP	r18,`6*$SIZE_T`($sp)
+	$POP	r19,`7*$SIZE_T`($sp)
+	$POP	r20,`8*$SIZE_T`($sp)
+	$POP	r21,`9*$SIZE_T`($sp)
+	$POP	r22,`10*$SIZE_T`($sp)
+	$POP	r23,`11*$SIZE_T`($sp)
+	lfd	f14,`12*$SIZE_T+0`($sp)
+	lfd	f15,`12*$SIZE_T+8`($sp)
+	lfd	f16,`12*$SIZE_T+16`($sp)
+	lfd	f17,`12*$SIZE_T+24`($sp)
+	lfd	f18,`12*$SIZE_T+32`($sp)
+	lfd	f19,`12*$SIZE_T+40`($sp)
+	lfd	f20,`12*$SIZE_T+48`($sp)
+	lfd	f21,`12*$SIZE_T+56`($sp)
+	lfd	f22,`12*$SIZE_T+64`($sp)
+	lfd	f23,`12*$SIZE_T+72`($sp)
+	lfd	f24,`12*$SIZE_T+80`($sp)
+	lfd	f25,`12*$SIZE_T+88`($sp)
+	$POP	$sp,0($sp)
 	li	r3,1	; signal "handled"
-	$POP	r22,`-12*8-10*$SIZE_T`($i)
-	$POP	r23,`-12*8-9*$SIZE_T`($i)
-	$POP	r24,`-12*8-8*$SIZE_T`($i)
-	$POP	r25,`-12*8-7*$SIZE_T`($i)
-	$POP	r26,`-12*8-6*$SIZE_T`($i)
-	$POP	r27,`-12*8-5*$SIZE_T`($i)
-	$POP	r28,`-12*8-4*$SIZE_T`($i)
-	$POP	r29,`-12*8-3*$SIZE_T`($i)
-	$POP	r30,`-12*8-2*$SIZE_T`($i)
-	$POP	r31,`-12*8-1*$SIZE_T`($i)
-	lfd	f20,`-12*8`($i)
-	lfd	f21,`-11*8`($i)
-	lfd	f22,`-10*8`($i)
-	lfd	f23,`-9*8`($i)
-	lfd	f24,`-8*8`($i)
-	lfd	f25,`-7*8`($i)
-	lfd	f26,`-6*8`($i)
-	lfd	f27,`-5*8`($i)
-	lfd	f28,`-4*8`($i)
-	lfd	f29,`-3*8`($i)
-	lfd	f30,`-2*8`($i)
-	lfd	f31,`-1*8`($i)
-	mr	$sp,$i
 	blr
 	.long	0
-	.byte	0,12,4,0,0x8c,10,6,0
-	.long	0
-
-.asciz  "Montgomery Multiplication for PPC64, CRYPTOGAMS by <appro\@openssl.org>"
+.asciz  "Montgomery Multiplication for PPC64, CRYPTOGAMS by <appro\@fy.chalmers.se>"
 ___
 
 $code =~ s/\`([^\`]*)\`/eval $1/gem;