Initial import

8 files changed, 12623 insertions, 0 deletions

diff --git a/openssl/crypto/aes/asm/aes-586.pl b/openssl/crypto/aes/asm/aes-586.pl
new file mode 100755
index 00000000..aab40e6f
--- /dev/null
+++ b/openssl/crypto/aes/asm/aes-586.pl
@@ -0,0 +1,2980 @@
+#!/usr/bin/env perl
+#
+# ====================================================================
+# Written by Andy Polyakov <appro@fy.chalmers.se> 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/.
+# ====================================================================
+#
+# Version 4.3.
+#
+# You might fail to appreciate this module performance from the first
+# try. If compared to "vanilla" linux-ia32-icc target, i.e. considered
+# to be *the* best Intel C compiler without -KPIC, performance appears
+# to be virtually identical... But try to re-configure with shared
+# library support... Aha! Intel compiler "suddenly" lags behind by 30%
+# [on P4, more on others]:-) And if compared to position-independent
+# code generated by GNU C, this code performs *more* than *twice* as
+# fast! Yes, all this buzz about PIC means that unlike other hand-
+# coded implementations, this one was explicitly designed to be safe
+# to use even in shared library context... This also means that this
+# code isn't necessarily absolutely fastest "ever," because in order
+# to achieve position independence an extra register has to be
+# off-loaded to stack, which affects the benchmark result.
+#
+# Special note about instruction choice. Do you recall RC4_INT code
+# performing poorly on P4? It might be the time to figure out why.
+# RC4_INT code implies effective address calculations in base+offset*4
+# form. Trouble is that it seems that offset scaling turned to be
+# critical path... At least eliminating scaling resulted in 2.8x RC4
+# performance improvement [as you might recall]. As AES code is hungry
+# for scaling too, I [try to] avoid the latter by favoring off-by-2
+# shifts and masking the result with 0xFF<<2 instead of "boring" 0xFF.
+#
+# As was shown by Dean Gaudet <dean@arctic.org>, the above note turned
+# void. Performance improvement with off-by-2 shifts was observed on
+# intermediate implementation, which was spilling yet another register
+# to stack... Final offset*4 code below runs just a tad faster on P4,
+# but exhibits up to 10% improvement on other cores.
+#
+# Second version is "monolithic" replacement for aes_core.c, which in
+# addition to AES_[de|en]crypt implements AES_set_[de|en]cryption_key.
+# This made it possible to implement little-endian variant of the
+# algorithm without modifying the base C code. Motivating factor for
+# the undertaken effort was that it appeared that in tight IA-32
+# register window little-endian flavor could achieve slightly higher
+# Instruction Level Parallelism, and it indeed resulted in up to 15%
+# better performance on most recent µ-archs...
+#
+# Third version adds AES_cbc_encrypt implementation, which resulted in
+# up to 40% performance imrovement of CBC benchmark results. 40% was
+# observed on P4 core, where "overall" imrovement coefficient, i.e. if
+# compared to PIC generated by GCC and in CBC mode, was observed to be
+# as large as 4x:-) CBC performance is virtually identical to ECB now
+# and on some platforms even better, e.g. 17.6 "small" cycles/byte on
+# Opteron, because certain function prologues and epilogues are
+# effectively taken out of the loop...
+#
+# Version 3.2 implements compressed tables and prefetch of these tables
+# in CBC[!] mode. Former means that 3/4 of table references are now
+# misaligned, which unfortunately has negative impact on elder IA-32
+# implementations, Pentium suffered 30% penalty, PIII - 10%.
+#
+# Version 3.3 avoids L1 cache aliasing between stack frame and
+# S-boxes, and 3.4 - L1 cache aliasing even between key schedule. The
+# latter is achieved by copying the key schedule to controlled place in
+# stack. This unfortunately has rather strong impact on small block CBC
+# performance, ~2x deterioration on 16-byte block if compared to 3.3.
+#
+# Version 3.5 checks if there is L1 cache aliasing between user-supplied
+# key schedule and S-boxes and abstains from copying the former if
+# there is no. This allows end-user to consciously retain small block
+# performance by aligning key schedule in specific manner.
+#
+# Version 3.6 compresses Td4 to 256 bytes and prefetches it in ECB.
+#
+# Current ECB performance numbers for 128-bit key in CPU cycles per
+# processed byte [measure commonly used by AES benchmarkers] are:
+#
+#		small footprint		fully unrolled
+# P4		24			22
+# AMD K8	20			19
+# PIII		25			23
+# Pentium	81			78
+#
+# Version 3.7 reimplements outer rounds as "compact." Meaning that
+# first and last rounds reference compact 256 bytes S-box. This means
+# that first round consumes a lot more CPU cycles and that encrypt
+# and decrypt performance becomes asymmetric. Encrypt performance
+# drops by 10-12%, while decrypt - by 20-25%:-( 256 bytes S-box is
+# aggressively pre-fetched.
+#
+# Version 4.0 effectively rolls back to 3.6 and instead implements
+# additional set of functions, _[x86|sse]_AES_[en|de]crypt_compact,
+# which use exclusively 256 byte S-box. These functions are to be
+# called in modes not concealing plain text, such as ECB, or when
+# we're asked to process smaller amount of data [or unconditionally
+# on hyper-threading CPU]. Currently it's called unconditionally from
+# AES_[en|de]crypt, which affects all modes, but CBC. CBC routine
+# still needs to be modified to switch between slower and faster
+# mode when appropriate... But in either case benchmark landscape
+# changes dramatically and below numbers are CPU cycles per processed
+# byte for 128-bit key.
+#
+#		ECB encrypt	ECB decrypt	CBC large chunk
+# P4		56[60]		84[100]		23
+# AMD K8	48[44]		70[79]		18
+# PIII		41[50]		61[91]		24
+# Core 2	32[38]		45[70]		18.5
+# Pentium	120		160		77
+#
+# Version 4.1 switches to compact S-box even in key schedule setup.
+#
+# Version 4.2 prefetches compact S-box in every SSE round or in other
+# words every cache-line is *guaranteed* to be accessed within ~50
+# cycles window. Why just SSE? Because it's needed on hyper-threading
+# CPU! Which is also why it's prefetched with 64 byte stride. Best
+# part is that it has no negative effect on performance:-)  
+#
+# Version 4.3 implements switch between compact and non-compact block
+# functions in AES_cbc_encrypt depending on how much data was asked
+# to be processed in one stroke.
+#
+######################################################################
+# Timing attacks are classified in two classes: synchronous when
+# attacker consciously initiates cryptographic operation and collects
+# timing data of various character afterwards, and asynchronous when
+# malicious code is executed on same CPU simultaneously with AES,
+# instruments itself and performs statistical analysis of this data.
+#
+# As far as synchronous attacks go the root to the AES timing
+# vulnerability is twofold. Firstly, of 256 S-box elements at most 160
+# are referred to in single 128-bit block operation. Well, in C
+# implementation with 4 distinct tables it's actually as little as 40
+# references per 256 elements table, but anyway... Secondly, even
+# though S-box elements are clustered into smaller amount of cache-
+# lines, smaller than 160 and even 40, it turned out that for certain
+# plain-text pattern[s] or simply put chosen plain-text and given key
+# few cache-lines remain unaccessed during block operation. Now, if
+# attacker can figure out this access pattern, he can deduct the key
+# [or at least part of it]. The natural way to mitigate this kind of
+# attacks is to minimize the amount of cache-lines in S-box and/or
+# prefetch them to ensure that every one is accessed for more uniform
+# timing. But note that *if* plain-text was concealed in such way that
+# input to block function is distributed *uniformly*, then attack
+# wouldn't apply. Now note that some encryption modes, most notably
+# CBC, do mask the plain-text in this exact way [secure cipher output
+# is distributed uniformly]. Yes, one still might find input that
+# would reveal the information about given key, but if amount of
+# candidate inputs to be tried is larger than amount of possible key
+# combinations then attack becomes infeasible. This is why revised
+# AES_cbc_encrypt "dares" to switch to larger S-box when larger chunk
+# of data is to be processed in one stroke. The current size limit of
+# 512 bytes is chosen to provide same [diminishigly low] probability
+# for cache-line to remain untouched in large chunk operation with
+# large S-box as for single block operation with compact S-box and
+# surely needs more careful consideration...
+#
+# As for asynchronous attacks. There are two flavours: attacker code
+# being interleaved with AES on hyper-threading CPU at *instruction*
+# level, and two processes time sharing single core. As for latter.
+# Two vectors. 1. Given that attacker process has higher priority,
+# yield execution to process performing AES just before timer fires
+# off the scheduler, immediately regain control of CPU and analyze the
+# cache state. For this attack to be efficient attacker would have to
+# effectively slow down the operation by several *orders* of magnitute,
+# by ratio of time slice to duration of handful of AES rounds, which
+# unlikely to remain unnoticed. Not to mention that this also means
+# that he would spend correspondigly more time to collect enough
+# statistical data to mount the attack. It's probably appropriate to
+# say that if adeversary reckons that this attack is beneficial and
+# risks to be noticed, you probably have larger problems having him
+# mere opportunity. In other words suggested code design expects you
+# to preclude/mitigate this attack by overall system security design.
+# 2. Attacker manages to make his code interrupt driven. In order for
+# this kind of attack to be feasible, interrupt rate has to be high
+# enough, again comparable to duration of handful of AES rounds. But
+# is there interrupt source of such rate? Hardly, not even 1Gbps NIC
+# generates interrupts at such raging rate...
+#
+# And now back to the former, hyper-threading CPU or more specifically
+# Intel P4. Recall that asynchronous attack implies that malicious
+# code instruments itself. And naturally instrumentation granularity
+# has be noticeably lower than duration of codepath accessing S-box.
+# Given that all cache-lines are accessed during that time that is.
+# Current implementation accesses *all* cache-lines within ~50 cycles
+# window, which is actually *less* than RDTSC latency on Intel P4!
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+push(@INC,"${dir}","${dir}../../perlasm");
+require "x86asm.pl";
+
+&asm_init($ARGV[0],"aes-586.pl",$x86only = $ARGV[$#ARGV] eq "386");
+&static_label("AES_Te");
+&static_label("AES_Td");
+
+$s0="eax";
+$s1="ebx";
+$s2="ecx";
+$s3="edx";
+$key="edi";
+$acc="esi";
+$tbl="ebp";
+
+# stack frame layout in _[x86|sse]_AES_* routines, frame is allocated
+# by caller
+$__ra=&DWP(0,"esp");	# return address
+$__s0=&DWP(4,"esp");	# s0 backing store
+$__s1=&DWP(8,"esp");	# s1 backing store
+$__s2=&DWP(12,"esp");	# s2 backing store
+$__s3=&DWP(16,"esp");	# s3 backing store
+$__key=&DWP(20,"esp");	# pointer to key schedule
+$__end=&DWP(24,"esp");	# pointer to end of key schedule
+$__tbl=&DWP(28,"esp");	# %ebp backing store
+
+# stack frame layout in AES_[en|crypt] routines, which differs from
+# above by 4 and overlaps by %ebp backing store
+$_tbl=&DWP(24,"esp");
+$_esp=&DWP(28,"esp");
+
+sub _data_word() { my $i; while(defined($i=shift)) { &data_word($i,$i); } }
+
+$speed_limit=512;	# chunks smaller than $speed_limit are
+			# processed with compact routine in CBC mode
+$small_footprint=1;	# $small_footprint=1 code is ~5% slower [on
+			# recent µ-archs], but ~5 times smaller!
+			# I favor compact code to minimize cache
+			# contention and in hope to "collect" 5% back
+			# in real-life applications...
+
+$vertical_spin=0;	# shift "verticaly" defaults to 0, because of
+			# its proof-of-concept status...
+# Note that there is no decvert(), as well as last encryption round is
+# performed with "horizontal" shifts. This is because this "vertical"
+# implementation [one which groups shifts on a given $s[i] to form a
+# "column," unlike "horizontal" one, which groups shifts on different
+# $s[i] to form a "row"] is work in progress. It was observed to run
+# few percents faster on Intel cores, but not AMD. On AMD K8 core it's
+# whole 12% slower:-( So we face a trade-off... Shall it be resolved
+# some day? Till then the code is considered experimental and by
+# default remains dormant...
+
+sub encvert()
+{ my ($te,@s) = @_;
+  my $v0 = $acc, $v1 = $key;
+
+	&mov	($v0,$s[3]);				# copy s3
+	&mov	(&DWP(4,"esp"),$s[2]);			# save s2
+	&mov	($v1,$s[0]);				# copy s0
+	&mov	(&DWP(8,"esp"),$s[1]);			# save s1
+
+	&movz	($s[2],&HB($s[0]));
+	&and	($s[0],0xFF);
+	&mov	($s[0],&DWP(0,$te,$s[0],8));		# s0>>0
+	&shr	($v1,16);
+	&mov	($s[3],&DWP(3,$te,$s[2],8));		# s0>>8
+	&movz	($s[1],&HB($v1));
+	&and	($v1,0xFF);
+	&mov	($s[2],&DWP(2,$te,$v1,8));		# s0>>16
+	 &mov	($v1,$v0);
+	&mov	($s[1],&DWP(1,$te,$s[1],8));		# s0>>24
+
+	&and	($v0,0xFF);
+	&xor	($s[3],&DWP(0,$te,$v0,8));		# s3>>0
+	&movz	($v0,&HB($v1));
+	&shr	($v1,16);
+	&xor	($s[2],&DWP(3,$te,$v0,8));		# s3>>8
+	&movz	($v0,&HB($v1));
+	&and	($v1,0xFF);
+	&xor	($s[1],&DWP(2,$te,$v1,8));		# s3>>16
+	 &mov	($v1,&DWP(4,"esp"));			# restore s2
+	&xor	($s[0],&DWP(1,$te,$v0,8));		# s3>>24
+
+	&mov	($v0,$v1);
+	&and	($v1,0xFF);
+	&xor	($s[2],&DWP(0,$te,$v1,8));		# s2>>0
+	&movz	($v1,&HB($v0));
+	&shr	($v0,16);
+	&xor	($s[1],&DWP(3,$te,$v1,8));		# s2>>8
+	&movz	($v1,&HB($v0));
+	&and	($v0,0xFF);
+	&xor	($s[0],&DWP(2,$te,$v0,8));		# s2>>16
+	 &mov	($v0,&DWP(8,"esp"));			# restore s1
+	&xor	($s[3],&DWP(1,$te,$v1,8));		# s2>>24
+
+	&mov	($v1,$v0);
+	&and	($v0,0xFF);
+	&xor	($s[1],&DWP(0,$te,$v0,8));		# s1>>0
+	&movz	($v0,&HB($v1));
+	&shr	($v1,16);
+	&xor	($s[0],&DWP(3,$te,$v0,8));		# s1>>8
+	&movz	($v0,&HB($v1));
+	&and	($v1,0xFF);
+	&xor	($s[3],&DWP(2,$te,$v1,8));		# s1>>16
+	 &mov	($key,$__key);				# reincarnate v1 as key
+	&xor	($s[2],&DWP(1,$te,$v0,8));		# s1>>24
+}
+
+# Another experimental routine, which features "horizontal spin," but
+# eliminates one reference to stack. Strangely enough runs slower...
+sub enchoriz()
+{ my $v0 = $key, $v1 = $acc;
+
+	&movz	($v0,&LB($s0));			#  3, 2, 1, 0*
+	&rotr	($s2,8);			#  8,11,10, 9
+	&mov	($v1,&DWP(0,$te,$v0,8));	#  0
+	&movz	($v0,&HB($s1));			#  7, 6, 5*, 4
+	&rotr	($s3,16);			# 13,12,15,14
+	&xor	($v1,&DWP(3,$te,$v0,8));	#  5
+	&movz	($v0,&HB($s2));			#  8,11,10*, 9
+	&rotr	($s0,16);			#  1, 0, 3, 2
+	&xor	($v1,&DWP(2,$te,$v0,8));	# 10
+	&movz	($v0,&HB($s3));			# 13,12,15*,14
+	&xor	($v1,&DWP(1,$te,$v0,8));	# 15, t[0] collected
+	&mov	($__s0,$v1);			# t[0] saved
+
+	&movz	($v0,&LB($s1));			#  7, 6, 5, 4*
+	&shr	($s1,16);			#  -, -, 7, 6
+	&mov	($v1,&DWP(0,$te,$v0,8));	#  4
+	&movz	($v0,&LB($s3));			# 13,12,15,14*
+	&xor	($v1,&DWP(2,$te,$v0,8));	# 14
+	&movz	($v0,&HB($s0));			#  1, 0, 3*, 2
+	&and	($s3,0xffff0000);		# 13,12, -, -
+	&xor	($v1,&DWP(1,$te,$v0,8));	#  3
+	&movz	($v0,&LB($s2));			#  8,11,10, 9*
+	&or	($s3,$s1);			# 13,12, 7, 6
+	&xor	($v1,&DWP(3,$te,$v0,8));	#  9, t[1] collected
+	&mov	($s1,$v1);			#  s[1]=t[1]
+
+	&movz	($v0,&LB($s0));			#  1, 0, 3, 2*
+	&shr	($s2,16);			#  -, -, 8,11
+	&mov	($v1,&DWP(2,$te,$v0,8));	#  2
+	&movz	($v0,&HB($s3));			# 13,12, 7*, 6
+	&xor	($v1,&DWP(1,$te,$v0,8));	#  7
+	&movz	($v0,&HB($s2));			#  -, -, 8*,11
+	&xor	($v1,&DWP(0,$te,$v0,8));	#  8
+	&mov	($v0,$s3);
+	&shr	($v0,24);			# 13
+	&xor	($v1,&DWP(3,$te,$v0,8));	# 13, t[2] collected
+
+	&movz	($v0,&LB($s2));			#  -, -, 8,11*
+	&shr	($s0,24);			#  1*
+	&mov	($s2,&DWP(1,$te,$v0,8));	# 11
+	&xor	($s2,&DWP(3,$te,$s0,8));	#  1
+	&mov	($s0,$__s0);			# s[0]=t[0]
+	&movz	($v0,&LB($s3));			# 13,12, 7, 6*
+	&shr	($s3,16);			#   ,  ,13,12
+	&xor	($s2,&DWP(2,$te,$v0,8));	#  6
+	&mov	($key,$__key);			# reincarnate v0 as key
+	&and	($s3,0xff);			#   ,  ,13,12*
+	&mov	($s3,&DWP(0,$te,$s3,8));	# 12
+	&xor	($s3,$s2);			# s[2]=t[3] collected
+	&mov	($s2,$v1);			# s[2]=t[2]
+}
+
+# More experimental code... SSE one... Even though this one eliminates
+# *all* references to stack, it's not faster...
+sub sse_encbody()
+{
+	&movz	($acc,&LB("eax"));		#  0
+	&mov	("ecx",&DWP(0,$tbl,$acc,8));	#  0
+	&pshufw	("mm2","mm0",0x0d);		#  7, 6, 3, 2
+	&movz	("edx",&HB("eax"));		#  1
+	&mov	("edx",&DWP(3,$tbl,"edx",8));	#  1
+	&shr	("eax",16);			#  5, 4
+
+	&movz	($acc,&LB("ebx"));		# 10
+	&xor	("ecx",&DWP(2,$tbl,$acc,8));	# 10
+	&pshufw	("mm6","mm4",0x08);		# 13,12, 9, 8
+	&movz	($acc,&HB("ebx"));		# 11
+	&xor	("edx",&DWP(1,$tbl,$acc,8));	# 11
+	&shr	("ebx",16);			# 15,14
+
+	&movz	($acc,&HB("eax"));		#  5
+	&xor	("ecx",&DWP(3,$tbl,$acc,8));	#  5
+	&movq	("mm3",QWP(16,$key));
+	&movz	($acc,&HB("ebx"));		# 15
+	&xor	("ecx",&DWP(1,$tbl,$acc,8));	# 15
+	&movd	("mm0","ecx");			# t[0] collected
+
+	&movz	($acc,&LB("eax"));		#  4
+	&mov	("ecx",&DWP(0,$tbl,$acc,8));	#  4
+	&movd	("eax","mm2");			#  7, 6, 3, 2
+	&movz	($acc,&LB("ebx"));		# 14
+	&xor	("ecx",&DWP(2,$tbl,$acc,8));	# 14
+	&movd	("ebx","mm6");			# 13,12, 9, 8
+
+	&movz	($acc,&HB("eax"));		#  3
+	&xor	("ecx",&DWP(1,$tbl,$acc,8));	#  3
+	&movz	($acc,&HB("ebx"));		#  9
+	&xor	("ecx",&DWP(3,$tbl,$acc,8));	#  9
+	&movd	("mm1","ecx");			# t[1] collected
+
+	&movz	($acc,&LB("eax"));		#  2
+	&mov	("ecx",&DWP(2,$tbl,$acc,8));	#  2
+	&shr	("eax",16);			#  7, 6
+	&punpckldq	("mm0","mm1");		# t[0,1] collected
+	&movz	($acc,&LB("ebx"));		#  8
+	&xor	("ecx",&DWP(0,$tbl,$acc,8));	#  8
+	&shr	("ebx",16);			# 13,12
+
+	&movz	($acc,&HB("eax"));		#  7
+	&xor	("ecx",&DWP(1,$tbl,$acc,8));	#  7
+	&pxor	("mm0","mm3");
+	&movz	("eax",&LB("eax"));		#  6
+	&xor	("edx",&DWP(2,$tbl,"eax",8));	#  6
+	&pshufw	("mm1","mm0",0x08);		#  5, 4, 1, 0
+	&movz	($acc,&HB("ebx"));		# 13
+	&xor	("ecx",&DWP(3,$tbl,$acc,8));	# 13
+	&xor	("ecx",&DWP(24,$key));		# t[2]
+	&movd	("mm4","ecx");			# t[2] collected
+	&movz	("ebx",&LB("ebx"));		# 12
+	&xor	("edx",&DWP(0,$tbl,"ebx",8));	# 12
+	&shr	("ecx",16);
+	&movd	("eax","mm1");			#  5, 4, 1, 0
+	&mov	("ebx",&DWP(28,$key));		# t[3]
+	&xor	("ebx","edx");
+	&movd	("mm5","ebx");			# t[3] collected
+	&and	("ebx",0xffff0000);
+	&or	("ebx","ecx");
+
+	&punpckldq	("mm4","mm5");		# t[2,3] collected
+}
+
+######################################################################
+# "Compact" block function
+######################################################################
+
+sub enccompact()
+{ my $Fn = mov;
+  while ($#_>5) { pop(@_); $Fn=sub{}; }
+  my ($i,$te,@s)=@_;
+  my $tmp = $key;
+  my $out = $i==3?$s[0]:$acc;
+
+	# $Fn is used in first compact round and its purpose is to
+	# void restoration of some values from stack, so that after
+	# 4xenccompact with extra argument $key value is left there...
+	if ($i==3)  {	&$Fn	($key,$__key);			}##%edx
+	else        {	&mov	($out,$s[0]);			}
+			&and	($out,0xFF);
+	if ($i==1)  {	&shr	($s[0],16);			}#%ebx[1]
+	if ($i==2)  {	&shr	($s[0],24);			}#%ecx[2]
+			&movz	($out,&BP(-128,$te,$out,1));
+
+	if ($i==3)  {	$tmp=$s[1];				}##%eax
+			&movz	($tmp,&HB($s[1]));
+			&movz	($tmp,&BP(-128,$te,$tmp,1));
+			&shl	($tmp,8);
+			&xor	($out,$tmp);
+
+	if ($i==3)  {	$tmp=$s[2]; &mov ($s[1],$__s0);		}##%ebx
+	else        {	&mov	($tmp,$s[2]);
+			&shr	($tmp,16);			}
+	if ($i==2)  {	&and	($s[1],0xFF);			}#%edx[2]
+			&and	($tmp,0xFF);
+			&movz	($tmp,&BP(-128,$te,$tmp,1));
+			&shl	($tmp,16);
+			&xor	($out,$tmp);
+
+	if ($i==3)  {	$tmp=$s[3]; &mov ($s[2],$__s1);		}##%ecx
+	elsif($i==2){	&movz	($tmp,&HB($s[3]));		}#%ebx[2]
+	else        {	&mov	($tmp,$s[3]);
+			&shr	($tmp,24);			}
+			&movz	($tmp,&BP(-128,$te,$tmp,1));
+			&shl	($tmp,24);
+			&xor	($out,$tmp);
+	if ($i<2)   {	&mov	(&DWP(4+4*$i,"esp"),$out);	}
+	if ($i==3)  {	&mov	($s[3],$acc);			}
+	&comment();
+}
+
+sub enctransform()
+{ my @s = ($s0,$s1,$s2,$s3);
+  my $i = shift;
+  my $tmp = $tbl;
+  my $r2  = $key ;
+
+	&mov	($acc,$s[$i]);
+	&and	($acc,0x80808080);
+	&mov	($tmp,$acc);
+	&shr	($tmp,7);
+	&lea	($r2,&DWP(0,$s[$i],$s[$i]));
+	&sub	($acc,$tmp);
+	&and	($r2,0xfefefefe);
+	&and	($acc,0x1b1b1b1b);
+	&mov	($tmp,$s[$i]);
+	&xor	($acc,$r2);	# r2
+
+	&xor	($s[$i],$acc);	# r0 ^ r2
+	&rotl	($s[$i],24);
+	&xor	($s[$i],$acc)	# ROTATE(r2^r0,24) ^ r2
+	&rotr	($tmp,16);
+	&xor	($s[$i],$tmp);
+	&rotr	($tmp,8);
+	&xor	($s[$i],$tmp);
+}
+
+&function_begin_B("_x86_AES_encrypt_compact");
+	# note that caller is expected to allocate stack frame for me!
+	&mov	($__key,$key);			# save key
+
+	&xor	($s0,&DWP(0,$key));		# xor with key
+	&xor	($s1,&DWP(4,$key));
+	&xor	($s2,&DWP(8,$key));
+	&xor	($s3,&DWP(12,$key));
+
+	&mov	($acc,&DWP(240,$key));		# load key->rounds
+	&lea	($acc,&DWP(-2,$acc,$acc));
+	&lea	($acc,&DWP(0,$key,$acc,8));
+	&mov	($__end,$acc);			# end of key schedule
+
+	# prefetch Te4
+	&mov	($key,&DWP(0-128,$tbl));
+	&mov	($acc,&DWP(32-128,$tbl));
+	&mov	($key,&DWP(64-128,$tbl));
+	&mov	($acc,&DWP(96-128,$tbl));
+	&mov	($key,&DWP(128-128,$tbl));
+	&mov	($acc,&DWP(160-128,$tbl));
+	&mov	($key,&DWP(192-128,$tbl));
+	&mov	($acc,&DWP(224-128,$tbl));
+
+	&set_label("loop",16);
+
+		&enccompact(0,$tbl,$s0,$s1,$s2,$s3,1);
+		&enccompact(1,$tbl,$s1,$s2,$s3,$s0,1);
+		&enccompact(2,$tbl,$s2,$s3,$s0,$s1,1);
+		&enccompact(3,$tbl,$s3,$s0,$s1,$s2,1);
+		&enctransform(2);
+		&enctransform(3);
+		&enctransform(0);
+		&enctransform(1);
+		&mov 	($key,$__key);
+		&mov	($tbl,$__tbl);
+		&add	($key,16);		# advance rd_key
+		&xor	($s0,&DWP(0,$key));
+		&xor	($s1,&DWP(4,$key));
+		&xor	($s2,&DWP(8,$key));
+		&xor	($s3,&DWP(12,$key));
+
+	&cmp	($key,$__end);
+	&mov	($__key,$key);
+	&jb	(&label("loop"));
+
+	&enccompact(0,$tbl,$s0,$s1,$s2,$s3);
+	&enccompact(1,$tbl,$s1,$s2,$s3,$s0);
+	&enccompact(2,$tbl,$s2,$s3,$s0,$s1);
+	&enccompact(3,$tbl,$s3,$s0,$s1,$s2);
+
+	&xor	($s0,&DWP(16,$key));
+	&xor	($s1,&DWP(20,$key));
+	&xor	($s2,&DWP(24,$key));
+	&xor	($s3,&DWP(28,$key));
+
+	&ret	();
+&function_end_B("_x86_AES_encrypt_compact");
+
+######################################################################
+# "Compact" SSE block function.
+######################################################################
+#
+# Performance is not actually extraordinary in comparison to pure
+# x86 code. In particular encrypt performance is virtually the same.
+# Decrypt performance on the other hand is 15-20% better on newer
+# µ-archs [but we're thankful for *any* improvement here], and ~50%
+# better on PIII:-) And additionally on the pros side this code
+# eliminates redundant references to stack and thus relieves/
+# minimizes the pressure on the memory bus.
+#
+# MMX register layout                           lsb
+# +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
+# |          mm4          |          mm0          |
+# +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
+# |     s3    |     s2    |     s1    |     s0    |    
+# +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
+# |15|14|13|12|11|10| 9| 8| 7| 6| 5| 4| 3| 2| 1| 0|
+# +--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+
+#
+# Indexes translate as s[N/4]>>(8*(N%4)), e.g. 5 means s1>>8.
+# In this terms encryption and decryption "compact" permutation
+# matrices can be depicted as following:
+#
+# encryption              lsb	# decryption              lsb
+# +----++----+----+----+----+	# +----++----+----+----+----+
+# | t0 || 15 | 10 |  5 |  0 |	# | t0 ||  7 | 10 | 13 |  0 |
+# +----++----+----+----+----+	# +----++----+----+----+----+
+# | t1 ||  3 | 14 |  9 |  4 |	# | t1 || 11 | 14 |  1 |  4 |
+# +----++----+----+----+----+	# +----++----+----+----+----+
+# | t2 ||  7 |  2 | 13 |  8 |	# | t2 || 15 |  2 |  5 |  8 |
+# +----++----+----+----+----+	# +----++----+----+----+----+
+# | t3 || 11 |  6 |  1 | 12 |	# | t3 ||  3 |  6 |  9 | 12 |
+# +----++----+----+----+----+	# +----++----+----+----+----+
+#
+######################################################################
+# Why not xmm registers? Short answer. It was actually tested and
+# was not any faster, but *contrary*, most notably on Intel CPUs.
+# Longer answer. Main advantage of using mm registers is that movd
+# latency is lower, especially on Intel P4. While arithmetic
+# instructions are twice as many, they can be scheduled every cycle
+# and not every second one when they are operating on xmm register,
+# so that "arithmetic throughput" remains virtually the same. And
+# finally the code can be executed even on elder SSE-only CPUs:-)
+
+sub sse_enccompact()
+{
+	&pshufw	("mm1","mm0",0x08);		#  5, 4, 1, 0
+	&pshufw	("mm5","mm4",0x0d);		# 15,14,11,10
+	&movd	("eax","mm1");			#  5, 4, 1, 0
+	&movd	("ebx","mm5");			# 15,14,11,10
+
+	&movz	($acc,&LB("eax"));		#  0
+	&movz	("ecx",&BP(-128,$tbl,$acc,1));	#  0
+	&pshufw	("mm2","mm0",0x0d);		#  7, 6, 3, 2
+	&movz	("edx",&HB("eax"));		#  1
+	&movz	("edx",&BP(-128,$tbl,"edx",1));	#  1
+	&shl	("edx",8);			#  1
+	&shr	("eax",16);			#  5, 4
+
+	&movz	($acc,&LB("ebx"));		# 10
+	&movz	($acc,&BP(-128,$tbl,$acc,1));	# 10
+	&shl	($acc,16);			# 10
+	&or	("ecx",$acc);			# 10
+	&pshufw	("mm6","mm4",0x08);		# 13,12, 9, 8
+	&movz	($acc,&HB("ebx"));		# 11
+	&movz	($acc,&BP(-128,$tbl,$acc,1));	# 11
+	&shl	($acc,24);			# 11
+	&or	("edx",$acc);			# 11
+	&shr	("ebx",16);			# 15,14
+
+	&movz	($acc,&HB("eax"));		#  5
+	&movz	($acc,&BP(-128,$tbl,$acc,1));	#  5
+	&shl	($acc,8);			#  5
+	&or	("ecx",$acc);			#  5
+	&movz	($acc,&HB("ebx"));		# 15
+	&movz	($acc,&BP(-128,$tbl,$acc,1));	# 15
+	&shl	($acc,24);			# 15
+	&or	("ecx",$acc);			# 15
+	&movd	("mm0","ecx");			# t[0] collected
+
+	&movz	($acc,&LB("eax"));		#  4
+	&movz	("ecx",&BP(-128,$tbl,$acc,1));	#  4
+	&movd	("eax","mm2");			#  7, 6, 3, 2
+	&movz	($acc,&LB("ebx"));		# 14
+	&movz	($acc,&BP(-128,$tbl,$acc,1));	# 14
+	&shl	($acc,16);			# 14
+	&or	("ecx",$acc);			# 14
+
+	&movd	("ebx","mm6");			# 13,12, 9, 8
+	&movz	($acc,&HB("eax"));		#  3
+	&movz	($acc,&BP(-128,$tbl,$acc,1));	#  3
+	&shl	($acc,24);			#  3
+	&or	("ecx",$acc);			#  3
+	&movz	($acc,&HB("ebx"));		#  9
+	&movz	($acc,&BP(-128,$tbl,$acc,1));	#  9
+	&shl	($acc,8);			#  9
+	&or	("ecx",$acc);			#  9
+	&movd	("mm1","ecx");			# t[1] collected
+
+	&movz	($acc,&LB("ebx"));		#  8
+	&movz	("ecx",&BP(-128,$tbl,$acc,1));	#  8
+	&shr	("ebx",16);			# 13,12
+	&movz	($acc,&LB("eax"));		#  2
+	&movz	($acc,&BP(-128,$tbl,$acc,1));	#  2
+	&shl	($acc,16);			#  2
+	&or	("ecx",$acc);			#  2
+	&shr	("eax",16);			#  7, 6
+
+	&punpckldq	("mm0","mm1");		# t[0,1] collected
+
+	&movz	($acc,&HB("eax"));		#  7
+	&movz	($acc,&BP(-128,$tbl,$acc,1));	#  7
+	&shl	($acc,24);			#  7
+	&or	("ecx",$acc);			#  7
+	&and	("eax",0xff);			#  6
+	&movz	("eax",&BP(-128,$tbl,"eax",1));	#  6
+	&shl	("eax",16);			#  6
+	&or	("edx","eax");			#  6
+	&movz	($acc,&HB("ebx"));		# 13
+	&movz	($acc,&BP(-128,$tbl,$acc,1));	# 13
+	&shl	($acc,8);			# 13
+	&or	("ecx",$acc);			# 13
+	&movd	("mm4","ecx");			# t[2] collected
+	&and	("ebx",0xff);			# 12
+	&movz	("ebx",&BP(-128,$tbl,"ebx",1));	# 12
+	&or	("edx","ebx");			# 12
+	&movd	("mm5","edx");			# t[3] collected
+
+	&punpckldq	("mm4","mm5");		# t[2,3] collected
+}
+
+					if (!$x86only) {
+&function_begin_B("_sse_AES_encrypt_compact");
+	&pxor	("mm0",&QWP(0,$key));	#  7, 6, 5, 4, 3, 2, 1, 0
+	&pxor	("mm4",&QWP(8,$key));	# 15,14,13,12,11,10, 9, 8
+
+	# note that caller is expected to allocate stack frame for me!
+	&mov	($acc,&DWP(240,$key));		# load key->rounds
+	&lea	($acc,&DWP(-2,$acc,$acc));
+	&lea	($acc,&DWP(0,$key,$acc,8));
+	&mov	($__end,$acc);			# end of key schedule
+
+	&mov	($s0,0x1b1b1b1b);		# magic constant
+	&mov	(&DWP(8,"esp"),$s0);
+	&mov	(&DWP(12,"esp"),$s0);
+
+	# prefetch Te4
+	&mov	($s0,&DWP(0-128,$tbl));
+	&mov	($s1,&DWP(32-128,$tbl));
+	&mov	($s2,&DWP(64-128,$tbl));
+	&mov	($s3,&DWP(96-128,$tbl));
+	&mov	($s0,&DWP(128-128,$tbl));
+	&mov	($s1,&DWP(160-128,$tbl));
+	&mov	($s2,&DWP(192-128,$tbl));
+	&mov	($s3,&DWP(224-128,$tbl));
+
+	&set_label("loop",16);
+		&sse_enccompact();
+		&add	($key,16);
+		&cmp	($key,$__end);
+		&ja	(&label("out"));
+
+		&movq	("mm2",&QWP(8,"esp"));
+		&pxor	("mm3","mm3");		&pxor	("mm7","mm7");
+		&movq	("mm1","mm0");		&movq	("mm5","mm4");	# r0
+		&pcmpgtb("mm3","mm0");		&pcmpgtb("mm7","mm4");
+		&pand	("mm3","mm2");		&pand	("mm7","mm2");
+		&pshufw	("mm2","mm0",0xb1);	&pshufw	("mm6","mm4",0xb1);# ROTATE(r0,16)
+		&paddb	("mm0","mm0");		&paddb	("mm4","mm4");
+		&pxor	("mm0","mm3");		&pxor	("mm4","mm7");	# = r2
+		&pshufw	("mm3","mm2",0xb1);	&pshufw	("mm7","mm6",0xb1);# r0
+		&pxor	("mm1","mm0");		&pxor	("mm5","mm4");	# r0^r2
+		&pxor	("mm0","mm2");		&pxor	("mm4","mm6");	# ^= ROTATE(r0,16)
+
+		&movq	("mm2","mm3");		&movq	("mm6","mm7");
+		&pslld	("mm3",8);		&pslld	("mm7",8);
+		&psrld	("mm2",24);		&psrld	("mm6",24);
+		&pxor	("mm0","mm3");		&pxor	("mm4","mm7");	# ^= r0<<8
+		&pxor	("mm0","mm2");		&pxor	("mm4","mm6");	# ^= r0>>24
+
+		&movq	("mm3","mm1");		&movq	("mm7","mm5");
+		&movq	("mm2",&QWP(0,$key));	&movq	("mm6",&QWP(8,$key));
+		&psrld	("mm1",8);		&psrld	("mm5",8);
+		&mov	($s0,&DWP(0-128,$tbl));
+		&pslld	("mm3",24);		&pslld	("mm7",24);
+		&mov	($s1,&DWP(64-128,$tbl));
+		&pxor	("mm0","mm1");		&pxor	("mm4","mm5");	# ^= (r2^r0)<<8
+		&mov	($s2,&DWP(128-128,$tbl));
+		&pxor	("mm0","mm3");		&pxor	("mm4","mm7");	# ^= (r2^r0)>>24
+		&mov	($s3,&DWP(192-128,$tbl));
+
+		&pxor	("mm0","mm2");		&pxor	("mm4","mm6");
+	&jmp	(&label("loop"));
+
+	&set_label("out",16);
+	&pxor	("mm0",&QWP(0,$key));
+	&pxor	("mm4",&QWP(8,$key));
+
+	&ret	();
+&function_end_B("_sse_AES_encrypt_compact");
+					}
+
+######################################################################
+# Vanilla block function.
+######################################################################
+
+sub encstep()
+{ my ($i,$te,@s) = @_;
+  my $tmp = $key;
+  my $out = $i==3?$s[0]:$acc;
+
+	# lines marked with #%e?x[i] denote "reordered" instructions...
+	if ($i==3)  {	&mov	($key,$__key);			}##%edx
+	else        {	&mov	($out,$s[0]);
+			&and	($out,0xFF);			}
+	if ($i==1)  {	&shr	($s[0],16);			}#%ebx[1]
+	if ($i==2)  {	&shr	($s[0],24);			}#%ecx[2]
+			&mov	($out,&DWP(0,$te,$out,8));
+
+	if ($i==3)  {	$tmp=$s[1];				}##%eax
+			&movz	($tmp,&HB($s[1]));
+			&xor	($out,&DWP(3,$te,$tmp,8));
+
+	if ($i==3)  {	$tmp=$s[2]; &mov ($s[1],$__s0);		}##%ebx
+	else        {	&mov	($tmp,$s[2]);
+			&shr	($tmp,16);			}
+	if ($i==2)  {	&and	($s[1],0xFF);			}#%edx[2]
+			&and	($tmp,0xFF);
+			&xor	($out,&DWP(2,$te,$tmp,8));
+
+	if ($i==3)  {	$tmp=$s[3]; &mov ($s[2],$__s1);		}##%ecx
+	elsif($i==2){	&movz	($tmp,&HB($s[3]));		}#%ebx[2]
+	else        {	&mov	($tmp,$s[3]); 
+			&shr	($tmp,24)			}
+			&xor	($out,&DWP(1,$te,$tmp,8));
+	if ($i<2)   {	&mov	(&DWP(4+4*$i,"esp"),$out);	}
+	if ($i==3)  {	&mov	($s[3],$acc);			}
+			&comment();
+}
+
+sub enclast()
+{ my ($i,$te,@s)=@_;
+  my $tmp = $key;
+  my $out = $i==3?$s[0]:$acc;
+
+	if ($i==3)  {	&mov	($key,$__key);			}##%edx
+	else        {	&mov	($out,$s[0]);			}
+			&and	($out,0xFF);
+	if ($i==1)  {	&shr	($s[0],16);			}#%ebx[1]
+	if ($i==2)  {	&shr	($s[0],24);			}#%ecx[2]
+			&mov	($out,&DWP(2,$te,$out,8));
+			&and	($out,0x000000ff);
+
+	if ($i==3)  {	$tmp=$s[1];				}##%eax
+			&movz	($tmp,&HB($s[1]));
+			&mov	($tmp,&DWP(0,$te,$tmp,8));
+			&and	($tmp,0x0000ff00);
+			&xor	($out,$tmp);
+
+	if ($i==3)  {	$tmp=$s[2]; &mov ($s[1],$__s0);		}##%ebx
+	else        {	&mov	($tmp,$s[2]);
+			&shr	($tmp,16);			}
+	if ($i==2)  {	&and	($s[1],0xFF);			}#%edx[2]
+			&and	($tmp,0xFF);
+			&mov	($tmp,&DWP(0,$te,$tmp,8));
+			&and	($tmp,0x00ff0000);
+			&xor	($out,$tmp);
+
+	if ($i==3)  {	$tmp=$s[3]; &mov ($s[2],$__s1);		}##%ecx
+	elsif($i==2){	&movz	($tmp,&HB($s[3]));		}#%ebx[2]
+	else        {	&mov	($tmp,$s[3]);
+			&shr	($tmp,24);			}
+			&mov	($tmp,&DWP(2,$te,$tmp,8));
+			&and	($tmp,0xff000000);
+			&xor	($out,$tmp);
+	if ($i<2)   {	&mov	(&DWP(4+4*$i,"esp"),$out);	}
+	if ($i==3)  {	&mov	($s[3],$acc);			}
+}
+
+&function_begin_B("_x86_AES_encrypt");
+	if ($vertical_spin) {
+		# I need high parts of volatile registers to be accessible...
+		&exch	($s1="edi",$key="ebx");
+		&mov	($s2="esi",$acc="ecx");
+	}
+
+	# note that caller is expected to allocate stack frame for me!
+	&mov	($__key,$key);			# save key
+
+	&xor	($s0,&DWP(0,$key));		# xor with key
+	&xor	($s1,&DWP(4,$key));
+	&xor	($s2,&DWP(8,$key));
+	&xor	($s3,&DWP(12,$key));
+
+	&mov	($acc,&DWP(240,$key));		# load key->rounds
+
+	if ($small_footprint) {
+	    &lea	($acc,&DWP(-2,$acc,$acc));
+	    &lea	($acc,&DWP(0,$key,$acc,8));
+	    &mov	($__end,$acc);		# end of key schedule
+
+	    &set_label("loop",16);
+		if ($vertical_spin) {
+		    &encvert($tbl,$s0,$s1,$s2,$s3);
+		} else {
+		    &encstep(0,$tbl,$s0,$s1,$s2,$s3);
+		    &encstep(1,$tbl,$s1,$s2,$s3,$s0);
+		    &encstep(2,$tbl,$s2,$s3,$s0,$s1);
+		    &encstep(3,$tbl,$s3,$s0,$s1,$s2);
+		}
+		&add	($key,16);		# advance rd_key
+		&xor	($s0,&DWP(0,$key));
+		&xor	($s1,&DWP(4,$key));
+		&xor	($s2,&DWP(8,$key));
+		&xor	($s3,&DWP(12,$key));
+	    &cmp	($key,$__end);
+	    &mov	($__key,$key);
+	    &jb		(&label("loop"));
+	}
+	else {
+	    &cmp	($acc,10);
+	    &jle	(&label("10rounds"));
+	    &cmp	($acc,12);
+	    &jle	(&label("12rounds"));
+
+	&set_label("14rounds",4);
+	    for ($i=1;$i<3;$i++) {
+		if ($vertical_spin) {
+		    &encvert($tbl,$s0,$s1,$s2,$s3);
+		} else {
+		    &encstep(0,$tbl,$s0,$s1,$s2,$s3);
+		    &encstep(1,$tbl,$s1,$s2,$s3,$s0);
+		    &encstep(2,$tbl,$s2,$s3,$s0,$s1);
+		    &encstep(3,$tbl,$s3,$s0,$s1,$s2);
+		}
+		&xor	($s0,&DWP(16*$i+0,$key));
+		&xor	($s1,&DWP(16*$i+4,$key));
+		&xor	($s2,&DWP(16*$i+8,$key));
+		&xor	($s3,&DWP(16*$i+12,$key));
+	    }
+	    &add	($key,32);
+	    &mov	($__key,$key);		# advance rd_key
+	&set_label("12rounds",4);
+	    for ($i=1;$i<3;$i++) {
+		if ($vertical_spin) {
+		    &encvert($tbl,$s0,$s1,$s2,$s3);
+		} else {
+		    &encstep(0,$tbl,$s0,$s1,$s2,$s3);
+		    &encstep(1,$tbl,$s1,$s2,$s3,$s0);
+		    &encstep(2,$tbl,$s2,$s3,$s0,$s1);
+		    &encstep(3,$tbl,$s3,$s0,$s1,$s2);
+		}
+		&xor	($s0,&DWP(16*$i+0,$key));
+		&xor	($s1,&DWP(16*$i+4,$key));
+		&xor	($s2,&DWP(16*$i+8,$key));
+		&xor	($s3,&DWP(16*$i+12,$key));
+	    }
+	    &add	($key,32);
+	    &mov	($__key,$key);		# advance rd_key
+	&set_label("10rounds",4);
+	    for ($i=1;$i<10;$i++) {
+		if ($vertical_spin) {
+		    &encvert($tbl,$s0,$s1,$s2,$s3);
+		} else {
+		    &encstep(0,$tbl,$s0,$s1,$s2,$s3);
+		    &encstep(1,$tbl,$s1,$s2,$s3,$s0);
+		    &encstep(2,$tbl,$s2,$s3,$s0,$s1);
+		    &encstep(3,$tbl,$s3,$s0,$s1,$s2);
+		}
+		&xor	($s0,&DWP(16*$i+0,$key));
+		&xor	($s1,&DWP(16*$i+4,$key));
+		&xor	($s2,&DWP(16*$i+8,$key));
+		&xor	($s3,&DWP(16*$i+12,$key));
+	    }
+	}
+
+	if ($vertical_spin) {
+	    # "reincarnate" some registers for "horizontal" spin...
+	    &mov	($s1="ebx",$key="edi");
+	    &mov	($s2="ecx",$acc="esi");
+	}
+	&enclast(0,$tbl,$s0,$s1,$s2,$s3);
+	&enclast(1,$tbl,$s1,$s2,$s3,$s0);
+	&enclast(2,$tbl,$s2,$s3,$s0,$s1);
+	&enclast(3,$tbl,$s3,$s0,$s1,$s2);
+
+	&add	($key,$small_footprint?16:160);
+	&xor	($s0,&DWP(0,$key));
+	&xor	($s1,&DWP(4,$key));
+	&xor	($s2,&DWP(8,$key));
+	&xor	($s3,&DWP(12,$key));
+
+	&ret	();
+
+&set_label("AES_Te",64);	# Yes! I keep it in the code segment!
+	&_data_word(0xa56363c6, 0x847c7cf8, 0x997777ee, 0x8d7b7bf6);
+	&_data_word(0x0df2f2ff, 0xbd6b6bd6, 0xb16f6fde, 0x54c5c591);
+	&_data_word(0x50303060, 0x03010102, 0xa96767ce, 0x7d2b2b56);
+	&_data_word(0x19fefee7, 0x62d7d7b5, 0xe6abab4d, 0x9a7676ec);
+	&_data_word(0x45caca8f, 0x9d82821f, 0x40c9c989, 0x877d7dfa);
+	&_data_word(0x15fafaef, 0xeb5959b2, 0xc947478e, 0x0bf0f0fb);
+	&_data_word(0xecadad41, 0x67d4d4b3, 0xfda2a25f, 0xeaafaf45);
+	&_data_word(0xbf9c9c23, 0xf7a4a453, 0x967272e4, 0x5bc0c09b);
+	&_data_word(0xc2b7b775, 0x1cfdfde1, 0xae93933d, 0x6a26264c);
+	&_data_word(0x5a36366c, 0x413f3f7e, 0x02f7f7f5, 0x4fcccc83);
+	&_data_word(0x5c343468, 0xf4a5a551, 0x34e5e5d1, 0x08f1f1f9);
+	&_data_word(0x937171e2, 0x73d8d8ab, 0x53313162, 0x3f15152a);
+	&_data_word(0x0c040408, 0x52c7c795, 0x65232346, 0x5ec3c39d);
+	&_data_word(0x28181830, 0xa1969637, 0x0f05050a, 0xb59a9a2f);
+	&_data_word(0x0907070e, 0x36121224, 0x9b80801b, 0x3de2e2df);
+	&_data_word(0x26ebebcd, 0x6927274e, 0xcdb2b27f, 0x9f7575ea);
+	&_data_word(0x1b090912, 0x9e83831d, 0x742c2c58, 0x2e1a1a34);
+	&_data_word(0x2d1b1b36, 0xb26e6edc, 0xee5a5ab4, 0xfba0a05b);
+	&_data_word(0xf65252a4, 0x4d3b3b76, 0x61d6d6b7, 0xceb3b37d);
+	&_data_word(0x7b292952, 0x3ee3e3dd, 0x712f2f5e, 0x97848413);
+	&_data_word(0xf55353a6, 0x68d1d1b9, 0x00000000, 0x2cededc1);
+	&_data_word(0x60202040, 0x1ffcfce3, 0xc8b1b179, 0xed5b5bb6);
+	&_data_word(0xbe6a6ad4, 0x46cbcb8d, 0xd9bebe67, 0x4b393972);
+	&_data_word(0xde4a4a94, 0xd44c4c98, 0xe85858b0, 0x4acfcf85);
+	&_data_word(0x6bd0d0bb, 0x2aefefc5, 0xe5aaaa4f, 0x16fbfbed);
+	&_data_word(0xc5434386, 0xd74d4d9a, 0x55333366, 0x94858511);
+	&_data_word(0xcf45458a, 0x10f9f9e9, 0x06020204, 0x817f7ffe);
+	&_data_word(0xf05050a0, 0x443c3c78, 0xba9f9f25, 0xe3a8a84b);
+	&_data_word(0xf35151a2, 0xfea3a35d, 0xc0404080, 0x8a8f8f05);
+	&_data_word(0xad92923f, 0xbc9d9d21, 0x48383870, 0x04f5f5f1);
+	&_data_word(0xdfbcbc63, 0xc1b6b677, 0x75dadaaf, 0x63212142);
+	&_data_word(0x30101020, 0x1affffe5, 0x0ef3f3fd, 0x6dd2d2bf);
+	&_data_word(0x4ccdcd81, 0x140c0c18, 0x35131326, 0x2fececc3);
+	&_data_word(0xe15f5fbe, 0xa2979735, 0xcc444488, 0x3917172e);
+	&_data_word(0x57c4c493, 0xf2a7a755, 0x827e7efc, 0x473d3d7a);
+	&_data_word(0xac6464c8, 0xe75d5dba, 0x2b191932, 0x957373e6);
+	&_data_word(0xa06060c0, 0x98818119, 0xd14f4f9e, 0x7fdcdca3);
+	&_data_word(0x66222244, 0x7e2a2a54, 0xab90903b, 0x8388880b);
+	&_data_word(0xca46468c, 0x29eeeec7, 0xd3b8b86b, 0x3c141428);
+	&_data_word(0x79dedea7, 0xe25e5ebc, 0x1d0b0b16, 0x76dbdbad);
+	&_data_word(0x3be0e0db, 0x56323264, 0x4e3a3a74, 0x1e0a0a14);
+	&_data_word(0xdb494992, 0x0a06060c, 0x6c242448, 0xe45c5cb8);
+	&_data_word(0x5dc2c29f, 0x6ed3d3bd, 0xefacac43, 0xa66262c4);
+	&_data_word(0xa8919139, 0xa4959531, 0x37e4e4d3, 0x8b7979f2);
+	&_data_word(0x32e7e7d5, 0x43c8c88b, 0x5937376e, 0xb76d6dda);
+	&_data_word(0x8c8d8d01, 0x64d5d5b1, 0xd24e4e9c, 0xe0a9a949);
+	&_data_word(0xb46c6cd8, 0xfa5656ac, 0x07f4f4f3, 0x25eaeacf);
+	&_data_word(0xaf6565ca, 0x8e7a7af4, 0xe9aeae47, 0x18080810);
+	&_data_word(0xd5baba6f, 0x887878f0, 0x6f25254a, 0x722e2e5c);
+	&_data_word(0x241c1c38, 0xf1a6a657, 0xc7b4b473, 0x51c6c697);
+	&_data_word(0x23e8e8cb, 0x7cdddda1, 0x9c7474e8, 0x211f1f3e);
+	&_data_word(0xdd4b4b96, 0xdcbdbd61, 0x868b8b0d, 0x858a8a0f);
+	&_data_word(0x907070e0, 0x423e3e7c, 0xc4b5b571, 0xaa6666cc);
+	&_data_word(0xd8484890, 0x05030306, 0x01f6f6f7, 0x120e0e1c);
+	&_data_word(0xa36161c2, 0x5f35356a, 0xf95757ae, 0xd0b9b969);
+	&_data_word(0x91868617, 0x58c1c199, 0x271d1d3a, 0xb99e9e27);
+	&_data_word(0x38e1e1d9, 0x13f8f8eb, 0xb398982b, 0x33111122);
+	&_data_word(0xbb6969d2, 0x70d9d9a9, 0x898e8e07, 0xa7949433);
+	&_data_word(0xb69b9b2d, 0x221e1e3c, 0x92878715, 0x20e9e9c9);
+	&_data_word(0x49cece87, 0xff5555aa, 0x78282850, 0x7adfdfa5);
+	&_data_word(0x8f8c8c03, 0xf8a1a159, 0x80898909, 0x170d0d1a);
+	&_data_word(0xdabfbf65, 0x31e6e6d7, 0xc6424284, 0xb86868d0);
+	&_data_word(0xc3414182, 0xb0999929, 0x772d2d5a, 0x110f0f1e);
+	&_data_word(0xcbb0b07b, 0xfc5454a8, 0xd6bbbb6d, 0x3a16162c);
+
+#Te4	# four copies of Te4 to choose from to avoid L1 aliasing
+	&data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5);
+	&data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76);
+	&data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0);
+	&data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0);
+	&data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc);
+	&data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15);
+	&data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a);
+	&data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75);
+	&data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0);
+	&data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84);
+	&data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b);
+	&data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf);
+	&data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85);
+	&data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8);
+	&data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5);
+	&data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2);
+	&data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17);
+	&data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73);
+	&data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88);
+	&data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb);
+	&data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c);
+	&data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79);
+	&data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9);
+	&data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08);
+	&data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6);
+	&data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a);
+	&data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e);
+	&data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e);
+	&data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94);
+	&data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf);
+	&data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68);
+	&data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16);
+
+	&data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5);
+	&data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76);
+	&data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0);
+	&data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0);
+	&data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc);
+	&data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15);
+	&data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a);
+	&data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75);
+	&data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0);
+	&data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84);
+	&data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b);
+	&data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf);
+	&data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85);
+	&data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8);
+	&data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5);
+	&data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2);
+	&data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17);
+	&data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73);
+	&data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88);
+	&data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb);
+	&data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c);
+	&data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79);
+	&data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9);
+	&data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08);
+	&data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6);
+	&data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a);
+	&data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e);
+	&data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e);
+	&data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94);
+	&data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf);
+	&data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68);
+	&data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16);
+
+	&data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5);
+	&data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76);
+	&data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0);
+	&data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0);
+	&data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc);
+	&data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15);
+	&data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a);
+	&data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75);
+	&data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0);
+	&data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84);
+	&data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b);
+	&data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf);
+	&data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85);
+	&data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8);
+	&data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5);
+	&data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2);
+	&data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17);
+	&data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73);
+	&data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88);
+	&data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb);
+	&data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c);
+	&data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79);
+	&data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9);
+	&data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08);
+	&data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6);
+	&data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a);
+	&data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e);
+	&data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e);
+	&data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94);
+	&data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf);
+	&data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68);
+	&data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16);
+
+	&data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5);
+	&data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76);
+	&data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0);
+	&data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0);
+	&data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc);
+	&data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15);
+	&data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a);
+	&data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75);
+	&data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0);
+	&data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84);
+	&data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b);
+	&data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf);
+	&data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85);
+	&data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8);
+	&data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5);
+	&data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2);
+	&data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17);
+	&data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73);
+	&data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88);
+	&data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb);
+	&data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c);
+	&data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79);
+	&data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9);
+	&data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08);
+	&data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6);
+	&data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a);
+	&data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e);
+	&data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e);
+	&data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94);
+	&data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf);
+	&data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68);
+	&data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16);
+#rcon:
+	&data_word(0x00000001, 0x00000002, 0x00000004, 0x00000008);
+	&data_word(0x00000010, 0x00000020, 0x00000040, 0x00000080);
+	&data_word(0x0000001b, 0x00000036, 0x00000000, 0x00000000);
+	&data_word(0x00000000, 0x00000000, 0x00000000, 0x00000000);
+&function_end_B("_x86_AES_encrypt");
+
+# void AES_encrypt (const void *inp,void *out,const AES_KEY *key);
+&function_begin("AES_encrypt");
+	&mov	($acc,&wparam(0));		# load inp
+	&mov	($key,&wparam(2));		# load key
+
+	&mov	($s0,"esp");
+	&sub	("esp",36);
+	&and	("esp",-64);			# align to cache-line
+
+	# place stack frame just "above" the key schedule
+	&lea	($s1,&DWP(-64-63,$key));
+	&sub	($s1,"esp");
+	&neg	($s1);
+	&and	($s1,0x3C0);	# modulo 1024, but aligned to cache-line
+	&sub	("esp",$s1);
+	&add	("esp",4);	# 4 is reserved for caller's return address
+	&mov	($_esp,$s0);			# save stack pointer
+
+	&call   (&label("pic_point"));          # make it PIC!
+	&set_label("pic_point");
+	&blindpop($tbl);
+	&picmeup($s0,"OPENSSL_ia32cap_P",$tbl,&label("pic_point")) if (!$x86only);
+	&lea    ($tbl,&DWP(&label("AES_Te")."-".&label("pic_point"),$tbl));
+
+	# pick Te4 copy which can't "overlap" with stack frame or key schedule
+	&lea	($s1,&DWP(768-4,"esp"));
+	&sub	($s1,$tbl);
+	&and	($s1,0x300);
+	&lea	($tbl,&DWP(2048+128,$tbl,$s1));
+
+					if (!$x86only) {
+	&bt	(&DWP(0,$s0),25);	# check for SSE bit
+	&jnc	(&label("x86"));
+
+	&movq	("mm0",&QWP(0,$acc));
+	&movq	("mm4",&QWP(8,$acc));
+	&call	("_sse_AES_encrypt_compact");
+	&mov	("esp",$_esp);			# restore stack pointer
+	&mov	($acc,&wparam(1));		# load out
+	&movq	(&QWP(0,$acc),"mm0");		# write output data
+	&movq	(&QWP(8,$acc),"mm4");
+	&emms	();
+	&function_end_A();
+					}
+	&set_label("x86",16);
+	&mov	($_tbl,$tbl);
+	&mov	($s0,&DWP(0,$acc));		# load input data
+	&mov	($s1,&DWP(4,$acc));
+	&mov	($s2,&DWP(8,$acc));
+	&mov	($s3,&DWP(12,$acc));
+	&call	("_x86_AES_encrypt_compact");
+	&mov	("esp",$_esp);			# restore stack pointer
+	&mov	($acc,&wparam(1));		# load out
+	&mov	(&DWP(0,$acc),$s0);		# write output data
+	&mov	(&DWP(4,$acc),$s1);
+	&mov	(&DWP(8,$acc),$s2);
+	&mov	(&DWP(12,$acc),$s3);
+&function_end("AES_encrypt");
+
+#--------------------------------------------------------------------#
+
+######################################################################
+# "Compact" block function
+######################################################################
+
+sub deccompact()
+{ my $Fn = mov;
+  while ($#_>5) { pop(@_); $Fn=sub{}; }
+  my ($i,$td,@s)=@_;
+  my $tmp = $key;
+  my $out = $i==3?$s[0]:$acc;
+
+	# $Fn is used in first compact round and its purpose is to
+	# void restoration of some values from stack, so that after
+	# 4xdeccompact with extra argument $key, $s0 and $s1 values
+	# are left there...
+	if($i==3)   {	&$Fn	($key,$__key);			}
+	else        {	&mov	($out,$s[0]);			}
+			&and	($out,0xFF);
+			&movz	($out,&BP(-128,$td,$out,1));
+
+	if ($i==3)  {	$tmp=$s[1];				}
+			&movz	($tmp,&HB($s[1]));
+			&movz	($tmp,&BP(-128,$td,$tmp,1));
+			&shl	($tmp,8);
+			&xor	($out,$tmp);
+
+	if ($i==3)  {	$tmp=$s[2]; &mov ($s[1],$acc);		}
+	else        {	mov	($tmp,$s[2]);			}
+			&shr	($tmp,16);
+			&and	($tmp,0xFF);
+			&movz	($tmp,&BP(-128,$td,$tmp,1));
+			&shl	($tmp,16);
+			&xor	($out,$tmp);
+
+	if ($i==3)  {	$tmp=$s[3]; &$Fn ($s[2],$__s1);		}
+	else        {	&mov	($tmp,$s[3]);			}
+			&shr	($tmp,24);
+			&movz	($tmp,&BP(-128,$td,$tmp,1));
+			&shl	($tmp,24);
+			&xor	($out,$tmp);
+	if ($i<2)   {	&mov	(&DWP(4+4*$i,"esp"),$out);	}
+	if ($i==3)  {	&$Fn	($s[3],$__s0);			}
+}
+
+# must be called with 2,3,0,1 as argument sequence!!!
+sub dectransform()
+{ my @s = ($s0,$s1,$s2,$s3);
+  my $i = shift;
+  my $tmp = $key;
+  my $tp2 = @s[($i+2)%4]; $tp2 = @s[2] if ($i==1);
+  my $tp4 = @s[($i+3)%4]; $tp4 = @s[3] if ($i==1);
+  my $tp8 = $tbl;
+
+	&mov	($acc,$s[$i]);
+	&and	($acc,0x80808080);
+	&mov	($tmp,$acc);
+	&shr	($tmp,7);
+	&lea	($tp2,&DWP(0,$s[$i],$s[$i]));
+	&sub	($acc,$tmp);
+	&and	($tp2,0xfefefefe);
+	&and	($acc,0x1b1b1b1b);
+	&xor	($acc,$tp2);
+	&mov	($tp2,$acc);
+
+	&and	($acc,0x80808080);
+	&mov	($tmp,$acc);
+	&shr	($tmp,7);
+	&lea	($tp4,&DWP(0,$tp2,$tp2));
+	&sub	($acc,$tmp);
+	&and	($tp4,0xfefefefe);
+	&and	($acc,0x1b1b1b1b);
+	 &xor	($tp2,$s[$i]);	# tp2^tp1
+	&xor	($acc,$tp4);
+	&mov	($tp4,$acc);
+
+	&and	($acc,0x80808080);
+	&mov	($tmp,$acc);
+	&shr	($tmp,7);
+	&lea	($tp8,&DWP(0,$tp4,$tp4));
+	&sub	($acc,$tmp);
+	&and	($tp8,0xfefefefe);
+	&and	($acc,0x1b1b1b1b);
+	 &xor	($tp4,$s[$i]);	# tp4^tp1
+	 &rotl	($s[$i],8);	# = ROTATE(tp1,8)
+	&xor	($tp8,$acc);
+
+	&xor	($s[$i],$tp2);
+	&xor	($tp2,$tp8);
+	&rotl	($tp2,24);
+	&xor	($s[$i],$tp4);
+	&xor	($tp4,$tp8);
+	&rotl	($tp4,16);
+	&xor	($s[$i],$tp8);	# ^= tp8^(tp4^tp1)^(tp2^tp1)
+	&rotl	($tp8,8);
+	&xor	($s[$i],$tp2);	# ^= ROTATE(tp8^tp2^tp1,24)
+	&xor	($s[$i],$tp4);	# ^= ROTATE(tp8^tp4^tp1,16)
+	 &mov	($s[0],$__s0)			if($i==2); #prefetch $s0
+	 &mov	($s[1],$__s1)			if($i==3); #prefetch $s1
+	 &mov	($s[2],$__s2)			if($i==1);
+	&xor	($s[$i],$tp8);	# ^= ROTATE(tp8,8)
+
+	&mov	($s[3],$__s3)			if($i==1);
+	&mov	(&DWP(4+4*$i,"esp"),$s[$i])	if($i>=2);
+}
+
+&function_begin_B("_x86_AES_decrypt_compact");
+	# note that caller is expected to allocate stack frame for me!
+	&mov	($__key,$key);			# save key
+
+	&xor	($s0,&DWP(0,$key));		# xor with key
+	&xor	($s1,&DWP(4,$key));
+	&xor	($s2,&DWP(8,$key));
+	&xor	($s3,&DWP(12,$key));
+
+	&mov	($acc,&DWP(240,$key));		# load key->rounds
+
+	&lea	($acc,&DWP(-2,$acc,$acc));
+	&lea	($acc,&DWP(0,$key,$acc,8));
+	&mov	($__end,$acc);			# end of key schedule
+
+	# prefetch Td4
+	&mov	($key,&DWP(0-128,$tbl));
+	&mov	($acc,&DWP(32-128,$tbl));
+	&mov	($key,&DWP(64-128,$tbl));
+	&mov	($acc,&DWP(96-128,$tbl));
+	&mov	($key,&DWP(128-128,$tbl));
+	&mov	($acc,&DWP(160-128,$tbl));
+	&mov	($key,&DWP(192-128,$tbl));
+	&mov	($acc,&DWP(224-128,$tbl));
+
+	&set_label("loop",16);
+
+		&deccompact(0,$tbl,$s0,$s3,$s2,$s1,1);
+		&deccompact(1,$tbl,$s1,$s0,$s3,$s2,1);
+		&deccompact(2,$tbl,$s2,$s1,$s0,$s3,1);
+		&deccompact(3,$tbl,$s3,$s2,$s1,$s0,1);
+		&dectransform(2);
+		&dectransform(3);
+		&dectransform(0);
+		&dectransform(1);
+		&mov 	($key,$__key);
+		&mov	($tbl,$__tbl);
+		&add	($key,16);		# advance rd_key
+		&xor	($s0,&DWP(0,$key));
+		&xor	($s1,&DWP(4,$key));
+		&xor	($s2,&DWP(8,$key));
+		&xor	($s3,&DWP(12,$key));
+
+	&cmp	($key,$__end);
+	&mov	($__key,$key);
+	&jb	(&label("loop"));
+
+	&deccompact(0,$tbl,$s0,$s3,$s2,$s1);
+	&deccompact(1,$tbl,$s1,$s0,$s3,$s2);
+	&deccompact(2,$tbl,$s2,$s1,$s0,$s3);
+	&deccompact(3,$tbl,$s3,$s2,$s1,$s0);
+
+	&xor	($s0,&DWP(16,$key));
+	&xor	($s1,&DWP(20,$key));
+	&xor	($s2,&DWP(24,$key));
+	&xor	($s3,&DWP(28,$key));
+
+	&ret	();
+&function_end_B("_x86_AES_decrypt_compact");
+
+######################################################################
+# "Compact" SSE block function.
+######################################################################
+
+sub sse_deccompact()
+{
+	&pshufw	("mm1","mm0",0x0c);		#  7, 6, 1, 0
+	&movd	("eax","mm1");			#  7, 6, 1, 0
+
+	&pshufw	("mm5","mm4",0x09);		# 13,12,11,10
+	&movz	($acc,&LB("eax"));		#  0
+	&movz	("ecx",&BP(-128,$tbl,$acc,1));	#  0
+	&movd	("ebx","mm5");			# 13,12,11,10
+	&movz	("edx",&HB("eax"));		#  1
+	&movz	("edx",&BP(-128,$tbl,"edx",1));	#  1
+	&shl	("edx",8);			#  1
+
+	&pshufw	("mm2","mm0",0x06);		#  3, 2, 5, 4
+	&movz	($acc,&LB("ebx"));		# 10
+	&movz	($acc,&BP(-128,$tbl,$acc,1));	# 10
+	&shl	($acc,16);			# 10
+	&or	("ecx",$acc);			# 10
+	&shr	("eax",16);			#  7, 6
+	&movz	($acc,&HB("ebx"));		# 11
+	&movz	($acc,&BP(-128,$tbl,$acc,1));	# 11
+	&shl	($acc,24);			# 11
+	&or	("edx",$acc);			# 11
+	&shr	("ebx",16);			# 13,12
+
+	&pshufw	("mm6","mm4",0x03);		# 9, 8,15,14
+	&movz	($acc,&HB("eax"));		#  7
+	&movz	($acc,&BP(-128,$tbl,$acc,1));	#  7
+	&shl	($acc,24);			#  7
+	&or	("ecx",$acc);			#  7
+	&movz	($acc,&HB("ebx"));		# 13
+	&movz	($acc,&BP(-128,$tbl,$acc,1));	# 13
+	&shl	($acc,8);			# 13
+	&or	("ecx",$acc);			# 13
+	&movd	("mm0","ecx");			# t[0] collected
+
+	&movz	($acc,&LB("eax"));		#  6
+	&movd	("eax","mm2");			#  3, 2, 5, 4
+	&movz	("ecx",&BP(-128,$tbl,$acc,1));	#  6
+	&shl	("ecx",16);			#  6
+	&movz	($acc,&LB("ebx"));		# 12
+	&movd	("ebx","mm6");			#  9, 8,15,14
+	&movz	($acc,&BP(-128,$tbl,$acc,1));	# 12
+	&or	("ecx",$acc);			# 12
+
+	&movz	($acc,&LB("eax"));		#  4
+	&movz	($acc,&BP(-128,$tbl,$acc,1));	#  4
+	&or	("edx",$acc);			#  4
+	&movz	($acc,&LB("ebx"));		# 14
+	&movz	($acc,&BP(-128,$tbl,$acc,1));	# 14
+	&shl	($acc,16);			# 14
+	&or	("edx",$acc);			# 14
+	&movd	("mm1","edx");			# t[1] collected
+
+	&movz	($acc,&HB("eax"));		#  5
+	&movz	("edx",&BP(-128,$tbl,$acc,1));	#  5
+	&shl	("edx",8);			#  5
+	&movz	($acc,&HB("ebx"));		# 15
+	&shr	("eax",16);			#  3, 2
+	&movz	($acc,&BP(-128,$tbl,$acc,1));	# 15
+	&shl	($acc,24);			# 15
+	&or	("edx",$acc);			# 15
+	&shr	("ebx",16);			#  9, 8
+
+	&punpckldq	("mm0","mm1");		# t[0,1] collected
+
+	&movz	($acc,&HB("ebx"));		#  9
+	&movz	($acc,&BP(-128,$tbl,$acc,1));	#  9
+	&shl	($acc,8);			#  9
+	&or	("ecx",$acc);			#  9
+	&and	("ebx",0xff);			#  8
+	&movz	("ebx",&BP(-128,$tbl,"ebx",1));	#  8
+	&or	("edx","ebx");			#  8
+	&movz	($acc,&LB("eax"));		#  2
+	&movz	($acc,&BP(-128,$tbl,$acc,1));	#  2
+	&shl	($acc,16);			#  2
+	&or	("edx",$acc);			#  2
+	&movd	("mm4","edx");			# t[2] collected
+	&movz	("eax",&HB("eax"));		#  3
+	&movz	("eax",&BP(-128,$tbl,"eax",1));	#  3
+	&shl	("eax",24);			#  3
+	&or	("ecx","eax");			#  3
+	&movd	("mm5","ecx");			# t[3] collected
+
+	&punpckldq	("mm4","mm5");		# t[2,3] collected
+}
+
+					if (!$x86only) {
+&function_begin_B("_sse_AES_decrypt_compact");
+	&pxor	("mm0",&QWP(0,$key));	#  7, 6, 5, 4, 3, 2, 1, 0
+	&pxor	("mm4",&QWP(8,$key));	# 15,14,13,12,11,10, 9, 8
+
+	# note that caller is expected to allocate stack frame for me!
+	&mov	($acc,&DWP(240,$key));		# load key->rounds
+	&lea	($acc,&DWP(-2,$acc,$acc));
+	&lea	($acc,&DWP(0,$key,$acc,8));
+	&mov	($__end,$acc);			# end of key schedule
+
+	&mov	($s0,0x1b1b1b1b);		# magic constant
+	&mov	(&DWP(8,"esp"),$s0);
+	&mov	(&DWP(12,"esp"),$s0);
+
+	# prefetch Td4
+	&mov	($s0,&DWP(0-128,$tbl));
+	&mov	($s1,&DWP(32-128,$tbl));
+	&mov	($s2,&DWP(64-128,$tbl));
+	&mov	($s3,&DWP(96-128,$tbl));
+	&mov	($s0,&DWP(128-128,$tbl));
+	&mov	($s1,&DWP(160-128,$tbl));
+	&mov	($s2,&DWP(192-128,$tbl));
+	&mov	($s3,&DWP(224-128,$tbl));
+
+	&set_label("loop",16);
+		&sse_deccompact();
+		&add	($key,16);
+		&cmp	($key,$__end);
+		&ja	(&label("out"));
+
+		# ROTATE(x^y,N) == ROTATE(x,N)^ROTATE(y,N)
+		&movq	("mm3","mm0");		&movq	("mm7","mm4");
+		&movq	("mm2","mm0",1);	&movq	("mm6","mm4",1);
+		&movq	("mm1","mm0");		&movq	("mm5","mm4");
+		&pshufw	("mm0","mm0",0xb1);	&pshufw	("mm4","mm4",0xb1);# = ROTATE(tp0,16)
+		&pslld	("mm2",8);		&pslld	("mm6",8);
+		&psrld	("mm3",8);		&psrld	("mm7",8);
+		&pxor	("mm0","mm2");		&pxor	("mm4","mm6");	# ^= tp0<<8
+		&pxor	("mm0","mm3");		&pxor	("mm4","mm7");	# ^= tp0>>8
+		&pslld	("mm2",16);		&pslld	("mm6",16);
+		&psrld	("mm3",16);		&psrld	("mm7",16);
+		&pxor	("mm0","mm2");		&pxor	("mm4","mm6");	# ^= tp0<<24
+		&pxor	("mm0","mm3");		&pxor	("mm4","mm7");	# ^= tp0>>24
+
+		&movq	("mm3",&QWP(8,"esp"));
+		&pxor	("mm2","mm2");		&pxor	("mm6","mm6");
+		&pcmpgtb("mm2","mm1");		&pcmpgtb("mm6","mm5");
+		&pand	("mm2","mm3");		&pand	("mm6","mm3");
+		&paddb	("mm1","mm1");		&paddb	("mm5","mm5");
+		&pxor	("mm1","mm2");		&pxor	("mm5","mm6");	# tp2
+		&movq	("mm3","mm1");		&movq	("mm7","mm5");
+		&movq	("mm2","mm1");		&movq	("mm6","mm5");
+		&pxor	("mm0","mm1");		&pxor	("mm4","mm5");	# ^= tp2
+		&pslld	("mm3",24);		&pslld	("mm7",24);
+		&psrld	("mm2",8);		&psrld	("mm6",8);
+		&pxor	("mm0","mm3");		&pxor	("mm4","mm7");	# ^= tp2<<24
+		&pxor	("mm0","mm2");		&pxor	("mm4","mm6");	# ^= tp2>>8
+
+		&movq	("mm2",&QWP(8,"esp"));
+		&pxor	("mm3","mm3");		&pxor	("mm7","mm7");
+		&pcmpgtb("mm3","mm1");		&pcmpgtb("mm7","mm5");
+		&pand	("mm3","mm2");		&pand	("mm7","mm2");
+		&paddb	("mm1","mm1");		&paddb	("mm5","mm5");
+		&pxor	("mm1","mm3");		&pxor	("mm5","mm7");	# tp4
+		&pshufw	("mm3","mm1",0xb1);	&pshufw	("mm7","mm5",0xb1);
+		&pxor	("mm0","mm1");		&pxor	("mm4","mm5");	# ^= tp4
+		&pxor	("mm0","mm3");		&pxor	("mm4","mm7");	# ^= ROTATE(tp4,16)	
+
+		&pxor	("mm3","mm3");		&pxor	("mm7","mm7");
+		&pcmpgtb("mm3","mm1");		&pcmpgtb("mm7","mm5");
+		&pand	("mm3","mm2");		&pand	("mm7","mm2");
+		&paddb	("mm1","mm1");		&paddb	("mm5","mm5");
+		&pxor	("mm1","mm3");		&pxor	("mm5","mm7");	# tp8
+		&pxor	("mm0","mm1");		&pxor	("mm4","mm5");	# ^= tp8
+		&movq	("mm3","mm1");		&movq	("mm7","mm5");
+		&pshufw	("mm2","mm1",0xb1);	&pshufw	("mm6","mm5",0xb1);
+		&pxor	("mm0","mm2");		&pxor	("mm4","mm6");	# ^= ROTATE(tp8,16)
+		&pslld	("mm1",8);		&pslld	("mm5",8);
+		&psrld	("mm3",8);		&psrld	("mm7",8);
+		&movq	("mm2",&QWP(0,$key));	&movq	("mm6",&QWP(8,$key));
+		&pxor	("mm0","mm1");		&pxor	("mm4","mm5");	# ^= tp8<<8
+		&pxor	("mm0","mm3");		&pxor	("mm4","mm7");	# ^= tp8>>8
+		&mov	($s0,&DWP(0-128,$tbl));
+		&pslld	("mm1",16);		&pslld	("mm5",16);
+		&mov	($s1,&DWP(64-128,$tbl));
+		&psrld	("mm3",16);		&psrld	("mm7",16);
+		&mov	($s2,&DWP(128-128,$tbl));
+		&pxor	("mm0","mm1");		&pxor	("mm4","mm5");	# ^= tp8<<24
+		&mov	($s3,&DWP(192-128,$tbl));
+		&pxor	("mm0","mm3");		&pxor	("mm4","mm7");	# ^= tp8>>24
+
+		&pxor	("mm0","mm2");		&pxor	("mm4","mm6");
+	&jmp	(&label("loop"));
+
+	&set_label("out",16);
+	&pxor	("mm0",&QWP(0,$key));
+	&pxor	("mm4",&QWP(8,$key));
+
+	&ret	();
+&function_end_B("_sse_AES_decrypt_compact");
+					}
+
+######################################################################
+# Vanilla block function.
+######################################################################
+
+sub decstep()
+{ my ($i,$td,@s) = @_;
+  my $tmp = $key;
+  my $out = $i==3?$s[0]:$acc;
+
+	# no instructions are reordered, as performance appears
+	# optimal... or rather that all attempts to reorder didn't
+	# result in better performance [which by the way is not a
+	# bit lower than ecryption].
+	if($i==3)   {	&mov	($key,$__key);			}
+	else        {	&mov	($out,$s[0]);			}
+			&and	($out,0xFF);
+			&mov	($out,&DWP(0,$td,$out,8));
+
+	if ($i==3)  {	$tmp=$s[1];				}
+			&movz	($tmp,&HB($s[1]));
+			&xor	($out,&DWP(3,$td,$tmp,8));
+
+	if ($i==3)  {	$tmp=$s[2]; &mov ($s[1],$acc);		}
+	else        {	&mov	($tmp,$s[2]);			}
+			&shr	($tmp,16);
+			&and	($tmp,0xFF);
+			&xor	($out,&DWP(2,$td,$tmp,8));
+
+	if ($i==3)  {	$tmp=$s[3]; &mov ($s[2],$__s1);		}
+	else        {	&mov	($tmp,$s[3]);			}
+			&shr	($tmp,24);
+			&xor	($out,&DWP(1,$td,$tmp,8));
+	if ($i<2)   {	&mov	(&DWP(4+4*$i,"esp"),$out);	}
+	if ($i==3)  {	&mov	($s[3],$__s0);			}
+			&comment();
+}
+
+sub declast()
+{ my ($i,$td,@s)=@_;
+  my $tmp = $key;
+  my $out = $i==3?$s[0]:$acc;
+
+	if($i==0)   {	&lea	($td,&DWP(2048+128,$td));
+			&mov	($tmp,&DWP(0-128,$td));
+			&mov	($acc,&DWP(32-128,$td));
+			&mov	($tmp,&DWP(64-128,$td));
+			&mov	($acc,&DWP(96-128,$td));
+			&mov	($tmp,&DWP(128-128,$td));
+			&mov	($acc,&DWP(160-128,$td));
+			&mov	($tmp,&DWP(192-128,$td));
+			&mov	($acc,&DWP(224-128,$td));
+			&lea	($td,&DWP(-128,$td));		}
+	if($i==3)   {	&mov	($key,$__key);			}
+	else        {	&mov	($out,$s[0]);			}
+			&and	($out,0xFF);
+			&movz	($out,&BP(0,$td,$out,1));
+
+	if ($i==3)  {	$tmp=$s[1];				}
+			&movz	($tmp,&HB($s[1]));
+			&movz	($tmp,&BP(0,$td,$tmp,1));
+			&shl	($tmp,8);
+			&xor	($out,$tmp);
+
+	if ($i==3)  {	$tmp=$s[2]; &mov ($s[1],$acc);		}
+	else        {	mov	($tmp,$s[2]);			}
+			&shr	($tmp,16);
+			&and	($tmp,0xFF);
+			&movz	($tmp,&BP(0,$td,$tmp,1));
+			&shl	($tmp,16);
+			&xor	($out,$tmp);
+
+	if ($i==3)  {	$tmp=$s[3]; &mov ($s[2],$__s1);		}
+	else        {	&mov	($tmp,$s[3]);			}
+			&shr	($tmp,24);
+			&movz	($tmp,&BP(0,$td,$tmp,1));
+			&shl	($tmp,24);
+			&xor	($out,$tmp);
+	if ($i<2)   {	&mov	(&DWP(4+4*$i,"esp"),$out);	}
+	if ($i==3)  {	&mov	($s[3],$__s0);
+			&lea	($td,&DWP(-2048,$td));		}
+}
+
+&function_begin_B("_x86_AES_decrypt");
+	# note that caller is expected to allocate stack frame for me!
+	&mov	($__key,$key);			# save key
+
+	&xor	($s0,&DWP(0,$key));		# xor with key
+	&xor	($s1,&DWP(4,$key));
+	&xor	($s2,&DWP(8,$key));
+	&xor	($s3,&DWP(12,$key));
+
+	&mov	($acc,&DWP(240,$key));		# load key->rounds
+
+	if ($small_footprint) {
+	    &lea	($acc,&DWP(-2,$acc,$acc));
+	    &lea	($acc,&DWP(0,$key,$acc,8));
+	    &mov	($__end,$acc);		# end of key schedule
+	    &set_label("loop",16);
+		&decstep(0,$tbl,$s0,$s3,$s2,$s1);
+		&decstep(1,$tbl,$s1,$s0,$s3,$s2);
+		&decstep(2,$tbl,$s2,$s1,$s0,$s3);
+		&decstep(3,$tbl,$s3,$s2,$s1,$s0);
+		&add	($key,16);		# advance rd_key
+		&xor	($s0,&DWP(0,$key));
+		&xor	($s1,&DWP(4,$key));
+		&xor	($s2,&DWP(8,$key));
+		&xor	($s3,&DWP(12,$key));
+	    &cmp	($key,$__end);
+	    &mov	($__key,$key);
+	    &jb		(&label("loop"));
+	}
+	else {
+	    &cmp	($acc,10);
+	    &jle	(&label("10rounds"));
+	    &cmp	($acc,12);
+	    &jle	(&label("12rounds"));
+
+	&set_label("14rounds",4);
+	    for ($i=1;$i<3;$i++) {
+		&decstep(0,$tbl,$s0,$s3,$s2,$s1);
+		&decstep(1,$tbl,$s1,$s0,$s3,$s2);
+		&decstep(2,$tbl,$s2,$s1,$s0,$s3);
+		&decstep(3,$tbl,$s3,$s2,$s1,$s0);
+		&xor	($s0,&DWP(16*$i+0,$key));
+		&xor	($s1,&DWP(16*$i+4,$key));
+		&xor	($s2,&DWP(16*$i+8,$key));
+		&xor	($s3,&DWP(16*$i+12,$key));
+	    }
+	    &add	($key,32);
+	    &mov	($__key,$key);		# advance rd_key
+	&set_label("12rounds",4);
+	    for ($i=1;$i<3;$i++) {
+		&decstep(0,$tbl,$s0,$s3,$s2,$s1);
+		&decstep(1,$tbl,$s1,$s0,$s3,$s2);
+		&decstep(2,$tbl,$s2,$s1,$s0,$s3);
+		&decstep(3,$tbl,$s3,$s2,$s1,$s0);
+		&xor	($s0,&DWP(16*$i+0,$key));
+		&xor	($s1,&DWP(16*$i+4,$key));
+		&xor	($s2,&DWP(16*$i+8,$key));
+		&xor	($s3,&DWP(16*$i+12,$key));
+	    }
+	    &add	($key,32);
+	    &mov	($__key,$key);		# advance rd_key
+	&set_label("10rounds",4);
+	    for ($i=1;$i<10;$i++) {
+		&decstep(0,$tbl,$s0,$s3,$s2,$s1);
+		&decstep(1,$tbl,$s1,$s0,$s3,$s2);
+		&decstep(2,$tbl,$s2,$s1,$s0,$s3);
+		&decstep(3,$tbl,$s3,$s2,$s1,$s0);
+		&xor	($s0,&DWP(16*$i+0,$key));
+		&xor	($s1,&DWP(16*$i+4,$key));
+		&xor	($s2,&DWP(16*$i+8,$key));
+		&xor	($s3,&DWP(16*$i+12,$key));
+	    }
+	}
+
+	&declast(0,$tbl,$s0,$s3,$s2,$s1);
+	&declast(1,$tbl,$s1,$s0,$s3,$s2);
+	&declast(2,$tbl,$s2,$s1,$s0,$s3);
+	&declast(3,$tbl,$s3,$s2,$s1,$s0);
+
+	&add	($key,$small_footprint?16:160);
+	&xor	($s0,&DWP(0,$key));
+	&xor	($s1,&DWP(4,$key));
+	&xor	($s2,&DWP(8,$key));
+	&xor	($s3,&DWP(12,$key));
+
+	&ret	();
+
+&set_label("AES_Td",64);	# Yes! I keep it in the code segment!
+	&_data_word(0x50a7f451, 0x5365417e, 0xc3a4171a, 0x965e273a);
+	&_data_word(0xcb6bab3b, 0xf1459d1f, 0xab58faac, 0x9303e34b);
+	&_data_word(0x55fa3020, 0xf66d76ad, 0x9176cc88, 0x254c02f5);
+	&_data_word(0xfcd7e54f, 0xd7cb2ac5, 0x80443526, 0x8fa362b5);
+	&_data_word(0x495ab1de, 0x671bba25, 0x980eea45, 0xe1c0fe5d);
+	&_data_word(0x02752fc3, 0x12f04c81, 0xa397468d, 0xc6f9d36b);
+	&_data_word(0xe75f8f03, 0x959c9215, 0xeb7a6dbf, 0xda595295);
+	&_data_word(0x2d83bed4, 0xd3217458, 0x2969e049, 0x44c8c98e);
+	&_data_word(0x6a89c275, 0x78798ef4, 0x6b3e5899, 0xdd71b927);
+	&_data_word(0xb64fe1be, 0x17ad88f0, 0x66ac20c9, 0xb43ace7d);
+	&_data_word(0x184adf63, 0x82311ae5, 0x60335197, 0x457f5362);
+	&_data_word(0xe07764b1, 0x84ae6bbb, 0x1ca081fe, 0x942b08f9);
+	&_data_word(0x58684870, 0x19fd458f, 0x876cde94, 0xb7f87b52);
+	&_data_word(0x23d373ab, 0xe2024b72, 0x578f1fe3, 0x2aab5566);
+	&_data_word(0x0728ebb2, 0x03c2b52f, 0x9a7bc586, 0xa50837d3);
+	&_data_word(0xf2872830, 0xb2a5bf23, 0xba6a0302, 0x5c8216ed);
+	&_data_word(0x2b1ccf8a, 0x92b479a7, 0xf0f207f3, 0xa1e2694e);
+	&_data_word(0xcdf4da65, 0xd5be0506, 0x1f6234d1, 0x8afea6c4);
+	&_data_word(0x9d532e34, 0xa055f3a2, 0x32e18a05, 0x75ebf6a4);
+	&_data_word(0x39ec830b, 0xaaef6040, 0x069f715e, 0x51106ebd);
+	&_data_word(0xf98a213e, 0x3d06dd96, 0xae053edd, 0x46bde64d);
+	&_data_word(0xb58d5491, 0x055dc471, 0x6fd40604, 0xff155060);
+	&_data_word(0x24fb9819, 0x97e9bdd6, 0xcc434089, 0x779ed967);
+	&_data_word(0xbd42e8b0, 0x888b8907, 0x385b19e7, 0xdbeec879);
+	&_data_word(0x470a7ca1, 0xe90f427c, 0xc91e84f8, 0x00000000);
+	&_data_word(0x83868009, 0x48ed2b32, 0xac70111e, 0x4e725a6c);
+	&_data_word(0xfbff0efd, 0x5638850f, 0x1ed5ae3d, 0x27392d36);
+	&_data_word(0x64d90f0a, 0x21a65c68, 0xd1545b9b, 0x3a2e3624);
+	&_data_word(0xb1670a0c, 0x0fe75793, 0xd296eeb4, 0x9e919b1b);
+	&_data_word(0x4fc5c080, 0xa220dc61, 0x694b775a, 0x161a121c);
+	&_data_word(0x0aba93e2, 0xe52aa0c0, 0x43e0223c, 0x1d171b12);
+	&_data_word(0x0b0d090e, 0xadc78bf2, 0xb9a8b62d, 0xc8a91e14);
+	&_data_word(0x8519f157, 0x4c0775af, 0xbbdd99ee, 0xfd607fa3);
+	&_data_word(0x9f2601f7, 0xbcf5725c, 0xc53b6644, 0x347efb5b);
+	&_data_word(0x7629438b, 0xdcc623cb, 0x68fcedb6, 0x63f1e4b8);
+	&_data_word(0xcadc31d7, 0x10856342, 0x40229713, 0x2011c684);
+	&_data_word(0x7d244a85, 0xf83dbbd2, 0x1132f9ae, 0x6da129c7);
+	&_data_word(0x4b2f9e1d, 0xf330b2dc, 0xec52860d, 0xd0e3c177);
+	&_data_word(0x6c16b32b, 0x99b970a9, 0xfa489411, 0x2264e947);
+	&_data_word(0xc48cfca8, 0x1a3ff0a0, 0xd82c7d56, 0xef903322);
+	&_data_word(0xc74e4987, 0xc1d138d9, 0xfea2ca8c, 0x360bd498);
+	&_data_word(0xcf81f5a6, 0x28de7aa5, 0x268eb7da, 0xa4bfad3f);
+	&_data_word(0xe49d3a2c, 0x0d927850, 0x9bcc5f6a, 0x62467e54);
+	&_data_word(0xc2138df6, 0xe8b8d890, 0x5ef7392e, 0xf5afc382);
+	&_data_word(0xbe805d9f, 0x7c93d069, 0xa92dd56f, 0xb31225cf);
+	&_data_word(0x3b99acc8, 0xa77d1810, 0x6e639ce8, 0x7bbb3bdb);
+	&_data_word(0x097826cd, 0xf418596e, 0x01b79aec, 0xa89a4f83);
+	&_data_word(0x656e95e6, 0x7ee6ffaa, 0x08cfbc21, 0xe6e815ef);
+	&_data_word(0xd99be7ba, 0xce366f4a, 0xd4099fea, 0xd67cb029);
+	&_data_word(0xafb2a431, 0x31233f2a, 0x3094a5c6, 0xc066a235);
+	&_data_word(0x37bc4e74, 0xa6ca82fc, 0xb0d090e0, 0x15d8a733);
+	&_data_word(0x4a9804f1, 0xf7daec41, 0x0e50cd7f, 0x2ff69117);
+	&_data_word(0x8dd64d76, 0x4db0ef43, 0x544daacc, 0xdf0496e4);
+	&_data_word(0xe3b5d19e, 0x1b886a4c, 0xb81f2cc1, 0x7f516546);
+	&_data_word(0x04ea5e9d, 0x5d358c01, 0x737487fa, 0x2e410bfb);
+	&_data_word(0x5a1d67b3, 0x52d2db92, 0x335610e9, 0x1347d66d);
+	&_data_word(0x8c61d79a, 0x7a0ca137, 0x8e14f859, 0x893c13eb);
+	&_data_word(0xee27a9ce, 0x35c961b7, 0xede51ce1, 0x3cb1477a);
+	&_data_word(0x59dfd29c, 0x3f73f255, 0x79ce1418, 0xbf37c773);
+	&_data_word(0xeacdf753, 0x5baafd5f, 0x146f3ddf, 0x86db4478);
+	&_data_word(0x81f3afca, 0x3ec468b9, 0x2c342438, 0x5f40a3c2);
+	&_data_word(0x72c31d16, 0x0c25e2bc, 0x8b493c28, 0x41950dff);
+	&_data_word(0x7101a839, 0xdeb30c08, 0x9ce4b4d8, 0x90c15664);
+	&_data_word(0x6184cb7b, 0x70b632d5, 0x745c6c48, 0x4257b8d0);
+
+#Td4:	# four copies of Td4 to choose from to avoid L1 aliasing
+	&data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38);
+	&data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb);
+	&data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87);
+	&data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb);
+	&data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d);
+	&data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e);
+	&data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2);
+	&data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25);
+	&data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16);
+	&data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92);
+	&data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda);
+	&data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84);
+	&data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a);
+	&data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06);
+	&data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02);
+	&data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b);
+	&data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea);
+	&data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73);
+	&data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85);
+	&data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e);
+	&data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89);
+	&data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b);
+	&data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20);
+	&data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4);
+	&data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31);
+	&data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f);
+	&data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d);
+	&data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef);
+	&data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0);
+	&data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61);
+	&data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26);
+	&data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d);
+
+	&data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38);
+	&data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb);
+	&data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87);
+	&data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb);
+	&data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d);
+	&data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e);
+	&data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2);
+	&data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25);
+	&data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16);
+	&data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92);
+	&data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda);
+	&data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84);
+	&data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a);
+	&data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06);
+	&data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02);
+	&data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b);
+	&data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea);
+	&data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73);
+	&data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85);
+	&data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e);
+	&data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89);
+	&data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b);
+	&data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20);
+	&data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4);
+	&data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31);
+	&data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f);
+	&data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d);
+	&data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef);
+	&data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0);
+	&data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61);
+	&data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26);
+	&data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d);
+
+	&data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38);
+	&data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb);
+	&data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87);
+	&data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb);
+	&data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d);
+	&data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e);
+	&data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2);
+	&data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25);
+	&data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16);
+	&data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92);
+	&data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda);
+	&data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84);
+	&data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a);
+	&data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06);
+	&data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02);
+	&data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b);
+	&data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea);
+	&data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73);
+	&data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85);
+	&data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e);
+	&data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89);
+	&data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b);
+	&data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20);
+	&data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4);
+	&data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31);
+	&data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f);
+	&data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d);
+	&data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef);
+	&data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0);
+	&data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61);
+	&data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26);
+	&data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d);
+
+	&data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38);
+	&data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb);
+	&data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87);
+	&data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb);
+	&data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d);
+	&data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e);
+	&data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2);
+	&data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25);
+	&data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16);
+	&data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92);
+	&data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda);
+	&data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84);
+	&data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a);
+	&data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06);
+	&data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02);
+	&data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b);
+	&data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea);
+	&data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73);
+	&data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85);
+	&data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e);
+	&data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89);
+	&data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b);
+	&data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20);
+	&data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4);
+	&data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31);
+	&data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f);
+	&data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d);
+	&data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef);
+	&data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0);
+	&data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61);
+	&data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26);
+	&data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d);
+&function_end_B("_x86_AES_decrypt");
+
+# void AES_decrypt (const void *inp,void *out,const AES_KEY *key);
+&function_begin("AES_decrypt");
+	&mov	($acc,&wparam(0));		# load inp
+	&mov	($key,&wparam(2));		# load key
+
+	&mov	($s0,"esp");
+	&sub	("esp",36);
+	&and	("esp",-64);			# align to cache-line
+
+	# place stack frame just "above" the key schedule
+	&lea	($s1,&DWP(-64-63,$key));
+	&sub	($s1,"esp");
+	&neg	($s1);
+	&and	($s1,0x3C0);	# modulo 1024, but aligned to cache-line
+	&sub	("esp",$s1);
+	&add	("esp",4);	# 4 is reserved for caller's return address
+	&mov	($_esp,$s0);	# save stack pointer
+
+	&call   (&label("pic_point"));          # make it PIC!
+	&set_label("pic_point");
+	&blindpop($tbl);
+	&picmeup($s0,"OPENSSL_ia32cap_P",$tbl,&label("pic_point")) if(!$x86only);
+	&lea    ($tbl,&DWP(&label("AES_Td")."-".&label("pic_point"),$tbl));
+
+	# pick Td4 copy which can't "overlap" with stack frame or key schedule
+	&lea	($s1,&DWP(768-4,"esp"));
+	&sub	($s1,$tbl);
+	&and	($s1,0x300);
+	&lea	($tbl,&DWP(2048+128,$tbl,$s1));
+
+					if (!$x86only) {
+	&bt	(&DWP(0,$s0),25);	# check for SSE bit
+	&jnc	(&label("x86"));
+
+	&movq	("mm0",&QWP(0,$acc));
+	&movq	("mm4",&QWP(8,$acc));
+	&call	("_sse_AES_decrypt_compact");
+	&mov	("esp",$_esp);			# restore stack pointer
+	&mov	($acc,&wparam(1));		# load out
+	&movq	(&QWP(0,$acc),"mm0");		# write output data
+	&movq	(&QWP(8,$acc),"mm4");
+	&emms	();
+	&function_end_A();
+					}
+	&set_label("x86",16);
+	&mov	($_tbl,$tbl);
+	&mov	($s0,&DWP(0,$acc));		# load input data
+	&mov	($s1,&DWP(4,$acc));
+	&mov	($s2,&DWP(8,$acc));
+	&mov	($s3,&DWP(12,$acc));
+	&call	("_x86_AES_decrypt_compact");
+	&mov	("esp",$_esp);			# restore stack pointer
+	&mov	($acc,&wparam(1));		# load out
+	&mov	(&DWP(0,$acc),$s0);		# write output data
+	&mov	(&DWP(4,$acc),$s1);
+	&mov	(&DWP(8,$acc),$s2);
+	&mov	(&DWP(12,$acc),$s3);
+&function_end("AES_decrypt");
+
+# void AES_cbc_encrypt (const void char *inp, unsigned char *out,
+#			size_t length, const AES_KEY *key,
+#			unsigned char *ivp,const int enc);
+{
+# stack frame layout
+#             -4(%esp)		# return address	 0(%esp)
+#              0(%esp)		# s0 backing store	 4(%esp)	
+#              4(%esp)		# s1 backing store	 8(%esp)
+#              8(%esp)		# s2 backing store	12(%esp)
+#             12(%esp)		# s3 backing store	16(%esp)
+#             16(%esp)		# key backup		20(%esp)
+#             20(%esp)		# end of key schedule	24(%esp)
+#             24(%esp)		# %ebp backup		28(%esp)
+#             28(%esp)		# %esp backup
+my $_inp=&DWP(32,"esp");	# copy of wparam(0)
+my $_out=&DWP(36,"esp");	# copy of wparam(1)
+my $_len=&DWP(40,"esp");	# copy of wparam(2)
+my $_key=&DWP(44,"esp");	# copy of wparam(3)
+my $_ivp=&DWP(48,"esp");	# copy of wparam(4)
+my $_tmp=&DWP(52,"esp");	# volatile variable
+#
+my $ivec=&DWP(60,"esp");	# ivec[16]
+my $aes_key=&DWP(76,"esp");	# copy of aes_key
+my $mark=&DWP(76+240,"esp");	# copy of aes_key->rounds
+
+&function_begin("AES_cbc_encrypt");
+	&mov	($s2 eq "ecx"? $s2 : "",&wparam(2));	# load len
+	&cmp	($s2,0);
+	&je	(&label("drop_out"));
+
+	&call   (&label("pic_point"));		# make it PIC!
+	&set_label("pic_point");
+	&blindpop($tbl);
+	&picmeup($s0,"OPENSSL_ia32cap_P",$tbl,&label("pic_point")) if(!$x86only);
+
+	&cmp	(&wparam(5),0);
+	&lea    ($tbl,&DWP(&label("AES_Te")."-".&label("pic_point"),$tbl));
+	&jne	(&label("picked_te"));
+	&lea	($tbl,&DWP(&label("AES_Td")."-".&label("AES_Te"),$tbl));
+	&set_label("picked_te");
+
+	# one can argue if this is required
+	&pushf	();
+	&cld	();
+
+	&cmp	($s2,$speed_limit);
+	&jb	(&label("slow_way"));
+	&test	($s2,15);
+	&jnz	(&label("slow_way"));
+					if (!$x86only) {
+	&bt	(&DWP(0,$s0),28);	# check for hyper-threading bit
+	&jc	(&label("slow_way"));
+					}
+	# pre-allocate aligned stack frame...
+	&lea	($acc,&DWP(-80-244,"esp"));
+	&and	($acc,-64);
+
+	# ... and make sure it doesn't alias with $tbl modulo 4096
+	&mov	($s0,$tbl);
+	&lea	($s1,&DWP(2048+256,$tbl));
+	&mov	($s3,$acc);
+	&and	($s0,0xfff);		# s = %ebp&0xfff
+	&and	($s1,0xfff);		# e = (%ebp+2048+256)&0xfff
+	&and	($s3,0xfff);		# p = %esp&0xfff
+
+	&cmp	($s3,$s1);		# if (p>=e) %esp =- (p-e);
+	&jb	(&label("tbl_break_out"));
+	&sub	($s3,$s1);
+	&sub	($acc,$s3);
+	&jmp	(&label("tbl_ok"));
+	&set_label("tbl_break_out",4);	# else %esp -= (p-s)&0xfff + framesz;
+	&sub	($s3,$s0);
+	&and	($s3,0xfff);
+	&add	($s3,384);
+	&sub	($acc,$s3);
+	&set_label("tbl_ok",4);
+
+	&lea	($s3,&wparam(0));	# obtain pointer to parameter block
+	&exch	("esp",$acc);		# allocate stack frame
+	&add	("esp",4);		# reserve for return address!
+	&mov	($_tbl,$tbl);		# save %ebp
+	&mov	($_esp,$acc);		# save %esp
+
+	&mov	($s0,&DWP(0,$s3));	# load inp
+	&mov	($s1,&DWP(4,$s3));	# load out
+	#&mov	($s2,&DWP(8,$s3));	# load len
+	&mov	($key,&DWP(12,$s3));	# load key
+	&mov	($acc,&DWP(16,$s3));	# load ivp
+	&mov	($s3,&DWP(20,$s3));	# load enc flag
+
+	&mov	($_inp,$s0);		# save copy of inp
+	&mov	($_out,$s1);		# save copy of out
+	&mov	($_len,$s2);		# save copy of len
+	&mov	($_key,$key);		# save copy of key
+	&mov	($_ivp,$acc);		# save copy of ivp
+
+	&mov	($mark,0);		# copy of aes_key->rounds = 0;
+	# do we copy key schedule to stack?
+	&mov	($s1 eq "ebx" ? $s1 : "",$key);
+	&mov	($s2 eq "ecx" ? $s2 : "",244/4);
+	&sub	($s1,$tbl);
+	&mov	("esi",$key);
+	&and	($s1,0xfff);
+	&lea	("edi",$aes_key);
+	&cmp	($s1,2048+256);
+	&jb	(&label("do_copy"));
+	&cmp	($s1,4096-244);
+	&jb	(&label("skip_copy"));
+	&set_label("do_copy",4);
+		&mov	($_key,"edi");
+		&data_word(0xA5F3F689);	# rep movsd
+	&set_label("skip_copy");
+
+	&mov	($key,16);
+	&set_label("prefetch_tbl",4);
+		&mov	($s0,&DWP(0,$tbl));
+		&mov	($s1,&DWP(32,$tbl));
+		&mov	($s2,&DWP(64,$tbl));
+		&mov	($acc,&DWP(96,$tbl));
+		&lea	($tbl,&DWP(128,$tbl));
+		&sub	($key,1);
+	&jnz	(&label("prefetch_tbl"));
+	&sub	($tbl,2048);
+
+	&mov	($acc,$_inp);
+	&mov	($key,$_ivp);
+
+	&cmp	($s3,0);
+	&je	(&label("fast_decrypt"));
+
+#----------------------------- ENCRYPT -----------------------------#
+	&mov	($s0,&DWP(0,$key));		# load iv
+	&mov	($s1,&DWP(4,$key));
+
+	&set_label("fast_enc_loop",16);
+		&mov	($s2,&DWP(8,$key));
+		&mov	($s3,&DWP(12,$key));
+
+		&xor	($s0,&DWP(0,$acc));	# xor input data
+		&xor	($s1,&DWP(4,$acc));
+		&xor	($s2,&DWP(8,$acc));
+		&xor	($s3,&DWP(12,$acc));
+
+		&mov	($key,$_key);		# load key
+		&call	("_x86_AES_encrypt");
+
+		&mov	($acc,$_inp);		# load inp
+		&mov	($key,$_out);		# load out
+
+		&mov	(&DWP(0,$key),$s0);	# save output data
+		&mov	(&DWP(4,$key),$s1);
+		&mov	(&DWP(8,$key),$s2);
+		&mov	(&DWP(12,$key),$s3);
+
+		&lea	($acc,&DWP(16,$acc));	# advance inp
+		&mov	($s2,$_len);		# load len
+		&mov	($_inp,$acc);		# save inp
+		&lea	($s3,&DWP(16,$key));	# advance out
+		&mov	($_out,$s3);		# save out
+		&sub	($s2,16);		# decrease len
+		&mov	($_len,$s2);		# save len
+	&jnz	(&label("fast_enc_loop"));
+	&mov	($acc,$_ivp);		# load ivp
+	&mov	($s2,&DWP(8,$key));	# restore last 2 dwords
+	&mov	($s3,&DWP(12,$key));
+	&mov	(&DWP(0,$acc),$s0);	# save ivec
+	&mov	(&DWP(4,$acc),$s1);
+	&mov	(&DWP(8,$acc),$s2);
+	&mov	(&DWP(12,$acc),$s3);
+
+	&cmp	($mark,0);		# was the key schedule copied?
+	&mov	("edi",$_key);
+	&je	(&label("skip_ezero"));
+	# zero copy of key schedule
+	&mov	("ecx",240/4);
+	&xor	("eax","eax");
+	&align	(4);
+	&data_word(0xABF3F689);	# rep stosd
+	&set_label("skip_ezero")
+	&mov	("esp",$_esp);
+	&popf	();
+    &set_label("drop_out");
+	&function_end_A();
+	&pushf	();			# kludge, never executed
+
+#----------------------------- DECRYPT -----------------------------#
+&set_label("fast_decrypt",16);
+
+	&cmp	($acc,$_out);
+	&je	(&label("fast_dec_in_place"));	# in-place processing...
+
+	&mov	($_tmp,$key);
+
+	&align	(4);
+	&set_label("fast_dec_loop",16);
+		&mov	($s0,&DWP(0,$acc));	# read input
+		&mov	($s1,&DWP(4,$acc));
+		&mov	($s2,&DWP(8,$acc));
+		&mov	($s3,&DWP(12,$acc));
+
+		&mov	($key,$_key);		# load key
+		&call	("_x86_AES_decrypt");
+
+		&mov	($key,$_tmp);		# load ivp
+		&mov	($acc,$_len);		# load len
+		&xor	($s0,&DWP(0,$key));	# xor iv
+		&xor	($s1,&DWP(4,$key));
+		&xor	($s2,&DWP(8,$key));
+		&xor	($s3,&DWP(12,$key));
+
+		&mov	($key,$_out);		# load out
+		&mov	($acc,$_inp);		# load inp
+
+		&mov	(&DWP(0,$key),$s0);	# write output
+		&mov	(&DWP(4,$key),$s1);
+		&mov	(&DWP(8,$key),$s2);
+		&mov	(&DWP(12,$key),$s3);
+
+		&mov	($s2,$_len);		# load len
+		&mov	($_tmp,$acc);		# save ivp
+		&lea	($acc,&DWP(16,$acc));	# advance inp
+		&mov	($_inp,$acc);		# save inp
+		&lea	($key,&DWP(16,$key));	# advance out
+		&mov	($_out,$key);		# save out
+		&sub	($s2,16);		# decrease len
+		&mov	($_len,$s2);		# save len
+	&jnz	(&label("fast_dec_loop"));
+	&mov	($key,$_tmp);		# load temp ivp
+	&mov	($acc,$_ivp);		# load user ivp
+	&mov	($s0,&DWP(0,$key));	# load iv
+	&mov	($s1,&DWP(4,$key));
+	&mov	($s2,&DWP(8,$key));
+	&mov	($s3,&DWP(12,$key));
+	&mov	(&DWP(0,$acc),$s0);	# copy back to user
+	&mov	(&DWP(4,$acc),$s1);
+	&mov	(&DWP(8,$acc),$s2);
+	&mov	(&DWP(12,$acc),$s3);
+	&jmp	(&label("fast_dec_out"));
+
+    &set_label("fast_dec_in_place",16);
+	&set_label("fast_dec_in_place_loop");
+		&mov	($s0,&DWP(0,$acc));	# read input
+		&mov	($s1,&DWP(4,$acc));
+		&mov	($s2,&DWP(8,$acc));
+		&mov	($s3,&DWP(12,$acc));
+
+		&lea	($key,$ivec);
+		&mov	(&DWP(0,$key),$s0);	# copy to temp
+		&mov	(&DWP(4,$key),$s1);
+		&mov	(&DWP(8,$key),$s2);
+		&mov	(&DWP(12,$key),$s3);
+
+		&mov	($key,$_key);		# load key
+		&call	("_x86_AES_decrypt");
+
+		&mov	($key,$_ivp);		# load ivp
+		&mov	($acc,$_out);		# load out
+		&xor	($s0,&DWP(0,$key));	# xor iv
+		&xor	($s1,&DWP(4,$key));
+		&xor	($s2,&DWP(8,$key));
+		&xor	($s3,&DWP(12,$key));
+
+		&mov	(&DWP(0,$acc),$s0);	# write output
+		&mov	(&DWP(4,$acc),$s1);
+		&mov	(&DWP(8,$acc),$s2);
+		&mov	(&DWP(12,$acc),$s3);
+
+		&lea	($acc,&DWP(16,$acc));	# advance out
+		&mov	($_out,$acc);		# save out
+
+		&lea	($acc,$ivec);
+		&mov	($s0,&DWP(0,$acc));	# read temp
+		&mov	($s1,&DWP(4,$acc));
+		&mov	($s2,&DWP(8,$acc));
+		&mov	($s3,&DWP(12,$acc));
+
+		&mov	(&DWP(0,$key),$s0);	# copy iv
+		&mov	(&DWP(4,$key),$s1);
+		&mov	(&DWP(8,$key),$s2);
+		&mov	(&DWP(12,$key),$s3);
+
+		&mov	($acc,$_inp);		# load inp
+		&mov	($s2,$_len);		# load len
+		&lea	($acc,&DWP(16,$acc));	# advance inp
+		&mov	($_inp,$acc);		# save inp
+		&sub	($s2,16);		# decrease len
+		&mov	($_len,$s2);		# save len
+	&jnz	(&label("fast_dec_in_place_loop"));
+
+    &set_label("fast_dec_out",4);
+	&cmp	($mark,0);		# was the key schedule copied?
+	&mov	("edi",$_key);
+	&je	(&label("skip_dzero"));
+	# zero copy of key schedule
+	&mov	("ecx",240/4);
+	&xor	("eax","eax");
+	&align	(4);
+	&data_word(0xABF3F689);	# rep stosd
+	&set_label("skip_dzero")
+	&mov	("esp",$_esp);
+	&popf	();
+	&function_end_A();
+	&pushf	();			# kludge, never executed
+
+#--------------------------- SLOW ROUTINE ---------------------------#
+&set_label("slow_way",16);
+
+	&mov	($s0,&DWP(0,$s0)) if (!$x86only);# load OPENSSL_ia32cap
+	&mov	($key,&wparam(3));	# load key
+
+	# pre-allocate aligned stack frame...
+	&lea	($acc,&DWP(-80,"esp"));
+	&and	($acc,-64);
+
+	# ... and make sure it doesn't alias with $key modulo 1024
+	&lea	($s1,&DWP(-80-63,$key));
+	&sub	($s1,$acc);
+	&neg	($s1);
+	&and	($s1,0x3C0);	# modulo 1024, but aligned to cache-line
+	&sub	($acc,$s1);
+
+	# pick S-box copy which can't overlap with stack frame or $key
+	&lea	($s1,&DWP(768,$acc));
+	&sub	($s1,$tbl);
+	&and	($s1,0x300);
+	&lea	($tbl,&DWP(2048+128,$tbl,$s1));
+
+	&lea	($s3,&wparam(0));	# pointer to parameter block
+
+	&exch	("esp",$acc);
+	&add	("esp",4);		# reserve for return address!
+	&mov	($_tbl,$tbl);		# save %ebp
+	&mov	($_esp,$acc);		# save %esp
+	&mov	($_tmp,$s0);		# save OPENSSL_ia32cap
+
+	&mov	($s0,&DWP(0,$s3));	# load inp
+	&mov	($s1,&DWP(4,$s3));	# load out
+	#&mov	($s2,&DWP(8,$s3));	# load len
+	#&mov	($key,&DWP(12,$s3));	# load key
+	&mov	($acc,&DWP(16,$s3));	# load ivp
+	&mov	($s3,&DWP(20,$s3));	# load enc flag
+
+	&mov	($_inp,$s0);		# save copy of inp
+	&mov	($_out,$s1);		# save copy of out
+	&mov	($_len,$s2);		# save copy of len
+	&mov	($_key,$key);		# save copy of key
+	&mov	($_ivp,$acc);		# save copy of ivp
+
+	&mov	($key,$acc);
+	&mov	($acc,$s0);
+
+	&cmp	($s3,0);
+	&je	(&label("slow_decrypt"));
+
+#--------------------------- SLOW ENCRYPT ---------------------------#
+	&cmp	($s2,16);
+	&mov	($s3,$s1);
+	&jb	(&label("slow_enc_tail"));
+
+					if (!$x86only) {
+	&bt	($_tmp,25);		# check for SSE bit
+	&jnc	(&label("slow_enc_x86"));
+
+	&movq	("mm0",&QWP(0,$key));	# load iv
+	&movq	("mm4",&QWP(8,$key));
+
+	&set_label("slow_enc_loop_sse",16);
+		&pxor	("mm0",&QWP(0,$acc));	# xor input data
+		&pxor	("mm4",&QWP(8,$acc));
+
+		&mov	($key,$_key);
+		&call	("_sse_AES_encrypt_compact");
+
+		&mov	($acc,$_inp);		# load inp
+		&mov	($key,$_out);		# load out
+		&mov	($s2,$_len);		# load len
+
+		&movq	(&QWP(0,$key),"mm0");	# save output data
+		&movq	(&QWP(8,$key),"mm4");
+
+		&lea	($acc,&DWP(16,$acc));	# advance inp
+		&mov	($_inp,$acc);		# save inp
+		&lea	($s3,&DWP(16,$key));	# advance out
+		&mov	($_out,$s3);		# save out
+		&sub	($s2,16);		# decrease len
+		&cmp	($s2,16);
+		&mov	($_len,$s2);		# save len
+	&jae	(&label("slow_enc_loop_sse"));
+	&test	($s2,15);
+	&jnz	(&label("slow_enc_tail"));
+	&mov	($acc,$_ivp);		# load ivp
+	&movq	(&QWP(0,$acc),"mm0");	# save ivec
+	&movq	(&QWP(8,$acc),"mm4");
+	&emms	();
+	&mov	("esp",$_esp);
+	&popf	();
+	&function_end_A();
+	&pushf	();			# kludge, never executed
+					}
+    &set_label("slow_enc_x86",16);
+	&mov	($s0,&DWP(0,$key));	# load iv
+	&mov	($s1,&DWP(4,$key));
+
+	&set_label("slow_enc_loop_x86",4);
+		&mov	($s2,&DWP(8,$key));
+		&mov	($s3,&DWP(12,$key));
+
+		&xor	($s0,&DWP(0,$acc));	# xor input data
+		&xor	($s1,&DWP(4,$acc));
+		&xor	($s2,&DWP(8,$acc));
+		&xor	($s3,&DWP(12,$acc));
+
+		&mov	($key,$_key);		# load key
+		&call	("_x86_AES_encrypt_compact");
+
+		&mov	($acc,$_inp);		# load inp
+		&mov	($key,$_out);		# load out
+
+		&mov	(&DWP(0,$key),$s0);	# save output data
+		&mov	(&DWP(4,$key),$s1);
+		&mov	(&DWP(8,$key),$s2);
+		&mov	(&DWP(12,$key),$s3);
+
+		&mov	($s2,$_len);		# load len
+		&lea	($acc,&DWP(16,$acc));	# advance inp
+		&mov	($_inp,$acc);		# save inp
+		&lea	($s3,&DWP(16,$key));	# advance out
+		&mov	($_out,$s3);		# save out
+		&sub	($s2,16);		# decrease len
+		&cmp	($s2,16);
+		&mov	($_len,$s2);		# save len
+	&jae	(&label("slow_enc_loop_x86"));
+	&test	($s2,15);
+	&jnz	(&label("slow_enc_tail"));
+	&mov	($acc,$_ivp);		# load ivp
+	&mov	($s2,&DWP(8,$key));	# restore last dwords
+	&mov	($s3,&DWP(12,$key));
+	&mov	(&DWP(0,$acc),$s0);	# save ivec
+	&mov	(&DWP(4,$acc),$s1);
+	&mov	(&DWP(8,$acc),$s2);
+	&mov	(&DWP(12,$acc),$s3);
+
+	&mov	("esp",$_esp);
+	&popf	();
+	&function_end_A();
+	&pushf	();			# kludge, never executed
+
+    &set_label("slow_enc_tail",16);
+	&emms	()	if (!$x86only);
+	&mov	($key eq "edi"? $key:"",$s3);	# load out to edi
+	&mov	($s1,16);
+	&sub	($s1,$s2);
+	&cmp	($key,$acc eq "esi"? $acc:"");	# compare with inp
+	&je	(&label("enc_in_place"));
+	&align	(4);
+	&data_word(0xA4F3F689);	# rep movsb	# copy input
+	&jmp	(&label("enc_skip_in_place"));
+    &set_label("enc_in_place");
+	&lea	($key,&DWP(0,$key,$s2));
+    &set_label("enc_skip_in_place");
+	&mov	($s2,$s1);
+	&xor	($s0,$s0);
+	&align	(4);
+	&data_word(0xAAF3F689);	# rep stosb	# zero tail
+
+	&mov	($key,$_ivp);			# restore ivp
+	&mov	($acc,$s3);			# output as input
+	&mov	($s0,&DWP(0,$key));
+	&mov	($s1,&DWP(4,$key));
+	&mov	($_len,16);			# len=16
+	&jmp	(&label("slow_enc_loop_x86"));	# one more spin...
+
+#--------------------------- SLOW DECRYPT ---------------------------#
+&set_label("slow_decrypt",16);
+					if (!$x86only) {
+	&bt	($_tmp,25);		# check for SSE bit
+	&jnc	(&label("slow_dec_loop_x86"));
+
+	&set_label("slow_dec_loop_sse",4);
+		&movq	("mm0",&QWP(0,$acc));	# read input
+		&movq	("mm4",&QWP(8,$acc));
+
+		&mov	($key,$_key);
+		&call	("_sse_AES_decrypt_compact");
+
+		&mov	($acc,$_inp);		# load inp
+		&lea	($s0,$ivec);
+		&mov	($s1,$_out);		# load out
+		&mov	($s2,$_len);		# load len
+		&mov	($key,$_ivp);		# load ivp
+
+		&movq	("mm1",&QWP(0,$acc));	# re-read input
+		&movq	("mm5",&QWP(8,$acc));
+
+		&pxor	("mm0",&QWP(0,$key));	# xor iv
+		&pxor	("mm4",&QWP(8,$key));
+
+		&movq	(&QWP(0,$key),"mm1");	# copy input to iv
+		&movq	(&QWP(8,$key),"mm5");
+
+		&sub	($s2,16);		# decrease len
+		&jc	(&label("slow_dec_partial_sse"));
+
+		&movq	(&QWP(0,$s1),"mm0");	# write output
+		&movq	(&QWP(8,$s1),"mm4");
+
+		&lea	($s1,&DWP(16,$s1));	# advance out
+		&mov	($_out,$s1);		# save out
+		&lea	($acc,&DWP(16,$acc));	# advance inp
+		&mov	($_inp,$acc);		# save inp
+		&mov	($_len,$s2);		# save len
+	&jnz	(&label("slow_dec_loop_sse"));
+	&emms	();
+	&mov	("esp",$_esp);
+	&popf	();
+	&function_end_A();
+	&pushf	();			# kludge, never executed
+
+    &set_label("slow_dec_partial_sse",16);
+	&movq	(&QWP(0,$s0),"mm0");	# save output to temp
+	&movq	(&QWP(8,$s0),"mm4");
+	&emms	();
+
+	&add	($s2 eq "ecx" ? "ecx":"",16);
+	&mov	("edi",$s1);		# out
+	&mov	("esi",$s0);		# temp
+	&align	(4);
+	&data_word(0xA4F3F689);		# rep movsb # copy partial output
+
+	&mov	("esp",$_esp);
+	&popf	();
+	&function_end_A();
+	&pushf	();			# kludge, never executed
+					}
+	&set_label("slow_dec_loop_x86",16);
+		&mov	($s0,&DWP(0,$acc));	# read input
+		&mov	($s1,&DWP(4,$acc));
+		&mov	($s2,&DWP(8,$acc));
+		&mov	($s3,&DWP(12,$acc));
+
+		&lea	($key,$ivec);
+		&mov	(&DWP(0,$key),$s0);	# copy to temp
+		&mov	(&DWP(4,$key),$s1);
+		&mov	(&DWP(8,$key),$s2);
+		&mov	(&DWP(12,$key),$s3);
+
+		&mov	($key,$_key);		# load key
+		&call	("_x86_AES_decrypt_compact");
+
+		&mov	($key,$_ivp);		# load ivp
+		&mov	($acc,$_len);		# load len
+		&xor	($s0,&DWP(0,$key));	# xor iv
+		&xor	($s1,&DWP(4,$key));
+		&xor	($s2,&DWP(8,$key));
+		&xor	($s3,&DWP(12,$key));
+
+		&sub	($acc,16);
+		&jc	(&label("slow_dec_partial_x86"));
+
+		&mov	($_len,$acc);		# save len
+		&mov	($acc,$_out);		# load out
+
+		&mov	(&DWP(0,$acc),$s0);	# write output
+		&mov	(&DWP(4,$acc),$s1);
+		&mov	(&DWP(8,$acc),$s2);
+		&mov	(&DWP(12,$acc),$s3);
+
+		&lea	($acc,&DWP(16,$acc));	# advance out
+		&mov	($_out,$acc);		# save out
+
+		&lea	($acc,$ivec);
+		&mov	($s0,&DWP(0,$acc));	# read temp
+		&mov	($s1,&DWP(4,$acc));
+		&mov	($s2,&DWP(8,$acc));
+		&mov	($s3,&DWP(12,$acc));
+
+		&mov	(&DWP(0,$key),$s0);	# copy it to iv
+		&mov	(&DWP(4,$key),$s1);
+		&mov	(&DWP(8,$key),$s2);
+		&mov	(&DWP(12,$key),$s3);
+
+		&mov	($acc,$_inp);		# load inp
+		&lea	($acc,&DWP(16,$acc));	# advance inp
+		&mov	($_inp,$acc);		# save inp
+	&jnz	(&label("slow_dec_loop_x86"));
+	&mov	("esp",$_esp);
+	&popf	();
+	&function_end_A();
+	&pushf	();			# kludge, never executed
+
+    &set_label("slow_dec_partial_x86",16);
+	&lea	($acc,$ivec);
+	&mov	(&DWP(0,$acc),$s0);	# save output to temp
+	&mov	(&DWP(4,$acc),$s1);
+	&mov	(&DWP(8,$acc),$s2);
+	&mov	(&DWP(12,$acc),$s3);
+
+	&mov	($acc,$_inp);
+	&mov	($s0,&DWP(0,$acc));	# re-read input
+	&mov	($s1,&DWP(4,$acc));
+	&mov	($s2,&DWP(8,$acc));
+	&mov	($s3,&DWP(12,$acc));
+
+	&mov	(&DWP(0,$key),$s0);	# copy it to iv
+	&mov	(&DWP(4,$key),$s1);
+	&mov	(&DWP(8,$key),$s2);
+	&mov	(&DWP(12,$key),$s3);
+
+	&mov	("ecx",$_len);
+	&mov	("edi",$_out);
+	&lea	("esi",$ivec);
+	&align	(4);
+	&data_word(0xA4F3F689);		# rep movsb # copy partial output
+
+	&mov	("esp",$_esp);
+	&popf	();
+&function_end("AES_cbc_encrypt");
+}
+
+#------------------------------------------------------------------#
+
+sub enckey()
+{
+	&movz	("esi",&LB("edx"));		# rk[i]>>0
+	&movz	("ebx",&BP(-128,$tbl,"esi",1));
+	&movz	("esi",&HB("edx"));		# rk[i]>>8
+	&shl	("ebx",24);
+	&xor	("eax","ebx");
+
+	&movz	("ebx",&BP(-128,$tbl,"esi",1));
+	&shr	("edx",16);
+	&movz	("esi",&LB("edx"));		# rk[i]>>16
+	&xor	("eax","ebx");
+
+	&movz	("ebx",&BP(-128,$tbl,"esi",1));
+	&movz	("esi",&HB("edx"));		# rk[i]>>24
+	&shl	("ebx",8);
+	&xor	("eax","ebx");
+
+	&movz	("ebx",&BP(-128,$tbl,"esi",1));
+	&shl	("ebx",16);
+	&xor	("eax","ebx");
+
+	&xor	("eax",&DWP(1024-128,$tbl,"ecx",4));	# rcon
+}
+
+&function_begin("_x86_AES_set_encrypt_key");
+	&mov	("esi",&wparam(1));		# user supplied key
+	&mov	("edi",&wparam(3));		# private key schedule
+
+	&test	("esi",-1);
+	&jz	(&label("badpointer"));
+	&test	("edi",-1);
+	&jz	(&label("badpointer"));
+
+	&call	(&label("pic_point"));
+	&set_label("pic_point");
+	&blindpop($tbl);
+	&lea	($tbl,&DWP(&label("AES_Te")."-".&label("pic_point"),$tbl));
+	&lea	($tbl,&DWP(2048+128,$tbl));
+
+	# prefetch Te4
+	&mov	("eax",&DWP(0-128,$tbl));
+	&mov	("ebx",&DWP(32-128,$tbl));
+	&mov	("ecx",&DWP(64-128,$tbl));
+	&mov	("edx",&DWP(96-128,$tbl));
+	&mov	("eax",&DWP(128-128,$tbl));
+	&mov	("ebx",&DWP(160-128,$tbl));
+	&mov	("ecx",&DWP(192-128,$tbl));
+	&mov	("edx",&DWP(224-128,$tbl));
+
+	&mov	("ecx",&wparam(2));		# number of bits in key
+	&cmp	("ecx",128);
+	&je	(&label("10rounds"));
+	&cmp	("ecx",192);
+	&je	(&label("12rounds"));
+	&cmp	("ecx",256);
+	&je	(&label("14rounds"));
+	&mov	("eax",-2);			# invalid number of bits
+	&jmp	(&label("exit"));
+
+    &set_label("10rounds");
+	&mov	("eax",&DWP(0,"esi"));		# copy first 4 dwords
+	&mov	("ebx",&DWP(4,"esi"));
+	&mov	("ecx",&DWP(8,"esi"));
+	&mov	("edx",&DWP(12,"esi"));
+	&mov	(&DWP(0,"edi"),"eax");
+	&mov	(&DWP(4,"edi"),"ebx");
+	&mov	(&DWP(8,"edi"),"ecx");
+	&mov	(&DWP(12,"edi"),"edx");
+
+	&xor	("ecx","ecx");
+	&jmp	(&label("10shortcut"));
+
+	&align	(4);
+	&set_label("10loop");
+		&mov	("eax",&DWP(0,"edi"));		# rk[0]
+		&mov	("edx",&DWP(12,"edi"));		# rk[3]
+	&set_label("10shortcut");
+		&enckey	();
+
+		&mov	(&DWP(16,"edi"),"eax");		# rk[4]
+		&xor	("eax",&DWP(4,"edi"));
+		&mov	(&DWP(20,"edi"),"eax");		# rk[5]
+		&xor	("eax",&DWP(8,"edi"));
+		&mov	(&DWP(24,"edi"),"eax");		# rk[6]
+		&xor	("eax",&DWP(12,"edi"));
+		&mov	(&DWP(28,"edi"),"eax");		# rk[7]
+		&inc	("ecx");
+		&add	("edi",16);
+		&cmp	("ecx",10);
+	&jl	(&label("10loop"));
+
+	&mov	(&DWP(80,"edi"),10);		# setup number of rounds
+	&xor	("eax","eax");
+	&jmp	(&label("exit"));
+		
+    &set_label("12rounds");
+	&mov	("eax",&DWP(0,"esi"));		# copy first 6 dwords
+	&mov	("ebx",&DWP(4,"esi"));
+	&mov	("ecx",&DWP(8,"esi"));
+	&mov	("edx",&DWP(12,"esi"));
+	&mov	(&DWP(0,"edi"),"eax");
+	&mov	(&DWP(4,"edi"),"ebx");
+	&mov	(&DWP(8,"edi"),"ecx");
+	&mov	(&DWP(12,"edi"),"edx");
+	&mov	("ecx",&DWP(16,"esi"));
+	&mov	("edx",&DWP(20,"esi"));
+	&mov	(&DWP(16,"edi"),"ecx");
+	&mov	(&DWP(20,"edi"),"edx");
+
+	&xor	("ecx","ecx");
+	&jmp	(&label("12shortcut"));
+
+	&align	(4);
+	&set_label("12loop");
+		&mov	("eax",&DWP(0,"edi"));		# rk[0]
+		&mov	("edx",&DWP(20,"edi"));		# rk[5]
+	&set_label("12shortcut");
+		&enckey	();
+
+		&mov	(&DWP(24,"edi"),"eax");		# rk[6]
+		&xor	("eax",&DWP(4,"edi"));
+		&mov	(&DWP(28,"edi"),"eax");		# rk[7]
+		&xor	("eax",&DWP(8,"edi"));
+		&mov	(&DWP(32,"edi"),"eax");		# rk[8]
+		&xor	("eax",&DWP(12,"edi"));
+		&mov	(&DWP(36,"edi"),"eax");		# rk[9]
+
+		&cmp	("ecx",7);
+		&je	(&label("12break"));
+		&inc	("ecx");
+
+		&xor	("eax",&DWP(16,"edi"));
+		&mov	(&DWP(40,"edi"),"eax");		# rk[10]
+		&xor	("eax",&DWP(20,"edi"));
+		&mov	(&DWP(44,"edi"),"eax");		# rk[11]
+
+		&add	("edi",24);
+	&jmp	(&label("12loop"));
+
+	&set_label("12break");
+	&mov	(&DWP(72,"edi"),12);		# setup number of rounds
+	&xor	("eax","eax");
+	&jmp	(&label("exit"));
+
+    &set_label("14rounds");
+	&mov	("eax",&DWP(0,"esi"));		# copy first 8 dwords
+	&mov	("ebx",&DWP(4,"esi"));
+	&mov	("ecx",&DWP(8,"esi"));
+	&mov	("edx",&DWP(12,"esi"));
+	&mov	(&DWP(0,"edi"),"eax");
+	&mov	(&DWP(4,"edi"),"ebx");
+	&mov	(&DWP(8,"edi"),"ecx");
+	&mov	(&DWP(12,"edi"),"edx");
+	&mov	("eax",&DWP(16,"esi"));
+	&mov	("ebx",&DWP(20,"esi"));
+	&mov	("ecx",&DWP(24,"esi"));
+	&mov	("edx",&DWP(28,"esi"));
+	&mov	(&DWP(16,"edi"),"eax");
+	&mov	(&DWP(20,"edi"),"ebx");
+	&mov	(&DWP(24,"edi"),"ecx");
+	&mov	(&DWP(28,"edi"),"edx");
+
+	&xor	("ecx","ecx");
+	&jmp	(&label("14shortcut"));
+
+	&align	(4);
+	&set_label("14loop");
+		&mov	("edx",&DWP(28,"edi"));		# rk[7]
+	&set_label("14shortcut");
+		&mov	("eax",&DWP(0,"edi"));		# rk[0]
+
+		&enckey	();
+
+		&mov	(&DWP(32,"edi"),"eax");		# rk[8]
+		&xor	("eax",&DWP(4,"edi"));
+		&mov	(&DWP(36,"edi"),"eax");		# rk[9]
+		&xor	("eax",&DWP(8,"edi"));
+		&mov	(&DWP(40,"edi"),"eax");		# rk[10]
+		&xor	("eax",&DWP(12,"edi"));
+		&mov	(&DWP(44,"edi"),"eax");		# rk[11]
+
+		&cmp	("ecx",6);
+		&je	(&label("14break"));
+		&inc	("ecx");
+
+		&mov	("edx","eax");
+		&mov	("eax",&DWP(16,"edi"));		# rk[4]
+		&movz	("esi",&LB("edx"));		# rk[11]>>0
+		&movz	("ebx",&BP(-128,$tbl,"esi",1));
+		&movz	("esi",&HB("edx"));		# rk[11]>>8
+		&xor	("eax","ebx");
+
+		&movz	("ebx",&BP(-128,$tbl,"esi",1));
+		&shr	("edx",16);
+		&shl	("ebx",8);
+		&movz	("esi",&LB("edx"));		# rk[11]>>16
+		&xor	("eax","ebx");
+
+		&movz	("ebx",&BP(-128,$tbl,"esi",1));
+		&movz	("esi",&HB("edx"));		# rk[11]>>24
+		&shl	("ebx",16);
+		&xor	("eax","ebx");
+
+		&movz	("ebx",&BP(-128,$tbl,"esi",1));
+		&shl	("ebx",24);
+		&xor	("eax","ebx");
+
+		&mov	(&DWP(48,"edi"),"eax");		# rk[12]
+		&xor	("eax",&DWP(20,"edi"));
+		&mov	(&DWP(52,"edi"),"eax");		# rk[13]
+		&xor	("eax",&DWP(24,"edi"));
+		&mov	(&DWP(56,"edi"),"eax");		# rk[14]
+		&xor	("eax",&DWP(28,"edi"));
+		&mov	(&DWP(60,"edi"),"eax");		# rk[15]
+
+		&add	("edi",32);
+	&jmp	(&label("14loop"));
+
+	&set_label("14break");
+	&mov	(&DWP(48,"edi"),14);		# setup number of rounds
+	&xor	("eax","eax");
+	&jmp	(&label("exit"));
+
+    &set_label("badpointer");
+	&mov	("eax",-1);
+    &set_label("exit");
+&function_end("_x86_AES_set_encrypt_key");
+
+# int AES_set_encrypt_key(const unsigned char *userKey, const int bits,
+#                        AES_KEY *key)
+&function_begin_B("AES_set_encrypt_key");
+	&call	("_x86_AES_set_encrypt_key");
+	&ret	();
+&function_end_B("AES_set_encrypt_key");
+
+sub deckey()
+{ my ($i,$key,$tp1,$tp2,$tp4,$tp8) = @_;
+  my $tmp = $tbl;
+
+	&mov	($acc,$tp1);
+	&and	($acc,0x80808080);
+	&mov	($tmp,$acc);
+	&shr	($tmp,7);
+	&lea	($tp2,&DWP(0,$tp1,$tp1));
+	&sub	($acc,$tmp);
+	&and	($tp2,0xfefefefe);
+	&and	($acc,0x1b1b1b1b);
+	&xor	($acc,$tp2);
+	&mov	($tp2,$acc);
+
+	&and	($acc,0x80808080);
+	&mov	($tmp,$acc);
+	&shr	($tmp,7);
+	&lea	($tp4,&DWP(0,$tp2,$tp2));
+	&sub	($acc,$tmp);
+	&and	($tp4,0xfefefefe);
+	&and	($acc,0x1b1b1b1b);
+	 &xor	($tp2,$tp1);	# tp2^tp1
+	&xor	($acc,$tp4);
+	&mov	($tp4,$acc);
+
+	&and	($acc,0x80808080);
+	&mov	($tmp,$acc);
+	&shr	($tmp,7);
+	&lea	($tp8,&DWP(0,$tp4,$tp4));
+	 &xor	($tp4,$tp1);	# tp4^tp1
+	&sub	($acc,$tmp);
+	&and	($tp8,0xfefefefe);
+	&and	($acc,0x1b1b1b1b);
+	 &rotl	($tp1,8);	# = ROTATE(tp1,8)
+	&xor	($tp8,$acc);
+
+	&mov	($tmp,&DWP(4*($i+1),$key));	# modulo-scheduled load
+
+	&xor	($tp1,$tp2);
+	&xor	($tp2,$tp8);
+	&xor	($tp1,$tp4);
+	&rotl	($tp2,24);
+	&xor	($tp4,$tp8);
+	&xor	($tp1,$tp8);	# ^= tp8^(tp4^tp1)^(tp2^tp1)
+	&rotl	($tp4,16);
+	&xor	($tp1,$tp2);	# ^= ROTATE(tp8^tp2^tp1,24)
+	&rotl	($tp8,8);
+	&xor	($tp1,$tp4);	# ^= ROTATE(tp8^tp4^tp1,16)
+	&mov	($tp2,$tmp);
+	&xor	($tp1,$tp8);	# ^= ROTATE(tp8,8)
+
+	&mov	(&DWP(4*$i,$key),$tp1);
+}
+
+# int AES_set_decrypt_key(const unsigned char *userKey, const int bits,
+#                        AES_KEY *key)
+&function_begin_B("AES_set_decrypt_key");
+	&call	("_x86_AES_set_encrypt_key");
+	&cmp	("eax",0);
+	&je	(&label("proceed"));
+	&ret	();
+
+    &set_label("proceed");
+	&push	("ebp");
+	&push	("ebx");
+	&push	("esi");
+	&push	("edi");
+
+	&mov	("esi",&wparam(2));
+	&mov	("ecx",&DWP(240,"esi"));	# pull number of rounds
+	&lea	("ecx",&DWP(0,"","ecx",4));
+	&lea	("edi",&DWP(0,"esi","ecx",4));	# pointer to last chunk
+
+	&set_label("invert",4);			# invert order of chunks
+		&mov	("eax",&DWP(0,"esi"));
+		&mov	("ebx",&DWP(4,"esi"));
+		&mov	("ecx",&DWP(0,"edi"));
+		&mov	("edx",&DWP(4,"edi"));
+		&mov	(&DWP(0,"edi"),"eax");
+		&mov	(&DWP(4,"edi"),"ebx");
+		&mov	(&DWP(0,"esi"),"ecx");
+		&mov	(&DWP(4,"esi"),"edx");
+		&mov	("eax",&DWP(8,"esi"));
+		&mov	("ebx",&DWP(12,"esi"));
+		&mov	("ecx",&DWP(8,"edi"));
+		&mov	("edx",&DWP(12,"edi"));
+		&mov	(&DWP(8,"edi"),"eax");
+		&mov	(&DWP(12,"edi"),"ebx");
+		&mov	(&DWP(8,"esi"),"ecx");
+		&mov	(&DWP(12,"esi"),"edx");
+		&add	("esi",16);
+		&sub	("edi",16);
+		&cmp	("esi","edi");
+	&jne	(&label("invert"));
+
+	&mov	($key,&wparam(2));
+	&mov	($acc,&DWP(240,$key));		# pull number of rounds
+	&lea	($acc,&DWP(-2,$acc,$acc));
+	&lea	($acc,&DWP(0,$key,$acc,8));
+	&mov	(&wparam(2),$acc);
+
+	&mov	($s0,&DWP(16,$key));		# modulo-scheduled load
+	&set_label("permute",4);		# permute the key schedule
+		&add	($key,16);
+		&deckey	(0,$key,$s0,$s1,$s2,$s3);
+		&deckey	(1,$key,$s1,$s2,$s3,$s0);
+		&deckey	(2,$key,$s2,$s3,$s0,$s1);
+		&deckey	(3,$key,$s3,$s0,$s1,$s2);
+		&cmp	($key,&wparam(2));
+	&jb	(&label("permute"));
+
+	&xor	("eax","eax");			# return success
+&function_end("AES_set_decrypt_key");
+&asciz("AES for x86, CRYPTOGAMS by <appro\@openssl.org>");
+
+&asm_finish();
diff --git a/openssl/crypto/aes/asm/aes-armv4.pl b/openssl/crypto/aes/asm/aes-armv4.pl
new file mode 100644
index 00000000..c51ee1fb
--- /dev/null
+++ b/openssl/crypto/aes/asm/aes-armv4.pl
@@ -0,0 +1,1030 @@
+#!/usr/bin/env perl
+
+# ====================================================================
+# Written by Andy Polyakov <appro@fy.chalmers.se> 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/.
+# ====================================================================
+
+# AES for ARMv4
+
+# January 2007.
+#
+# Code uses single 1K S-box and is >2 times faster than code generated
+# by gcc-3.4.1. This is thanks to unique feature of ARMv4 ISA, which
+# allows to merge logical or arithmetic operation with shift or rotate
+# in one instruction and emit combined result every cycle. The module
+# is endian-neutral. The performance is ~42 cycles/byte for 128-bit
+# key [on single-issue Xscale PXA250 core].
+
+# May 2007.
+#
+# AES_set_[en|de]crypt_key is added.
+
+# July 2010.
+#
+# Rescheduling for dual-issue pipeline resulted in 12% improvement on
+# Cortex A8 core and ~25 cycles per byte processed with 128-bit key.
+
+while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {}
+open STDOUT,">$output";
+
+$s0="r0";
+$s1="r1";
+$s2="r2";
+$s3="r3";
+$t1="r4";
+$t2="r5";
+$t3="r6";
+$i1="r7";
+$i2="r8";
+$i3="r9";
+
+$tbl="r10";
+$key="r11";
+$rounds="r12";
+
+$code=<<___;
+.text
+.code	32
+
+.type	AES_Te,%object
+.align	5
+AES_Te:
+.word	0xc66363a5, 0xf87c7c84, 0xee777799, 0xf67b7b8d
+.word	0xfff2f20d, 0xd66b6bbd, 0xde6f6fb1, 0x91c5c554
+.word	0x60303050, 0x02010103, 0xce6767a9, 0x562b2b7d
+.word	0xe7fefe19, 0xb5d7d762, 0x4dababe6, 0xec76769a
+.word	0x8fcaca45, 0x1f82829d, 0x89c9c940, 0xfa7d7d87
+.word	0xeffafa15, 0xb25959eb, 0x8e4747c9, 0xfbf0f00b
+.word	0x41adadec, 0xb3d4d467, 0x5fa2a2fd, 0x45afafea
+.word	0x239c9cbf, 0x53a4a4f7, 0xe4727296, 0x9bc0c05b
+.word	0x75b7b7c2, 0xe1fdfd1c, 0x3d9393ae, 0x4c26266a
+.word	0x6c36365a, 0x7e3f3f41, 0xf5f7f702, 0x83cccc4f
+.word	0x6834345c, 0x51a5a5f4, 0xd1e5e534, 0xf9f1f108
+.word	0xe2717193, 0xabd8d873, 0x62313153, 0x2a15153f
+.word	0x0804040c, 0x95c7c752, 0x46232365, 0x9dc3c35e
+.word	0x30181828, 0x379696a1, 0x0a05050f, 0x2f9a9ab5
+.word	0x0e070709, 0x24121236, 0x1b80809b, 0xdfe2e23d
+.word	0xcdebeb26, 0x4e272769, 0x7fb2b2cd, 0xea75759f
+.word	0x1209091b, 0x1d83839e, 0x582c2c74, 0x341a1a2e
+.word	0x361b1b2d, 0xdc6e6eb2, 0xb45a5aee, 0x5ba0a0fb
+.word	0xa45252f6, 0x763b3b4d, 0xb7d6d661, 0x7db3b3ce
+.word	0x5229297b, 0xdde3e33e, 0x5e2f2f71, 0x13848497
+.word	0xa65353f5, 0xb9d1d168, 0x00000000, 0xc1eded2c
+.word	0x40202060, 0xe3fcfc1f, 0x79b1b1c8, 0xb65b5bed
+.word	0xd46a6abe, 0x8dcbcb46, 0x67bebed9, 0x7239394b
+.word	0x944a4ade, 0x984c4cd4, 0xb05858e8, 0x85cfcf4a
+.word	0xbbd0d06b, 0xc5efef2a, 0x4faaaae5, 0xedfbfb16
+.word	0x864343c5, 0x9a4d4dd7, 0x66333355, 0x11858594
+.word	0x8a4545cf, 0xe9f9f910, 0x04020206, 0xfe7f7f81
+.word	0xa05050f0, 0x783c3c44, 0x259f9fba, 0x4ba8a8e3
+.word	0xa25151f3, 0x5da3a3fe, 0x804040c0, 0x058f8f8a
+.word	0x3f9292ad, 0x219d9dbc, 0x70383848, 0xf1f5f504
+.word	0x63bcbcdf, 0x77b6b6c1, 0xafdada75, 0x42212163
+.word	0x20101030, 0xe5ffff1a, 0xfdf3f30e, 0xbfd2d26d
+.word	0x81cdcd4c, 0x180c0c14, 0x26131335, 0xc3ecec2f
+.word	0xbe5f5fe1, 0x359797a2, 0x884444cc, 0x2e171739
+.word	0x93c4c457, 0x55a7a7f2, 0xfc7e7e82, 0x7a3d3d47
+.word	0xc86464ac, 0xba5d5de7, 0x3219192b, 0xe6737395
+.word	0xc06060a0, 0x19818198, 0x9e4f4fd1, 0xa3dcdc7f
+.word	0x44222266, 0x542a2a7e, 0x3b9090ab, 0x0b888883
+.word	0x8c4646ca, 0xc7eeee29, 0x6bb8b8d3, 0x2814143c
+.word	0xa7dede79, 0xbc5e5ee2, 0x160b0b1d, 0xaddbdb76
+.word	0xdbe0e03b, 0x64323256, 0x743a3a4e, 0x140a0a1e
+.word	0x924949db, 0x0c06060a, 0x4824246c, 0xb85c5ce4
+.word	0x9fc2c25d, 0xbdd3d36e, 0x43acacef, 0xc46262a6
+.word	0x399191a8, 0x319595a4, 0xd3e4e437, 0xf279798b
+.word	0xd5e7e732, 0x8bc8c843, 0x6e373759, 0xda6d6db7
+.word	0x018d8d8c, 0xb1d5d564, 0x9c4e4ed2, 0x49a9a9e0
+.word	0xd86c6cb4, 0xac5656fa, 0xf3f4f407, 0xcfeaea25
+.word	0xca6565af, 0xf47a7a8e, 0x47aeaee9, 0x10080818
+.word	0x6fbabad5, 0xf0787888, 0x4a25256f, 0x5c2e2e72
+.word	0x381c1c24, 0x57a6a6f1, 0x73b4b4c7, 0x97c6c651
+.word	0xcbe8e823, 0xa1dddd7c, 0xe874749c, 0x3e1f1f21
+.word	0x964b4bdd, 0x61bdbddc, 0x0d8b8b86, 0x0f8a8a85
+.word	0xe0707090, 0x7c3e3e42, 0x71b5b5c4, 0xcc6666aa
+.word	0x904848d8, 0x06030305, 0xf7f6f601, 0x1c0e0e12
+.word	0xc26161a3, 0x6a35355f, 0xae5757f9, 0x69b9b9d0
+.word	0x17868691, 0x99c1c158, 0x3a1d1d27, 0x279e9eb9
+.word	0xd9e1e138, 0xebf8f813, 0x2b9898b3, 0x22111133
+.word	0xd26969bb, 0xa9d9d970, 0x078e8e89, 0x339494a7
+.word	0x2d9b9bb6, 0x3c1e1e22, 0x15878792, 0xc9e9e920
+.word	0x87cece49, 0xaa5555ff, 0x50282878, 0xa5dfdf7a
+.word	0x038c8c8f, 0x59a1a1f8, 0x09898980, 0x1a0d0d17
+.word	0x65bfbfda, 0xd7e6e631, 0x844242c6, 0xd06868b8
+.word	0x824141c3, 0x299999b0, 0x5a2d2d77, 0x1e0f0f11
+.word	0x7bb0b0cb, 0xa85454fc, 0x6dbbbbd6, 0x2c16163a
+@ Te4[256]
+.byte	0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5
+.byte	0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76
+.byte	0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0
+.byte	0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0
+.byte	0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc
+.byte	0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15
+.byte	0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a
+.byte	0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75
+.byte	0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0
+.byte	0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84
+.byte	0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b
+.byte	0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf
+.byte	0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85
+.byte	0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8
+.byte	0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5
+.byte	0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2
+.byte	0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17
+.byte	0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73
+.byte	0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88
+.byte	0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb
+.byte	0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c
+.byte	0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79
+.byte	0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9
+.byte	0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08
+.byte	0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6
+.byte	0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a
+.byte	0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e
+.byte	0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e
+.byte	0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94
+.byte	0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf
+.byte	0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68
+.byte	0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16
+@ rcon[]
+.word	0x01000000, 0x02000000, 0x04000000, 0x08000000
+.word	0x10000000, 0x20000000, 0x40000000, 0x80000000
+.word	0x1B000000, 0x36000000, 0, 0, 0, 0, 0, 0
+.size	AES_Te,.-AES_Te
+
+@ void AES_encrypt(const unsigned char *in, unsigned char *out,
+@ 		 const AES_KEY *key) {
+.global AES_encrypt
+.type   AES_encrypt,%function
+.align	5
+AES_encrypt:
+	sub	r3,pc,#8		@ AES_encrypt
+	stmdb   sp!,{r1,r4-r12,lr}
+	mov	$rounds,r0		@ inp
+	mov	$key,r2
+	sub	$tbl,r3,#AES_encrypt-AES_Te	@ Te
+
+	ldrb	$s0,[$rounds,#3]	@ load input data in endian-neutral
+	ldrb	$t1,[$rounds,#2]	@ manner...
+	ldrb	$t2,[$rounds,#1]
+	ldrb	$t3,[$rounds,#0]
+	orr	$s0,$s0,$t1,lsl#8
+	ldrb	$s1,[$rounds,#7]
+	orr	$s0,$s0,$t2,lsl#16
+	ldrb	$t1,[$rounds,#6]
+	orr	$s0,$s0,$t3,lsl#24
+	ldrb	$t2,[$rounds,#5]
+	ldrb	$t3,[$rounds,#4]
+	orr	$s1,$s1,$t1,lsl#8
+	ldrb	$s2,[$rounds,#11]
+	orr	$s1,$s1,$t2,lsl#16
+	ldrb	$t1,[$rounds,#10]
+	orr	$s1,$s1,$t3,lsl#24
+	ldrb	$t2,[$rounds,#9]
+	ldrb	$t3,[$rounds,#8]
+	orr	$s2,$s2,$t1,lsl#8
+	ldrb	$s3,[$rounds,#15]
+	orr	$s2,$s2,$t2,lsl#16
+	ldrb	$t1,[$rounds,#14]
+	orr	$s2,$s2,$t3,lsl#24
+	ldrb	$t2,[$rounds,#13]
+	ldrb	$t3,[$rounds,#12]
+	orr	$s3,$s3,$t1,lsl#8
+	orr	$s3,$s3,$t2,lsl#16
+	orr	$s3,$s3,$t3,lsl#24
+
+	bl	_armv4_AES_encrypt
+
+	ldr	$rounds,[sp],#4		@ pop out
+	mov	$t1,$s0,lsr#24		@ write output in endian-neutral
+	mov	$t2,$s0,lsr#16		@ manner...
+	mov	$t3,$s0,lsr#8
+	strb	$t1,[$rounds,#0]
+	strb	$t2,[$rounds,#1]
+	mov	$t1,$s1,lsr#24
+	strb	$t3,[$rounds,#2]
+	mov	$t2,$s1,lsr#16
+	strb	$s0,[$rounds,#3]
+	mov	$t3,$s1,lsr#8
+	strb	$t1,[$rounds,#4]
+	strb	$t2,[$rounds,#5]
+	mov	$t1,$s2,lsr#24
+	strb	$t3,[$rounds,#6]
+	mov	$t2,$s2,lsr#16
+	strb	$s1,[$rounds,#7]
+	mov	$t3,$s2,lsr#8
+	strb	$t1,[$rounds,#8]
+	strb	$t2,[$rounds,#9]
+	mov	$t1,$s3,lsr#24
+	strb	$t3,[$rounds,#10]
+	mov	$t2,$s3,lsr#16
+	strb	$s2,[$rounds,#11]
+	mov	$t3,$s3,lsr#8
+	strb	$t1,[$rounds,#12]
+	strb	$t2,[$rounds,#13]
+	strb	$t3,[$rounds,#14]
+	strb	$s3,[$rounds,#15]
+
+	ldmia   sp!,{r4-r12,lr}
+	tst	lr,#1
+	moveq	pc,lr			@ be binary compatible with V4, yet
+	bx	lr			@ interoperable with Thumb ISA:-)
+.size	AES_encrypt,.-AES_encrypt
+
+.type   _armv4_AES_encrypt,%function
+.align	2
+_armv4_AES_encrypt:
+	str	lr,[sp,#-4]!		@ push lr
+	ldmia	$key!,{$t1-$i1}
+	eor	$s0,$s0,$t1
+	ldr	$rounds,[$key,#240-16]
+	eor	$s1,$s1,$t2
+	eor	$s2,$s2,$t3
+	eor	$s3,$s3,$i1
+	sub	$rounds,$rounds,#1
+	mov	lr,#255
+
+	and	$i1,lr,$s0
+	and	$i2,lr,$s0,lsr#8
+	and	$i3,lr,$s0,lsr#16
+	mov	$s0,$s0,lsr#24
+.Lenc_loop:
+	ldr	$t1,[$tbl,$i1,lsl#2]	@ Te3[s0>>0]
+	and	$i1,lr,$s1,lsr#16	@ i0
+	ldr	$t2,[$tbl,$i2,lsl#2]	@ Te2[s0>>8]
+	and	$i2,lr,$s1
+	ldr	$t3,[$tbl,$i3,lsl#2]	@ Te1[s0>>16]
+	and	$i3,lr,$s1,lsr#8
+	ldr	$s0,[$tbl,$s0,lsl#2]	@ Te0[s0>>24]
+	mov	$s1,$s1,lsr#24
+
+	ldr	$i1,[$tbl,$i1,lsl#2]	@ Te1[s1>>16]
+	ldr	$i2,[$tbl,$i2,lsl#2]	@ Te3[s1>>0]
+	ldr	$i3,[$tbl,$i3,lsl#2]	@ Te2[s1>>8]
+	eor	$s0,$s0,$i1,ror#8
+	ldr	$s1,[$tbl,$s1,lsl#2]	@ Te0[s1>>24]
+	and	$i1,lr,$s2,lsr#8	@ i0
+	eor	$t2,$t2,$i2,ror#8
+	and	$i2,lr,$s2,lsr#16	@ i1
+	eor	$t3,$t3,$i3,ror#8
+	and	$i3,lr,$s2
+	eor	$s1,$s1,$t1,ror#24
+	ldr	$i1,[$tbl,$i1,lsl#2]	@ Te2[s2>>8]
+	mov	$s2,$s2,lsr#24
+
+	ldr	$i2,[$tbl,$i2,lsl#2]	@ Te1[s2>>16]
+	ldr	$i3,[$tbl,$i3,lsl#2]	@ Te3[s2>>0]
+	eor	$s0,$s0,$i1,ror#16
+	ldr	$s2,[$tbl,$s2,lsl#2]	@ Te0[s2>>24]
+	and	$i1,lr,$s3		@ i0
+	eor	$s1,$s1,$i2,ror#8
+	and	$i2,lr,$s3,lsr#8	@ i1
+	eor	$t3,$t3,$i3,ror#16
+	and	$i3,lr,$s3,lsr#16	@ i2
+	eor	$s2,$s2,$t2,ror#16
+	ldr	$i1,[$tbl,$i1,lsl#2]	@ Te3[s3>>0]
+	mov	$s3,$s3,lsr#24
+
+	ldr	$i2,[$tbl,$i2,lsl#2]	@ Te2[s3>>8]
+	ldr	$i3,[$tbl,$i3,lsl#2]	@ Te1[s3>>16]
+	eor	$s0,$s0,$i1,ror#24
+	ldr	$s3,[$tbl,$s3,lsl#2]	@ Te0[s3>>24]
+	eor	$s1,$s1,$i2,ror#16
+	ldr	$i1,[$key],#16
+	eor	$s2,$s2,$i3,ror#8
+	ldr	$t1,[$key,#-12]
+	eor	$s3,$s3,$t3,ror#8
+
+	ldr	$t2,[$key,#-8]
+	eor	$s0,$s0,$i1
+	ldr	$t3,[$key,#-4]
+	and	$i1,lr,$s0
+	eor	$s1,$s1,$t1
+	and	$i2,lr,$s0,lsr#8
+	eor	$s2,$s2,$t2
+	and	$i3,lr,$s0,lsr#16
+	eor	$s3,$s3,$t3
+	mov	$s0,$s0,lsr#24
+
+	subs	$rounds,$rounds,#1
+	bne	.Lenc_loop
+
+	add	$tbl,$tbl,#2
+
+	ldrb	$t1,[$tbl,$i1,lsl#2]	@ Te4[s0>>0]
+	and	$i1,lr,$s1,lsr#16	@ i0
+	ldrb	$t2,[$tbl,$i2,lsl#2]	@ Te4[s0>>8]
+	and	$i2,lr,$s1
+	ldrb	$t3,[$tbl,$i3,lsl#2]	@ Te4[s0>>16]
+	and	$i3,lr,$s1,lsr#8
+	ldrb	$s0,[$tbl,$s0,lsl#2]	@ Te4[s0>>24]
+	mov	$s1,$s1,lsr#24
+
+	ldrb	$i1,[$tbl,$i1,lsl#2]	@ Te4[s1>>16]
+	ldrb	$i2,[$tbl,$i2,lsl#2]	@ Te4[s1>>0]
+	ldrb	$i3,[$tbl,$i3,lsl#2]	@ Te4[s1>>8]
+	eor	$s0,$i1,$s0,lsl#8
+	ldrb	$s1,[$tbl,$s1,lsl#2]	@ Te4[s1>>24]
+	and	$i1,lr,$s2,lsr#8	@ i0
+	eor	$t2,$i2,$t2,lsl#8
+	and	$i2,lr,$s2,lsr#16	@ i1
+	eor	$t3,$i3,$t3,lsl#8
+	and	$i3,lr,$s2
+	eor	$s1,$t1,$s1,lsl#24
+	ldrb	$i1,[$tbl,$i1,lsl#2]	@ Te4[s2>>8]
+	mov	$s2,$s2,lsr#24
+
+	ldrb	$i2,[$tbl,$i2,lsl#2]	@ Te4[s2>>16]
+	ldrb	$i3,[$tbl,$i3,lsl#2]	@ Te4[s2>>0]
+	eor	$s0,$i1,$s0,lsl#8
+	ldrb	$s2,[$tbl,$s2,lsl#2]	@ Te4[s2>>24]
+	and	$i1,lr,$s3		@ i0
+	eor	$s1,$s1,$i2,lsl#16
+	and	$i2,lr,$s3,lsr#8	@ i1
+	eor	$t3,$i3,$t3,lsl#8
+	and	$i3,lr,$s3,lsr#16	@ i2
+	eor	$s2,$t2,$s2,lsl#24
+	ldrb	$i1,[$tbl,$i1,lsl#2]	@ Te4[s3>>0]
+	mov	$s3,$s3,lsr#24
+
+	ldrb	$i2,[$tbl,$i2,lsl#2]	@ Te4[s3>>8]
+	ldrb	$i3,[$tbl,$i3,lsl#2]	@ Te4[s3>>16]
+	eor	$s0,$i1,$s0,lsl#8
+	ldrb	$s3,[$tbl,$s3,lsl#2]	@ Te4[s3>>24]
+	ldr	$i1,[$key,#0]
+	eor	$s1,$s1,$i2,lsl#8
+	ldr	$t1,[$key,#4]
+	eor	$s2,$s2,$i3,lsl#16
+	ldr	$t2,[$key,#8]
+	eor	$s3,$t3,$s3,lsl#24
+	ldr	$t3,[$key,#12]
+
+	eor	$s0,$s0,$i1
+	eor	$s1,$s1,$t1
+	eor	$s2,$s2,$t2
+	eor	$s3,$s3,$t3
+
+	sub	$tbl,$tbl,#2
+	ldr	pc,[sp],#4		@ pop and return
+.size	_armv4_AES_encrypt,.-_armv4_AES_encrypt
+
+.global AES_set_encrypt_key
+.type   AES_set_encrypt_key,%function
+.align	5
+AES_set_encrypt_key:
+	sub	r3,pc,#8		@ AES_set_encrypt_key
+	teq	r0,#0
+	moveq	r0,#-1
+	beq	.Labrt
+	teq	r2,#0
+	moveq	r0,#-1
+	beq	.Labrt
+
+	teq	r1,#128
+	beq	.Lok
+	teq	r1,#192
+	beq	.Lok
+	teq	r1,#256
+	movne	r0,#-1
+	bne	.Labrt
+
+.Lok:	stmdb   sp!,{r4-r12,lr}
+	sub	$tbl,r3,#AES_set_encrypt_key-AES_Te-1024	@ Te4
+
+	mov	$rounds,r0		@ inp
+	mov	lr,r1			@ bits
+	mov	$key,r2			@ key
+
+	ldrb	$s0,[$rounds,#3]	@ load input data in endian-neutral
+	ldrb	$t1,[$rounds,#2]	@ manner...
+	ldrb	$t2,[$rounds,#1]
+	ldrb	$t3,[$rounds,#0]
+	orr	$s0,$s0,$t1,lsl#8
+	ldrb	$s1,[$rounds,#7]
+	orr	$s0,$s0,$t2,lsl#16
+	ldrb	$t1,[$rounds,#6]
+	orr	$s0,$s0,$t3,lsl#24
+	ldrb	$t2,[$rounds,#5]
+	ldrb	$t3,[$rounds,#4]
+	orr	$s1,$s1,$t1,lsl#8
+	ldrb	$s2,[$rounds,#11]
+	orr	$s1,$s1,$t2,lsl#16
+	ldrb	$t1,[$rounds,#10]
+	orr	$s1,$s1,$t3,lsl#24
+	ldrb	$t2,[$rounds,#9]
+	ldrb	$t3,[$rounds,#8]
+	orr	$s2,$s2,$t1,lsl#8
+	ldrb	$s3,[$rounds,#15]
+	orr	$s2,$s2,$t2,lsl#16
+	ldrb	$t1,[$rounds,#14]
+	orr	$s2,$s2,$t3,lsl#24
+	ldrb	$t2,[$rounds,#13]
+	ldrb	$t3,[$rounds,#12]
+	orr	$s3,$s3,$t1,lsl#8
+	str	$s0,[$key],#16
+	orr	$s3,$s3,$t2,lsl#16
+	str	$s1,[$key,#-12]
+	orr	$s3,$s3,$t3,lsl#24
+	str	$s2,[$key,#-8]
+	str	$s3,[$key,#-4]
+
+	teq	lr,#128
+	bne	.Lnot128
+	mov	$rounds,#10
+	str	$rounds,[$key,#240-16]
+	add	$t3,$tbl,#256			@ rcon
+	mov	lr,#255
+
+.L128_loop:
+	and	$t2,lr,$s3,lsr#24
+	and	$i1,lr,$s3,lsr#16
+	ldrb	$t2,[$tbl,$t2]
+	and	$i2,lr,$s3,lsr#8
+	ldrb	$i1,[$tbl,$i1]
+	and	$i3,lr,$s3
+	ldrb	$i2,[$tbl,$i2]
+	orr	$t2,$t2,$i1,lsl#24
+	ldrb	$i3,[$tbl,$i3]
+	orr	$t2,$t2,$i2,lsl#16
+	ldr	$t1,[$t3],#4			@ rcon[i++]
+	orr	$t2,$t2,$i3,lsl#8
+	eor	$t2,$t2,$t1
+	eor	$s0,$s0,$t2			@ rk[4]=rk[0]^...
+	eor	$s1,$s1,$s0			@ rk[5]=rk[1]^rk[4]
+	str	$s0,[$key],#16
+	eor	$s2,$s2,$s1			@ rk[6]=rk[2]^rk[5]
+	str	$s1,[$key,#-12]
+	eor	$s3,$s3,$s2			@ rk[7]=rk[3]^rk[6]
+	str	$s2,[$key,#-8]
+	subs	$rounds,$rounds,#1
+	str	$s3,[$key,#-4]
+	bne	.L128_loop
+	sub	r2,$key,#176
+	b	.Ldone
+
+.Lnot128:
+	ldrb	$i2,[$rounds,#19]
+	ldrb	$t1,[$rounds,#18]
+	ldrb	$t2,[$rounds,#17]
+	ldrb	$t3,[$rounds,#16]
+	orr	$i2,$i2,$t1,lsl#8
+	ldrb	$i3,[$rounds,#23]
+	orr	$i2,$i2,$t2,lsl#16
+	ldrb	$t1,[$rounds,#22]
+	orr	$i2,$i2,$t3,lsl#24
+	ldrb	$t2,[$rounds,#21]
+	ldrb	$t3,[$rounds,#20]
+	orr	$i3,$i3,$t1,lsl#8
+	orr	$i3,$i3,$t2,lsl#16
+	str	$i2,[$key],#8
+	orr	$i3,$i3,$t3,lsl#24
+	str	$i3,[$key,#-4]
+
+	teq	lr,#192
+	bne	.Lnot192
+	mov	$rounds,#12
+	str	$rounds,[$key,#240-24]
+	add	$t3,$tbl,#256			@ rcon
+	mov	lr,#255
+	mov	$rounds,#8
+
+.L192_loop:
+	and	$t2,lr,$i3,lsr#24
+	and	$i1,lr,$i3,lsr#16
+	ldrb	$t2,[$tbl,$t2]
+	and	$i2,lr,$i3,lsr#8
+	ldrb	$i1,[$tbl,$i1]
+	and	$i3,lr,$i3
+	ldrb	$i2,[$tbl,$i2]
+	orr	$t2,$t2,$i1,lsl#24
+	ldrb	$i3,[$tbl,$i3]
+	orr	$t2,$t2,$i2,lsl#16
+	ldr	$t1,[$t3],#4			@ rcon[i++]
+	orr	$t2,$t2,$i3,lsl#8
+	eor	$i3,$t2,$t1
+	eor	$s0,$s0,$i3			@ rk[6]=rk[0]^...
+	eor	$s1,$s1,$s0			@ rk[7]=rk[1]^rk[6]
+	str	$s0,[$key],#24
+	eor	$s2,$s2,$s1			@ rk[8]=rk[2]^rk[7]
+	str	$s1,[$key,#-20]
+	eor	$s3,$s3,$s2			@ rk[9]=rk[3]^rk[8]
+	str	$s2,[$key,#-16]
+	subs	$rounds,$rounds,#1
+	str	$s3,[$key,#-12]
+	subeq	r2,$key,#216
+	beq	.Ldone
+
+	ldr	$i1,[$key,#-32]
+	ldr	$i2,[$key,#-28]
+	eor	$i1,$i1,$s3			@ rk[10]=rk[4]^rk[9]
+	eor	$i3,$i2,$i1			@ rk[11]=rk[5]^rk[10]
+	str	$i1,[$key,#-8]
+	str	$i3,[$key,#-4]
+	b	.L192_loop
+
+.Lnot192:
+	ldrb	$i2,[$rounds,#27]
+	ldrb	$t1,[$rounds,#26]
+	ldrb	$t2,[$rounds,#25]
+	ldrb	$t3,[$rounds,#24]
+	orr	$i2,$i2,$t1,lsl#8
+	ldrb	$i3,[$rounds,#31]
+	orr	$i2,$i2,$t2,lsl#16
+	ldrb	$t1,[$rounds,#30]
+	orr	$i2,$i2,$t3,lsl#24
+	ldrb	$t2,[$rounds,#29]
+	ldrb	$t3,[$rounds,#28]
+	orr	$i3,$i3,$t1,lsl#8
+	orr	$i3,$i3,$t2,lsl#16
+	str	$i2,[$key],#8
+	orr	$i3,$i3,$t3,lsl#24
+	str	$i3,[$key,#-4]
+
+	mov	$rounds,#14
+	str	$rounds,[$key,#240-32]
+	add	$t3,$tbl,#256			@ rcon
+	mov	lr,#255
+	mov	$rounds,#7
+
+.L256_loop:
+	and	$t2,lr,$i3,lsr#24
+	and	$i1,lr,$i3,lsr#16
+	ldrb	$t2,[$tbl,$t2]
+	and	$i2,lr,$i3,lsr#8
+	ldrb	$i1,[$tbl,$i1]
+	and	$i3,lr,$i3
+	ldrb	$i2,[$tbl,$i2]
+	orr	$t2,$t2,$i1,lsl#24
+	ldrb	$i3,[$tbl,$i3]
+	orr	$t2,$t2,$i2,lsl#16
+	ldr	$t1,[$t3],#4			@ rcon[i++]
+	orr	$t2,$t2,$i3,lsl#8
+	eor	$i3,$t2,$t1
+	eor	$s0,$s0,$i3			@ rk[8]=rk[0]^...
+	eor	$s1,$s1,$s0			@ rk[9]=rk[1]^rk[8]
+	str	$s0,[$key],#32
+	eor	$s2,$s2,$s1			@ rk[10]=rk[2]^rk[9]
+	str	$s1,[$key,#-28]
+	eor	$s3,$s3,$s2			@ rk[11]=rk[3]^rk[10]
+	str	$s2,[$key,#-24]
+	subs	$rounds,$rounds,#1
+	str	$s3,[$key,#-20]
+	subeq	r2,$key,#256
+	beq	.Ldone
+
+	and	$t2,lr,$s3
+	and	$i1,lr,$s3,lsr#8
+	ldrb	$t2,[$tbl,$t2]
+	and	$i2,lr,$s3,lsr#16
+	ldrb	$i1,[$tbl,$i1]
+	and	$i3,lr,$s3,lsr#24
+	ldrb	$i2,[$tbl,$i2]
+	orr	$t2,$t2,$i1,lsl#8
+	ldrb	$i3,[$tbl,$i3]
+	orr	$t2,$t2,$i2,lsl#16
+	ldr	$t1,[$key,#-48]
+	orr	$t2,$t2,$i3,lsl#24
+
+	ldr	$i1,[$key,#-44]
+	ldr	$i2,[$key,#-40]
+	eor	$t1,$t1,$t2			@ rk[12]=rk[4]^...
+	ldr	$i3,[$key,#-36]
+	eor	$i1,$i1,$t1			@ rk[13]=rk[5]^rk[12]
+	str	$t1,[$key,#-16]
+	eor	$i2,$i2,$i1			@ rk[14]=rk[6]^rk[13]
+	str	$i1,[$key,#-12]
+	eor	$i3,$i3,$i2			@ rk[15]=rk[7]^rk[14]
+	str	$i2,[$key,#-8]
+	str	$i3,[$key,#-4]
+	b	.L256_loop
+
+.Ldone:	mov	r0,#0
+	ldmia   sp!,{r4-r12,lr}
+.Labrt:	tst	lr,#1
+	moveq	pc,lr			@ be binary compatible with V4, yet
+	bx	lr			@ interoperable with Thumb ISA:-)
+.size	AES_set_encrypt_key,.-AES_set_encrypt_key
+
+.global AES_set_decrypt_key
+.type   AES_set_decrypt_key,%function
+.align	5
+AES_set_decrypt_key:
+	str	lr,[sp,#-4]!            @ push lr
+	bl	AES_set_encrypt_key
+	teq	r0,#0
+	ldrne	lr,[sp],#4              @ pop lr
+	bne	.Labrt
+
+	stmdb   sp!,{r4-r12}
+
+	ldr	$rounds,[r2,#240]	@ AES_set_encrypt_key preserves r2,
+	mov	$key,r2			@ which is AES_KEY *key
+	mov	$i1,r2
+	add	$i2,r2,$rounds,lsl#4
+
+.Linv:	ldr	$s0,[$i1]
+	ldr	$s1,[$i1,#4]
+	ldr	$s2,[$i1,#8]
+	ldr	$s3,[$i1,#12]
+	ldr	$t1,[$i2]
+	ldr	$t2,[$i2,#4]
+	ldr	$t3,[$i2,#8]
+	ldr	$i3,[$i2,#12]
+	str	$s0,[$i2],#-16
+	str	$s1,[$i2,#16+4]
+	str	$s2,[$i2,#16+8]
+	str	$s3,[$i2,#16+12]
+	str	$t1,[$i1],#16
+	str	$t2,[$i1,#-12]
+	str	$t3,[$i1,#-8]
+	str	$i3,[$i1,#-4]
+	teq	$i1,$i2
+	bne	.Linv
+___
+$mask80=$i1;
+$mask1b=$i2;
+$mask7f=$i3;
+$code.=<<___;
+	ldr	$s0,[$key,#16]!		@ prefetch tp1
+	mov	$mask80,#0x80
+	mov	$mask1b,#0x1b
+	orr	$mask80,$mask80,#0x8000
+	orr	$mask1b,$mask1b,#0x1b00
+	orr	$mask80,$mask80,$mask80,lsl#16
+	orr	$mask1b,$mask1b,$mask1b,lsl#16
+	sub	$rounds,$rounds,#1
+	mvn	$mask7f,$mask80
+	mov	$rounds,$rounds,lsl#2	@ (rounds-1)*4
+
+.Lmix:	and	$t1,$s0,$mask80
+	and	$s1,$s0,$mask7f
+	sub	$t1,$t1,$t1,lsr#7
+	and	$t1,$t1,$mask1b
+	eor	$s1,$t1,$s1,lsl#1	@ tp2
+
+	and	$t1,$s1,$mask80
+	and	$s2,$s1,$mask7f
+	sub	$t1,$t1,$t1,lsr#7
+	and	$t1,$t1,$mask1b
+	eor	$s2,$t1,$s2,lsl#1	@ tp4
+
+	and	$t1,$s2,$mask80
+	and	$s3,$s2,$mask7f
+	sub	$t1,$t1,$t1,lsr#7
+	and	$t1,$t1,$mask1b
+	eor	$s3,$t1,$s3,lsl#1	@ tp8
+
+	eor	$t1,$s1,$s2
+	eor	$t2,$s0,$s3		@ tp9
+	eor	$t1,$t1,$s3		@ tpe
+	eor	$t1,$t1,$s1,ror#24
+	eor	$t1,$t1,$t2,ror#24	@ ^= ROTATE(tpb=tp9^tp2,8)
+	eor	$t1,$t1,$s2,ror#16
+	eor	$t1,$t1,$t2,ror#16	@ ^= ROTATE(tpd=tp9^tp4,16)
+	eor	$t1,$t1,$t2,ror#8	@ ^= ROTATE(tp9,24)
+
+	ldr	$s0,[$key,#4]		@ prefetch tp1
+	str	$t1,[$key],#4
+	subs	$rounds,$rounds,#1
+	bne	.Lmix
+
+	mov	r0,#0
+	ldmia   sp!,{r4-r12,lr}
+	tst	lr,#1
+	moveq	pc,lr			@ be binary compatible with V4, yet
+	bx	lr			@ interoperable with Thumb ISA:-)
+.size	AES_set_decrypt_key,.-AES_set_decrypt_key
+
+.type	AES_Td,%object
+.align	5
+AES_Td:
+.word	0x51f4a750, 0x7e416553, 0x1a17a4c3, 0x3a275e96
+.word	0x3bab6bcb, 0x1f9d45f1, 0xacfa58ab, 0x4be30393
+.word	0x2030fa55, 0xad766df6, 0x88cc7691, 0xf5024c25
+.word	0x4fe5d7fc, 0xc52acbd7, 0x26354480, 0xb562a38f
+.word	0xdeb15a49, 0x25ba1b67, 0x45ea0e98, 0x5dfec0e1
+.word	0xc32f7502, 0x814cf012, 0x8d4697a3, 0x6bd3f9c6
+.word	0x038f5fe7, 0x15929c95, 0xbf6d7aeb, 0x955259da
+.word	0xd4be832d, 0x587421d3, 0x49e06929, 0x8ec9c844
+.word	0x75c2896a, 0xf48e7978, 0x99583e6b, 0x27b971dd
+.word	0xbee14fb6, 0xf088ad17, 0xc920ac66, 0x7dce3ab4
+.word	0x63df4a18, 0xe51a3182, 0x97513360, 0x62537f45
+.word	0xb16477e0, 0xbb6bae84, 0xfe81a01c, 0xf9082b94
+.word	0x70486858, 0x8f45fd19, 0x94de6c87, 0x527bf8b7
+.word	0xab73d323, 0x724b02e2, 0xe31f8f57, 0x6655ab2a
+.word	0xb2eb2807, 0x2fb5c203, 0x86c57b9a, 0xd33708a5
+.word	0x302887f2, 0x23bfa5b2, 0x02036aba, 0xed16825c
+.word	0x8acf1c2b, 0xa779b492, 0xf307f2f0, 0x4e69e2a1
+.word	0x65daf4cd, 0x0605bed5, 0xd134621f, 0xc4a6fe8a
+.word	0x342e539d, 0xa2f355a0, 0x058ae132, 0xa4f6eb75
+.word	0x0b83ec39, 0x4060efaa, 0x5e719f06, 0xbd6e1051
+.word	0x3e218af9, 0x96dd063d, 0xdd3e05ae, 0x4de6bd46
+.word	0x91548db5, 0x71c45d05, 0x0406d46f, 0x605015ff
+.word	0x1998fb24, 0xd6bde997, 0x894043cc, 0x67d99e77
+.word	0xb0e842bd, 0x07898b88, 0xe7195b38, 0x79c8eedb
+.word	0xa17c0a47, 0x7c420fe9, 0xf8841ec9, 0x00000000
+.word	0x09808683, 0x322bed48, 0x1e1170ac, 0x6c5a724e
+.word	0xfd0efffb, 0x0f853856, 0x3daed51e, 0x362d3927
+.word	0x0a0fd964, 0x685ca621, 0x9b5b54d1, 0x24362e3a
+.word	0x0c0a67b1, 0x9357e70f, 0xb4ee96d2, 0x1b9b919e
+.word	0x80c0c54f, 0x61dc20a2, 0x5a774b69, 0x1c121a16
+.word	0xe293ba0a, 0xc0a02ae5, 0x3c22e043, 0x121b171d
+.word	0x0e090d0b, 0xf28bc7ad, 0x2db6a8b9, 0x141ea9c8
+.word	0x57f11985, 0xaf75074c, 0xee99ddbb, 0xa37f60fd
+.word	0xf701269f, 0x5c72f5bc, 0x44663bc5, 0x5bfb7e34
+.word	0x8b432976, 0xcb23c6dc, 0xb6edfc68, 0xb8e4f163
+.word	0xd731dcca, 0x42638510, 0x13972240, 0x84c61120
+.word	0x854a247d, 0xd2bb3df8, 0xaef93211, 0xc729a16d
+.word	0x1d9e2f4b, 0xdcb230f3, 0x0d8652ec, 0x77c1e3d0
+.word	0x2bb3166c, 0xa970b999, 0x119448fa, 0x47e96422
+.word	0xa8fc8cc4, 0xa0f03f1a, 0x567d2cd8, 0x223390ef
+.word	0x87494ec7, 0xd938d1c1, 0x8ccaa2fe, 0x98d40b36
+.word	0xa6f581cf, 0xa57ade28, 0xdab78e26, 0x3fadbfa4
+.word	0x2c3a9de4, 0x5078920d, 0x6a5fcc9b, 0x547e4662
+.word	0xf68d13c2, 0x90d8b8e8, 0x2e39f75e, 0x82c3aff5
+.word	0x9f5d80be, 0x69d0937c, 0x6fd52da9, 0xcf2512b3
+.word	0xc8ac993b, 0x10187da7, 0xe89c636e, 0xdb3bbb7b
+.word	0xcd267809, 0x6e5918f4, 0xec9ab701, 0x834f9aa8
+.word	0xe6956e65, 0xaaffe67e, 0x21bccf08, 0xef15e8e6
+.word	0xbae79bd9, 0x4a6f36ce, 0xea9f09d4, 0x29b07cd6
+.word	0x31a4b2af, 0x2a3f2331, 0xc6a59430, 0x35a266c0
+.word	0x744ebc37, 0xfc82caa6, 0xe090d0b0, 0x33a7d815
+.word	0xf104984a, 0x41ecdaf7, 0x7fcd500e, 0x1791f62f
+.word	0x764dd68d, 0x43efb04d, 0xccaa4d54, 0xe49604df
+.word	0x9ed1b5e3, 0x4c6a881b, 0xc12c1fb8, 0x4665517f
+.word	0x9d5eea04, 0x018c355d, 0xfa877473, 0xfb0b412e
+.word	0xb3671d5a, 0x92dbd252, 0xe9105633, 0x6dd64713
+.word	0x9ad7618c, 0x37a10c7a, 0x59f8148e, 0xeb133c89
+.word	0xcea927ee, 0xb761c935, 0xe11ce5ed, 0x7a47b13c
+.word	0x9cd2df59, 0x55f2733f, 0x1814ce79, 0x73c737bf
+.word	0x53f7cdea, 0x5ffdaa5b, 0xdf3d6f14, 0x7844db86
+.word	0xcaaff381, 0xb968c43e, 0x3824342c, 0xc2a3405f
+.word	0x161dc372, 0xbce2250c, 0x283c498b, 0xff0d9541
+.word	0x39a80171, 0x080cb3de, 0xd8b4e49c, 0x6456c190
+.word	0x7bcb8461, 0xd532b670, 0x486c5c74, 0xd0b85742
+@ Td4[256]
+.byte	0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38
+.byte	0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb
+.byte	0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87
+.byte	0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb
+.byte	0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d
+.byte	0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e
+.byte	0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2
+.byte	0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25
+.byte	0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16
+.byte	0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92
+.byte	0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda
+.byte	0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84
+.byte	0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a
+.byte	0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06
+.byte	0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02
+.byte	0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b
+.byte	0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea
+.byte	0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73
+.byte	0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85
+.byte	0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e
+.byte	0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89
+.byte	0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b
+.byte	0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20
+.byte	0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4
+.byte	0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31
+.byte	0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f
+.byte	0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d
+.byte	0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef
+.byte	0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0
+.byte	0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61
+.byte	0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26
+.byte	0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d
+.size	AES_Td,.-AES_Td
+
+@ void AES_decrypt(const unsigned char *in, unsigned char *out,
+@ 		 const AES_KEY *key) {
+.global AES_decrypt
+.type   AES_decrypt,%function
+.align	5
+AES_decrypt:
+	sub	r3,pc,#8		@ AES_decrypt
+	stmdb   sp!,{r1,r4-r12,lr}
+	mov	$rounds,r0		@ inp
+	mov	$key,r2
+	sub	$tbl,r3,#AES_decrypt-AES_Td		@ Td
+
+	ldrb	$s0,[$rounds,#3]	@ load input data in endian-neutral
+	ldrb	$t1,[$rounds,#2]	@ manner...
+	ldrb	$t2,[$rounds,#1]
+	ldrb	$t3,[$rounds,#0]
+	orr	$s0,$s0,$t1,lsl#8
+	ldrb	$s1,[$rounds,#7]
+	orr	$s0,$s0,$t2,lsl#16
+	ldrb	$t1,[$rounds,#6]
+	orr	$s0,$s0,$t3,lsl#24
+	ldrb	$t2,[$rounds,#5]
+	ldrb	$t3,[$rounds,#4]
+	orr	$s1,$s1,$t1,lsl#8
+	ldrb	$s2,[$rounds,#11]
+	orr	$s1,$s1,$t2,lsl#16
+	ldrb	$t1,[$rounds,#10]
+	orr	$s1,$s1,$t3,lsl#24
+	ldrb	$t2,[$rounds,#9]
+	ldrb	$t3,[$rounds,#8]
+	orr	$s2,$s2,$t1,lsl#8
+	ldrb	$s3,[$rounds,#15]
+	orr	$s2,$s2,$t2,lsl#16
+	ldrb	$t1,[$rounds,#14]
+	orr	$s2,$s2,$t3,lsl#24
+	ldrb	$t2,[$rounds,#13]
+	ldrb	$t3,[$rounds,#12]
+	orr	$s3,$s3,$t1,lsl#8
+	orr	$s3,$s3,$t2,lsl#16
+	orr	$s3,$s3,$t3,lsl#24
+
+	bl	_armv4_AES_decrypt
+
+	ldr	$rounds,[sp],#4		@ pop out
+	mov	$t1,$s0,lsr#24		@ write output in endian-neutral
+	mov	$t2,$s0,lsr#16		@ manner...
+	mov	$t3,$s0,lsr#8
+	strb	$t1,[$rounds,#0]
+	strb	$t2,[$rounds,#1]
+	mov	$t1,$s1,lsr#24
+	strb	$t3,[$rounds,#2]
+	mov	$t2,$s1,lsr#16
+	strb	$s0,[$rounds,#3]
+	mov	$t3,$s1,lsr#8
+	strb	$t1,[$rounds,#4]
+	strb	$t2,[$rounds,#5]
+	mov	$t1,$s2,lsr#24
+	strb	$t3,[$rounds,#6]
+	mov	$t2,$s2,lsr#16
+	strb	$s1,[$rounds,#7]
+	mov	$t3,$s2,lsr#8
+	strb	$t1,[$rounds,#8]
+	strb	$t2,[$rounds,#9]
+	mov	$t1,$s3,lsr#24
+	strb	$t3,[$rounds,#10]
+	mov	$t2,$s3,lsr#16
+	strb	$s2,[$rounds,#11]
+	mov	$t3,$s3,lsr#8
+	strb	$t1,[$rounds,#12]
+	strb	$t2,[$rounds,#13]
+	strb	$t3,[$rounds,#14]
+	strb	$s3,[$rounds,#15]
+
+	ldmia   sp!,{r4-r12,lr}
+	tst	lr,#1
+	moveq	pc,lr			@ be binary compatible with V4, yet
+	bx	lr			@ interoperable with Thumb ISA:-)
+.size	AES_decrypt,.-AES_decrypt
+
+.type   _armv4_AES_decrypt,%function
+.align	2
+_armv4_AES_decrypt:
+	str	lr,[sp,#-4]!		@ push lr
+	ldmia	$key!,{$t1-$i1}
+	eor	$s0,$s0,$t1
+	ldr	$rounds,[$key,#240-16]
+	eor	$s1,$s1,$t2
+	eor	$s2,$s2,$t3
+	eor	$s3,$s3,$i1
+	sub	$rounds,$rounds,#1
+	mov	lr,#255
+
+	and	$i1,lr,$s0,lsr#16
+	and	$i2,lr,$s0,lsr#8
+	and	$i3,lr,$s0
+	mov	$s0,$s0,lsr#24
+.Ldec_loop:
+	ldr	$t1,[$tbl,$i1,lsl#2]	@ Td1[s0>>16]
+	and	$i1,lr,$s1		@ i0
+	ldr	$t2,[$tbl,$i2,lsl#2]	@ Td2[s0>>8]
+	and	$i2,lr,$s1,lsr#16
+	ldr	$t3,[$tbl,$i3,lsl#2]	@ Td3[s0>>0]
+	and	$i3,lr,$s1,lsr#8
+	ldr	$s0,[$tbl,$s0,lsl#2]	@ Td0[s0>>24]
+	mov	$s1,$s1,lsr#24
+
+	ldr	$i1,[$tbl,$i1,lsl#2]	@ Td3[s1>>0]
+	ldr	$i2,[$tbl,$i2,lsl#2]	@ Td1[s1>>16]
+	ldr	$i3,[$tbl,$i3,lsl#2]	@ Td2[s1>>8]
+	eor	$s0,$s0,$i1,ror#24
+	ldr	$s1,[$tbl,$s1,lsl#2]	@ Td0[s1>>24]
+	and	$i1,lr,$s2,lsr#8	@ i0
+	eor	$t2,$i2,$t2,ror#8
+	and	$i2,lr,$s2		@ i1
+	eor	$t3,$i3,$t3,ror#8
+	and	$i3,lr,$s2,lsr#16
+	eor	$s1,$s1,$t1,ror#8
+	ldr	$i1,[$tbl,$i1,lsl#2]	@ Td2[s2>>8]
+	mov	$s2,$s2,lsr#24
+
+	ldr	$i2,[$tbl,$i2,lsl#2]	@ Td3[s2>>0]
+	ldr	$i3,[$tbl,$i3,lsl#2]	@ Td1[s2>>16]
+	eor	$s0,$s0,$i1,ror#16
+	ldr	$s2,[$tbl,$s2,lsl#2]	@ Td0[s2>>24]
+	and	$i1,lr,$s3,lsr#16	@ i0
+	eor	$s1,$s1,$i2,ror#24
+	and	$i2,lr,$s3,lsr#8	@ i1
+	eor	$t3,$i3,$t3,ror#8
+	and	$i3,lr,$s3		@ i2
+	eor	$s2,$s2,$t2,ror#8
+	ldr	$i1,[$tbl,$i1,lsl#2]	@ Td1[s3>>16]
+	mov	$s3,$s3,lsr#24
+
+	ldr	$i2,[$tbl,$i2,lsl#2]	@ Td2[s3>>8]
+	ldr	$i3,[$tbl,$i3,lsl#2]	@ Td3[s3>>0]
+	eor	$s0,$s0,$i1,ror#8
+	ldr	$s3,[$tbl,$s3,lsl#2]	@ Td0[s3>>24]
+	eor	$s1,$s1,$i2,ror#16
+	eor	$s2,$s2,$i3,ror#24
+	ldr	$i1,[$key],#16
+	eor	$s3,$s3,$t3,ror#8
+
+	ldr	$t1,[$key,#-12]
+	ldr	$t2,[$key,#-8]
+	eor	$s0,$s0,$i1
+	ldr	$t3,[$key,#-4]
+	and	$i1,lr,$s0,lsr#16
+	eor	$s1,$s1,$t1
+	and	$i2,lr,$s0,lsr#8
+	eor	$s2,$s2,$t2
+	and	$i3,lr,$s0
+	eor	$s3,$s3,$t3
+	mov	$s0,$s0,lsr#24
+
+	subs	$rounds,$rounds,#1
+	bne	.Ldec_loop
+
+	add	$tbl,$tbl,#1024
+
+	ldr	$t2,[$tbl,#0]		@ prefetch Td4
+	ldr	$t3,[$tbl,#32]
+	ldr	$t1,[$tbl,#64]
+	ldr	$t2,[$tbl,#96]
+	ldr	$t3,[$tbl,#128]
+	ldr	$t1,[$tbl,#160]
+	ldr	$t2,[$tbl,#192]
+	ldr	$t3,[$tbl,#224]
+
+	ldrb	$s0,[$tbl,$s0]		@ Td4[s0>>24]
+	ldrb	$t1,[$tbl,$i1]		@ Td4[s0>>16]
+	and	$i1,lr,$s1		@ i0
+	ldrb	$t2,[$tbl,$i2]		@ Td4[s0>>8]
+	and	$i2,lr,$s1,lsr#16
+	ldrb	$t3,[$tbl,$i3]		@ Td4[s0>>0]
+	and	$i3,lr,$s1,lsr#8
+
+	ldrb	$i1,[$tbl,$i1]		@ Td4[s1>>0]
+	ldrb	$s1,[$tbl,$s1,lsr#24]	@ Td4[s1>>24]
+	ldrb	$i2,[$tbl,$i2]		@ Td4[s1>>16]
+	eor	$s0,$i1,$s0,lsl#24
+	ldrb	$i3,[$tbl,$i3]		@ Td4[s1>>8]
+	eor	$s1,$t1,$s1,lsl#8
+	and	$i1,lr,$s2,lsr#8	@ i0
+	eor	$t2,$t2,$i2,lsl#8
+	and	$i2,lr,$s2		@ i1
+	eor	$t3,$t3,$i3,lsl#8
+	ldrb	$i1,[$tbl,$i1]		@ Td4[s2>>8]
+	and	$i3,lr,$s2,lsr#16
+
+	ldrb	$i2,[$tbl,$i2]		@ Td4[s2>>0]
+	ldrb	$s2,[$tbl,$s2,lsr#24]	@ Td4[s2>>24]
+	eor	$s0,$s0,$i1,lsl#8
+	ldrb	$i3,[$tbl,$i3]		@ Td4[s2>>16]
+	eor	$s1,$i2,$s1,lsl#16
+	and	$i1,lr,$s3,lsr#16	@ i0
+	eor	$s2,$t2,$s2,lsl#16
+	and	$i2,lr,$s3,lsr#8	@ i1
+	eor	$t3,$t3,$i3,lsl#16
+	ldrb	$i1,[$tbl,$i1]		@ Td4[s3>>16]
+	and	$i3,lr,$s3		@ i2
+
+	ldrb	$i2,[$tbl,$i2]		@ Td4[s3>>8]
+	ldrb	$i3,[$tbl,$i3]		@ Td4[s3>>0]
+	ldrb	$s3,[$tbl,$s3,lsr#24]	@ Td4[s3>>24]
+	eor	$s0,$s0,$i1,lsl#16
+	ldr	$i1,[$key,#0]
+	eor	$s1,$s1,$i2,lsl#8
+	ldr	$t1,[$key,#4]
+	eor	$s2,$i3,$s2,lsl#8
+	ldr	$t2,[$key,#8]
+	eor	$s3,$t3,$s3,lsl#24
+	ldr	$t3,[$key,#12]
+
+	eor	$s0,$s0,$i1
+	eor	$s1,$s1,$t1
+	eor	$s2,$s2,$t2
+	eor	$s3,$s3,$t3
+
+	sub	$tbl,$tbl,#1024
+	ldr	pc,[sp],#4		@ pop and return
+.size	_armv4_AES_decrypt,.-_armv4_AES_decrypt
+.asciz	"AES for ARMv4, CRYPTOGAMS by <appro\@openssl.org>"
+.align	2
+___
+
+$code =~ s/\bbx\s+lr\b/.word\t0xe12fff1e/gm;	# make it possible to compile with -march=armv4
+print $code;
+close STDOUT;	# enforce flush
diff --git a/openssl/crypto/aes/asm/aes-armv4.s b/openssl/crypto/aes/asm/aes-armv4.s
new file mode 100644
index 00000000..27c681c7
--- /dev/null
+++ b/openssl/crypto/aes/asm/aes-armv4.s
@@ -0,0 +1,972 @@
+.text
+.code	32
+
+.type	AES_Te,%object
+.align	5
+AES_Te:
+.word	0xc66363a5, 0xf87c7c84, 0xee777799, 0xf67b7b8d
+.word	0xfff2f20d, 0xd66b6bbd, 0xde6f6fb1, 0x91c5c554
+.word	0x60303050, 0x02010103, 0xce6767a9, 0x562b2b7d
+.word	0xe7fefe19, 0xb5d7d762, 0x4dababe6, 0xec76769a
+.word	0x8fcaca45, 0x1f82829d, 0x89c9c940, 0xfa7d7d87
+.word	0xeffafa15, 0xb25959eb, 0x8e4747c9, 0xfbf0f00b
+.word	0x41adadec, 0xb3d4d467, 0x5fa2a2fd, 0x45afafea
+.word	0x239c9cbf, 0x53a4a4f7, 0xe4727296, 0x9bc0c05b
+.word	0x75b7b7c2, 0xe1fdfd1c, 0x3d9393ae, 0x4c26266a
+.word	0x6c36365a, 0x7e3f3f41, 0xf5f7f702, 0x83cccc4f
+.word	0x6834345c, 0x51a5a5f4, 0xd1e5e534, 0xf9f1f108
+.word	0xe2717193, 0xabd8d873, 0x62313153, 0x2a15153f
+.word	0x0804040c, 0x95c7c752, 0x46232365, 0x9dc3c35e
+.word	0x30181828, 0x379696a1, 0x0a05050f, 0x2f9a9ab5
+.word	0x0e070709, 0x24121236, 0x1b80809b, 0xdfe2e23d
+.word	0xcdebeb26, 0x4e272769, 0x7fb2b2cd, 0xea75759f
+.word	0x1209091b, 0x1d83839e, 0x582c2c74, 0x341a1a2e
+.word	0x361b1b2d, 0xdc6e6eb2, 0xb45a5aee, 0x5ba0a0fb
+.word	0xa45252f6, 0x763b3b4d, 0xb7d6d661, 0x7db3b3ce
+.word	0x5229297b, 0xdde3e33e, 0x5e2f2f71, 0x13848497
+.word	0xa65353f5, 0xb9d1d168, 0x00000000, 0xc1eded2c
+.word	0x40202060, 0xe3fcfc1f, 0x79b1b1c8, 0xb65b5bed
+.word	0xd46a6abe, 0x8dcbcb46, 0x67bebed9, 0x7239394b
+.word	0x944a4ade, 0x984c4cd4, 0xb05858e8, 0x85cfcf4a
+.word	0xbbd0d06b, 0xc5efef2a, 0x4faaaae5, 0xedfbfb16
+.word	0x864343c5, 0x9a4d4dd7, 0x66333355, 0x11858594
+.word	0x8a4545cf, 0xe9f9f910, 0x04020206, 0xfe7f7f81
+.word	0xa05050f0, 0x783c3c44, 0x259f9fba, 0x4ba8a8e3
+.word	0xa25151f3, 0x5da3a3fe, 0x804040c0, 0x058f8f8a
+.word	0x3f9292ad, 0x219d9dbc, 0x70383848, 0xf1f5f504
+.word	0x63bcbcdf, 0x77b6b6c1, 0xafdada75, 0x42212163
+.word	0x20101030, 0xe5ffff1a, 0xfdf3f30e, 0xbfd2d26d
+.word	0x81cdcd4c, 0x180c0c14, 0x26131335, 0xc3ecec2f
+.word	0xbe5f5fe1, 0x359797a2, 0x884444cc, 0x2e171739
+.word	0x93c4c457, 0x55a7a7f2, 0xfc7e7e82, 0x7a3d3d47
+.word	0xc86464ac, 0xba5d5de7, 0x3219192b, 0xe6737395
+.word	0xc06060a0, 0x19818198, 0x9e4f4fd1, 0xa3dcdc7f
+.word	0x44222266, 0x542a2a7e, 0x3b9090ab, 0x0b888883
+.word	0x8c4646ca, 0xc7eeee29, 0x6bb8b8d3, 0x2814143c
+.word	0xa7dede79, 0xbc5e5ee2, 0x160b0b1d, 0xaddbdb76
+.word	0xdbe0e03b, 0x64323256, 0x743a3a4e, 0x140a0a1e
+.word	0x924949db, 0x0c06060a, 0x4824246c, 0xb85c5ce4
+.word	0x9fc2c25d, 0xbdd3d36e, 0x43acacef, 0xc46262a6
+.word	0x399191a8, 0x319595a4, 0xd3e4e437, 0xf279798b
+.word	0xd5e7e732, 0x8bc8c843, 0x6e373759, 0xda6d6db7
+.word	0x018d8d8c, 0xb1d5d564, 0x9c4e4ed2, 0x49a9a9e0
+.word	0xd86c6cb4, 0xac5656fa, 0xf3f4f407, 0xcfeaea25
+.word	0xca6565af, 0xf47a7a8e, 0x47aeaee9, 0x10080818
+.word	0x6fbabad5, 0xf0787888, 0x4a25256f, 0x5c2e2e72
+.word	0x381c1c24, 0x57a6a6f1, 0x73b4b4c7, 0x97c6c651
+.word	0xcbe8e823, 0xa1dddd7c, 0xe874749c, 0x3e1f1f21
+.word	0x964b4bdd, 0x61bdbddc, 0x0d8b8b86, 0x0f8a8a85
+.word	0xe0707090, 0x7c3e3e42, 0x71b5b5c4, 0xcc6666aa
+.word	0x904848d8, 0x06030305, 0xf7f6f601, 0x1c0e0e12
+.word	0xc26161a3, 0x6a35355f, 0xae5757f9, 0x69b9b9d0
+.word	0x17868691, 0x99c1c158, 0x3a1d1d27, 0x279e9eb9
+.word	0xd9e1e138, 0xebf8f813, 0x2b9898b3, 0x22111133
+.word	0xd26969bb, 0xa9d9d970, 0x078e8e89, 0x339494a7
+.word	0x2d9b9bb6, 0x3c1e1e22, 0x15878792, 0xc9e9e920
+.word	0x87cece49, 0xaa5555ff, 0x50282878, 0xa5dfdf7a
+.word	0x038c8c8f, 0x59a1a1f8, 0x09898980, 0x1a0d0d17
+.word	0x65bfbfda, 0xd7e6e631, 0x844242c6, 0xd06868b8
+.word	0x824141c3, 0x299999b0, 0x5a2d2d77, 0x1e0f0f11
+.word	0x7bb0b0cb, 0xa85454fc, 0x6dbbbbd6, 0x2c16163a
+@ Te4[256]
+.byte	0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5
+.byte	0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76
+.byte	0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0
+.byte	0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0
+.byte	0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc
+.byte	0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15
+.byte	0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a
+.byte	0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75
+.byte	0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0
+.byte	0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84
+.byte	0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b
+.byte	0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf
+.byte	0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85
+.byte	0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8
+.byte	0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5
+.byte	0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2
+.byte	0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17
+.byte	0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73
+.byte	0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88
+.byte	0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb
+.byte	0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c
+.byte	0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79
+.byte	0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9
+.byte	0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08
+.byte	0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6
+.byte	0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a
+.byte	0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e
+.byte	0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e
+.byte	0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94
+.byte	0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf
+.byte	0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68
+.byte	0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16
+@ rcon[]
+.word	0x01000000, 0x02000000, 0x04000000, 0x08000000
+.word	0x10000000, 0x20000000, 0x40000000, 0x80000000
+.word	0x1B000000, 0x36000000, 0, 0, 0, 0, 0, 0
+.size	AES_Te,.-AES_Te
+
+@ void AES_encrypt(const unsigned char *in, unsigned char *out,
+@ 		 const AES_KEY *key) {
+.global AES_encrypt
+.type   AES_encrypt,%function
+.align	5
+AES_encrypt:
+	sub	r3,pc,#8		@ AES_encrypt
+	stmdb   sp!,{r1,r4-r12,lr}
+	mov	r12,r0		@ inp
+	mov	r11,r2
+	sub	r10,r3,#AES_encrypt-AES_Te	@ Te
+
+	ldrb	r0,[r12,#3]	@ load input data in endian-neutral
+	ldrb	r4,[r12,#2]	@ manner...
+	ldrb	r5,[r12,#1]
+	ldrb	r6,[r12,#0]
+	orr	r0,r0,r4,lsl#8
+	ldrb	r1,[r12,#7]
+	orr	r0,r0,r5,lsl#16
+	ldrb	r4,[r12,#6]
+	orr	r0,r0,r6,lsl#24
+	ldrb	r5,[r12,#5]
+	ldrb	r6,[r12,#4]
+	orr	r1,r1,r4,lsl#8
+	ldrb	r2,[r12,#11]
+	orr	r1,r1,r5,lsl#16
+	ldrb	r4,[r12,#10]
+	orr	r1,r1,r6,lsl#24
+	ldrb	r5,[r12,#9]
+	ldrb	r6,[r12,#8]
+	orr	r2,r2,r4,lsl#8
+	ldrb	r3,[r12,#15]
+	orr	r2,r2,r5,lsl#16
+	ldrb	r4,[r12,#14]
+	orr	r2,r2,r6,lsl#24
+	ldrb	r5,[r12,#13]
+	ldrb	r6,[r12,#12]
+	orr	r3,r3,r4,lsl#8
+	orr	r3,r3,r5,lsl#16
+	orr	r3,r3,r6,lsl#24
+
+	bl	_armv4_AES_encrypt
+
+	ldr	r12,[sp],#4		@ pop out
+	mov	r4,r0,lsr#24		@ write output in endian-neutral
+	mov	r5,r0,lsr#16		@ manner...
+	mov	r6,r0,lsr#8
+	strb	r4,[r12,#0]
+	strb	r5,[r12,#1]
+	mov	r4,r1,lsr#24
+	strb	r6,[r12,#2]
+	mov	r5,r1,lsr#16
+	strb	r0,[r12,#3]
+	mov	r6,r1,lsr#8
+	strb	r4,[r12,#4]
+	strb	r5,[r12,#5]
+	mov	r4,r2,lsr#24
+	strb	r6,[r12,#6]
+	mov	r5,r2,lsr#16
+	strb	r1,[r12,#7]
+	mov	r6,r2,lsr#8
+	strb	r4,[r12,#8]
+	strb	r5,[r12,#9]
+	mov	r4,r3,lsr#24
+	strb	r6,[r12,#10]
+	mov	r5,r3,lsr#16
+	strb	r2,[r12,#11]
+	mov	r6,r3,lsr#8
+	strb	r4,[r12,#12]
+	strb	r5,[r12,#13]
+	strb	r6,[r12,#14]
+	strb	r3,[r12,#15]
+
+	ldmia   sp!,{r4-r12,lr}
+	tst	lr,#1
+	moveq	pc,lr			@ be binary compatible with V4, yet
+	.word	0xe12fff1e			@ interoperable with Thumb ISA:-)
+.size	AES_encrypt,.-AES_encrypt
+
+.type   _armv4_AES_encrypt,%function
+.align	2
+_armv4_AES_encrypt:
+	str	lr,[sp,#-4]!		@ push lr
+	ldmia	r11!,{r4-r7}
+	eor	r0,r0,r4
+	ldr	r12,[r11,#240-16]
+	eor	r1,r1,r5
+	eor	r2,r2,r6
+	eor	r3,r3,r7
+	sub	r12,r12,#1
+	mov	lr,#255
+
+	and	r7,lr,r0
+	and	r8,lr,r0,lsr#8
+	and	r9,lr,r0,lsr#16
+	mov	r0,r0,lsr#24
+.Lenc_loop:
+	ldr	r4,[r10,r7,lsl#2]	@ Te3[s0>>0]
+	and	r7,lr,r1,lsr#16	@ i0
+	ldr	r5,[r10,r8,lsl#2]	@ Te2[s0>>8]
+	and	r8,lr,r1
+	ldr	r6,[r10,r9,lsl#2]	@ Te1[s0>>16]
+	and	r9,lr,r1,lsr#8
+	ldr	r0,[r10,r0,lsl#2]	@ Te0[s0>>24]
+	mov	r1,r1,lsr#24
+
+	ldr	r7,[r10,r7,lsl#2]	@ Te1[s1>>16]
+	ldr	r8,[r10,r8,lsl#2]	@ Te3[s1>>0]
+	ldr	r9,[r10,r9,lsl#2]	@ Te2[s1>>8]
+	eor	r0,r0,r7,ror#8
+	ldr	r1,[r10,r1,lsl#2]	@ Te0[s1>>24]
+	and	r7,lr,r2,lsr#8	@ i0
+	eor	r5,r5,r8,ror#8
+	and	r8,lr,r2,lsr#16	@ i1
+	eor	r6,r6,r9,ror#8
+	and	r9,lr,r2
+	eor	r1,r1,r4,ror#24
+	ldr	r7,[r10,r7,lsl#2]	@ Te2[s2>>8]
+	mov	r2,r2,lsr#24
+
+	ldr	r8,[r10,r8,lsl#2]	@ Te1[s2>>16]
+	ldr	r9,[r10,r9,lsl#2]	@ Te3[s2>>0]
+	eor	r0,r0,r7,ror#16
+	ldr	r2,[r10,r2,lsl#2]	@ Te0[s2>>24]
+	and	r7,lr,r3		@ i0
+	eor	r1,r1,r8,ror#8
+	and	r8,lr,r3,lsr#8	@ i1
+	eor	r6,r6,r9,ror#16
+	and	r9,lr,r3,lsr#16	@ i2
+	eor	r2,r2,r5,ror#16
+	ldr	r7,[r10,r7,lsl#2]	@ Te3[s3>>0]
+	mov	r3,r3,lsr#24
+
+	ldr	r8,[r10,r8,lsl#2]	@ Te2[s3>>8]
+	ldr	r9,[r10,r9,lsl#2]	@ Te1[s3>>16]
+	eor	r0,r0,r7,ror#24
+	ldr	r3,[r10,r3,lsl#2]	@ Te0[s3>>24]
+	eor	r1,r1,r8,ror#16
+	ldr	r7,[r11],#16
+	eor	r2,r2,r9,ror#8
+	ldr	r4,[r11,#-12]
+	eor	r3,r3,r6,ror#8
+
+	ldr	r5,[r11,#-8]
+	eor	r0,r0,r7
+	ldr	r6,[r11,#-4]
+	and	r7,lr,r0
+	eor	r1,r1,r4
+	and	r8,lr,r0,lsr#8
+	eor	r2,r2,r5
+	and	r9,lr,r0,lsr#16
+	eor	r3,r3,r6
+	mov	r0,r0,lsr#24
+
+	subs	r12,r12,#1
+	bne	.Lenc_loop
+
+	add	r10,r10,#2
+
+	ldrb	r4,[r10,r7,lsl#2]	@ Te4[s0>>0]
+	and	r7,lr,r1,lsr#16	@ i0
+	ldrb	r5,[r10,r8,lsl#2]	@ Te4[s0>>8]
+	and	r8,lr,r1
+	ldrb	r6,[r10,r9,lsl#2]	@ Te4[s0>>16]
+	and	r9,lr,r1,lsr#8
+	ldrb	r0,[r10,r0,lsl#2]	@ Te4[s0>>24]
+	mov	r1,r1,lsr#24
+
+	ldrb	r7,[r10,r7,lsl#2]	@ Te4[s1>>16]
+	ldrb	r8,[r10,r8,lsl#2]	@ Te4[s1>>0]
+	ldrb	r9,[r10,r9,lsl#2]	@ Te4[s1>>8]
+	eor	r0,r7,r0,lsl#8
+	ldrb	r1,[r10,r1,lsl#2]	@ Te4[s1>>24]
+	and	r7,lr,r2,lsr#8	@ i0
+	eor	r5,r8,r5,lsl#8
+	and	r8,lr,r2,lsr#16	@ i1
+	eor	r6,r9,r6,lsl#8
+	and	r9,lr,r2
+	eor	r1,r4,r1,lsl#24
+	ldrb	r7,[r10,r7,lsl#2]	@ Te4[s2>>8]
+	mov	r2,r2,lsr#24
+
+	ldrb	r8,[r10,r8,lsl#2]	@ Te4[s2>>16]
+	ldrb	r9,[r10,r9,lsl#2]	@ Te4[s2>>0]
+	eor	r0,r7,r0,lsl#8
+	ldrb	r2,[r10,r2,lsl#2]	@ Te4[s2>>24]
+	and	r7,lr,r3		@ i0
+	eor	r1,r1,r8,lsl#16
+	and	r8,lr,r3,lsr#8	@ i1
+	eor	r6,r9,r6,lsl#8
+	and	r9,lr,r3,lsr#16	@ i2
+	eor	r2,r5,r2,lsl#24
+	ldrb	r7,[r10,r7,lsl#2]	@ Te4[s3>>0]
+	mov	r3,r3,lsr#24
+
+	ldrb	r8,[r10,r8,lsl#2]	@ Te4[s3>>8]
+	ldrb	r9,[r10,r9,lsl#2]	@ Te4[s3>>16]
+	eor	r0,r7,r0,lsl#8
+	ldrb	r3,[r10,r3,lsl#2]	@ Te4[s3>>24]
+	ldr	r7,[r11,#0]
+	eor	r1,r1,r8,lsl#8
+	ldr	r4,[r11,#4]
+	eor	r2,r2,r9,lsl#16
+	ldr	r5,[r11,#8]
+	eor	r3,r6,r3,lsl#24
+	ldr	r6,[r11,#12]
+
+	eor	r0,r0,r7
+	eor	r1,r1,r4
+	eor	r2,r2,r5
+	eor	r3,r3,r6
+
+	sub	r10,r10,#2
+	ldr	pc,[sp],#4		@ pop and return
+.size	_armv4_AES_encrypt,.-_armv4_AES_encrypt
+
+.global AES_set_encrypt_key
+.type   AES_set_encrypt_key,%function
+.align	5
+AES_set_encrypt_key:
+	sub	r3,pc,#8		@ AES_set_encrypt_key
+	teq	r0,#0
+	moveq	r0,#-1
+	beq	.Labrt
+	teq	r2,#0
+	moveq	r0,#-1
+	beq	.Labrt
+
+	teq	r1,#128
+	beq	.Lok
+	teq	r1,#192
+	beq	.Lok
+	teq	r1,#256
+	movne	r0,#-1
+	bne	.Labrt
+
+.Lok:	stmdb   sp!,{r4-r12,lr}
+	sub	r10,r3,#AES_set_encrypt_key-AES_Te-1024	@ Te4
+
+	mov	r12,r0		@ inp
+	mov	lr,r1			@ bits
+	mov	r11,r2			@ key
+
+	ldrb	r0,[r12,#3]	@ load input data in endian-neutral
+	ldrb	r4,[r12,#2]	@ manner...
+	ldrb	r5,[r12,#1]
+	ldrb	r6,[r12,#0]
+	orr	r0,r0,r4,lsl#8
+	ldrb	r1,[r12,#7]
+	orr	r0,r0,r5,lsl#16
+	ldrb	r4,[r12,#6]
+	orr	r0,r0,r6,lsl#24
+	ldrb	r5,[r12,#5]
+	ldrb	r6,[r12,#4]
+	orr	r1,r1,r4,lsl#8
+	ldrb	r2,[r12,#11]
+	orr	r1,r1,r5,lsl#16
+	ldrb	r4,[r12,#10]
+	orr	r1,r1,r6,lsl#24
+	ldrb	r5,[r12,#9]
+	ldrb	r6,[r12,#8]
+	orr	r2,r2,r4,lsl#8
+	ldrb	r3,[r12,#15]
+	orr	r2,r2,r5,lsl#16
+	ldrb	r4,[r12,#14]
+	orr	r2,r2,r6,lsl#24
+	ldrb	r5,[r12,#13]
+	ldrb	r6,[r12,#12]
+	orr	r3,r3,r4,lsl#8
+	str	r0,[r11],#16
+	orr	r3,r3,r5,lsl#16
+	str	r1,[r11,#-12]
+	orr	r3,r3,r6,lsl#24
+	str	r2,[r11,#-8]
+	str	r3,[r11,#-4]
+
+	teq	lr,#128
+	bne	.Lnot128
+	mov	r12,#10
+	str	r12,[r11,#240-16]
+	add	r6,r10,#256			@ rcon
+	mov	lr,#255
+
+.L128_loop:
+	and	r5,lr,r3,lsr#24
+	and	r7,lr,r3,lsr#16
+	ldrb	r5,[r10,r5]
+	and	r8,lr,r3,lsr#8
+	ldrb	r7,[r10,r7]
+	and	r9,lr,r3
+	ldrb	r8,[r10,r8]
+	orr	r5,r5,r7,lsl#24
+	ldrb	r9,[r10,r9]
+	orr	r5,r5,r8,lsl#16
+	ldr	r4,[r6],#4			@ rcon[i++]
+	orr	r5,r5,r9,lsl#8
+	eor	r5,r5,r4
+	eor	r0,r0,r5			@ rk[4]=rk[0]^...
+	eor	r1,r1,r0			@ rk[5]=rk[1]^rk[4]
+	str	r0,[r11],#16
+	eor	r2,r2,r1			@ rk[6]=rk[2]^rk[5]
+	str	r1,[r11,#-12]
+	eor	r3,r3,r2			@ rk[7]=rk[3]^rk[6]
+	str	r2,[r11,#-8]
+	subs	r12,r12,#1
+	str	r3,[r11,#-4]
+	bne	.L128_loop
+	sub	r2,r11,#176
+	b	.Ldone
+
+.Lnot128:
+	ldrb	r8,[r12,#19]
+	ldrb	r4,[r12,#18]
+	ldrb	r5,[r12,#17]
+	ldrb	r6,[r12,#16]
+	orr	r8,r8,r4,lsl#8
+	ldrb	r9,[r12,#23]
+	orr	r8,r8,r5,lsl#16
+	ldrb	r4,[r12,#22]
+	orr	r8,r8,r6,lsl#24
+	ldrb	r5,[r12,#21]
+	ldrb	r6,[r12,#20]
+	orr	r9,r9,r4,lsl#8
+	orr	r9,r9,r5,lsl#16
+	str	r8,[r11],#8
+	orr	r9,r9,r6,lsl#24
+	str	r9,[r11,#-4]
+
+	teq	lr,#192
+	bne	.Lnot192
+	mov	r12,#12
+	str	r12,[r11,#240-24]
+	add	r6,r10,#256			@ rcon
+	mov	lr,#255
+	mov	r12,#8
+
+.L192_loop:
+	and	r5,lr,r9,lsr#24
+	and	r7,lr,r9,lsr#16
+	ldrb	r5,[r10,r5]
+	and	r8,lr,r9,lsr#8
+	ldrb	r7,[r10,r7]
+	and	r9,lr,r9
+	ldrb	r8,[r10,r8]
+	orr	r5,r5,r7,lsl#24
+	ldrb	r9,[r10,r9]
+	orr	r5,r5,r8,lsl#16
+	ldr	r4,[r6],#4			@ rcon[i++]
+	orr	r5,r5,r9,lsl#8
+	eor	r9,r5,r4
+	eor	r0,r0,r9			@ rk[6]=rk[0]^...
+	eor	r1,r1,r0			@ rk[7]=rk[1]^rk[6]
+	str	r0,[r11],#24
+	eor	r2,r2,r1			@ rk[8]=rk[2]^rk[7]
+	str	r1,[r11,#-20]
+	eor	r3,r3,r2			@ rk[9]=rk[3]^rk[8]
+	str	r2,[r11,#-16]
+	subs	r12,r12,#1
+	str	r3,[r11,#-12]
+	subeq	r2,r11,#216
+	beq	.Ldone
+
+	ldr	r7,[r11,#-32]
+	ldr	r8,[r11,#-28]
+	eor	r7,r7,r3			@ rk[10]=rk[4]^rk[9]
+	eor	r9,r8,r7			@ rk[11]=rk[5]^rk[10]
+	str	r7,[r11,#-8]
+	str	r9,[r11,#-4]
+	b	.L192_loop
+
+.Lnot192:
+	ldrb	r8,[r12,#27]
+	ldrb	r4,[r12,#26]
+	ldrb	r5,[r12,#25]
+	ldrb	r6,[r12,#24]
+	orr	r8,r8,r4,lsl#8
+	ldrb	r9,[r12,#31]
+	orr	r8,r8,r5,lsl#16
+	ldrb	r4,[r12,#30]
+	orr	r8,r8,r6,lsl#24
+	ldrb	r5,[r12,#29]
+	ldrb	r6,[r12,#28]
+	orr	r9,r9,r4,lsl#8
+	orr	r9,r9,r5,lsl#16
+	str	r8,[r11],#8
+	orr	r9,r9,r6,lsl#24
+	str	r9,[r11,#-4]
+
+	mov	r12,#14
+	str	r12,[r11,#240-32]
+	add	r6,r10,#256			@ rcon
+	mov	lr,#255
+	mov	r12,#7
+
+.L256_loop:
+	and	r5,lr,r9,lsr#24
+	and	r7,lr,r9,lsr#16
+	ldrb	r5,[r10,r5]
+	and	r8,lr,r9,lsr#8
+	ldrb	r7,[r10,r7]
+	and	r9,lr,r9
+	ldrb	r8,[r10,r8]
+	orr	r5,r5,r7,lsl#24
+	ldrb	r9,[r10,r9]
+	orr	r5,r5,r8,lsl#16
+	ldr	r4,[r6],#4			@ rcon[i++]
+	orr	r5,r5,r9,lsl#8
+	eor	r9,r5,r4
+	eor	r0,r0,r9			@ rk[8]=rk[0]^...
+	eor	r1,r1,r0			@ rk[9]=rk[1]^rk[8]
+	str	r0,[r11],#32
+	eor	r2,r2,r1			@ rk[10]=rk[2]^rk[9]
+	str	r1,[r11,#-28]
+	eor	r3,r3,r2			@ rk[11]=rk[3]^rk[10]
+	str	r2,[r11,#-24]
+	subs	r12,r12,#1
+	str	r3,[r11,#-20]
+	subeq	r2,r11,#256
+	beq	.Ldone
+
+	and	r5,lr,r3
+	and	r7,lr,r3,lsr#8
+	ldrb	r5,[r10,r5]
+	and	r8,lr,r3,lsr#16
+	ldrb	r7,[r10,r7]
+	and	r9,lr,r3,lsr#24
+	ldrb	r8,[r10,r8]
+	orr	r5,r5,r7,lsl#8
+	ldrb	r9,[r10,r9]
+	orr	r5,r5,r8,lsl#16
+	ldr	r4,[r11,#-48]
+	orr	r5,r5,r9,lsl#24
+
+	ldr	r7,[r11,#-44]
+	ldr	r8,[r11,#-40]
+	eor	r4,r4,r5			@ rk[12]=rk[4]^...
+	ldr	r9,[r11,#-36]
+	eor	r7,r7,r4			@ rk[13]=rk[5]^rk[12]
+	str	r4,[r11,#-16]
+	eor	r8,r8,r7			@ rk[14]=rk[6]^rk[13]
+	str	r7,[r11,#-12]
+	eor	r9,r9,r8			@ rk[15]=rk[7]^rk[14]
+	str	r8,[r11,#-8]
+	str	r9,[r11,#-4]
+	b	.L256_loop
+
+.Ldone:	mov	r0,#0
+	ldmia   sp!,{r4-r12,lr}
+.Labrt:	tst	lr,#1
+	moveq	pc,lr			@ be binary compatible with V4, yet
+	.word	0xe12fff1e			@ interoperable with Thumb ISA:-)
+.size	AES_set_encrypt_key,.-AES_set_encrypt_key
+
+.global AES_set_decrypt_key
+.type   AES_set_decrypt_key,%function
+.align	5
+AES_set_decrypt_key:
+	str	lr,[sp,#-4]!            @ push lr
+	bl	AES_set_encrypt_key
+	teq	r0,#0
+	ldrne	lr,[sp],#4              @ pop lr
+	bne	.Labrt
+
+	stmdb   sp!,{r4-r12}
+
+	ldr	r12,[r2,#240]	@ AES_set_encrypt_key preserves r2,
+	mov	r11,r2			@ which is AES_KEY *key
+	mov	r7,r2
+	add	r8,r2,r12,lsl#4
+
+.Linv:	ldr	r0,[r7]
+	ldr	r1,[r7,#4]
+	ldr	r2,[r7,#8]
+	ldr	r3,[r7,#12]
+	ldr	r4,[r8]
+	ldr	r5,[r8,#4]
+	ldr	r6,[r8,#8]
+	ldr	r9,[r8,#12]
+	str	r0,[r8],#-16
+	str	r1,[r8,#16+4]
+	str	r2,[r8,#16+8]
+	str	r3,[r8,#16+12]
+	str	r4,[r7],#16
+	str	r5,[r7,#-12]
+	str	r6,[r7,#-8]
+	str	r9,[r7,#-4]
+	teq	r7,r8
+	bne	.Linv
+	ldr	r0,[r11,#16]!		@ prefetch tp1
+	mov	r7,#0x80
+	mov	r8,#0x1b
+	orr	r7,r7,#0x8000
+	orr	r8,r8,#0x1b00
+	orr	r7,r7,r7,lsl#16
+	orr	r8,r8,r8,lsl#16
+	sub	r12,r12,#1
+	mvn	r9,r7
+	mov	r12,r12,lsl#2	@ (rounds-1)*4
+
+.Lmix:	and	r4,r0,r7
+	and	r1,r0,r9
+	sub	r4,r4,r4,lsr#7
+	and	r4,r4,r8
+	eor	r1,r4,r1,lsl#1	@ tp2
+
+	and	r4,r1,r7
+	and	r2,r1,r9
+	sub	r4,r4,r4,lsr#7
+	and	r4,r4,r8
+	eor	r2,r4,r2,lsl#1	@ tp4
+
+	and	r4,r2,r7
+	and	r3,r2,r9
+	sub	r4,r4,r4,lsr#7
+	and	r4,r4,r8
+	eor	r3,r4,r3,lsl#1	@ tp8
+
+	eor	r4,r1,r2
+	eor	r5,r0,r3		@ tp9
+	eor	r4,r4,r3		@ tpe
+	eor	r4,r4,r1,ror#24
+	eor	r4,r4,r5,ror#24	@ ^= ROTATE(tpb=tp9^tp2,8)
+	eor	r4,r4,r2,ror#16
+	eor	r4,r4,r5,ror#16	@ ^= ROTATE(tpd=tp9^tp4,16)
+	eor	r4,r4,r5,ror#8	@ ^= ROTATE(tp9,24)
+
+	ldr	r0,[r11,#4]		@ prefetch tp1
+	str	r4,[r11],#4
+	subs	r12,r12,#1
+	bne	.Lmix
+
+	mov	r0,#0
+	ldmia   sp!,{r4-r12,lr}
+	tst	lr,#1
+	moveq	pc,lr			@ be binary compatible with V4, yet
+	.word	0xe12fff1e			@ interoperable with Thumb ISA:-)
+.size	AES_set_decrypt_key,.-AES_set_decrypt_key
+
+.type	AES_Td,%object
+.align	5
+AES_Td:
+.word	0x51f4a750, 0x7e416553, 0x1a17a4c3, 0x3a275e96
+.word	0x3bab6bcb, 0x1f9d45f1, 0xacfa58ab, 0x4be30393
+.word	0x2030fa55, 0xad766df6, 0x88cc7691, 0xf5024c25
+.word	0x4fe5d7fc, 0xc52acbd7, 0x26354480, 0xb562a38f
+.word	0xdeb15a49, 0x25ba1b67, 0x45ea0e98, 0x5dfec0e1
+.word	0xc32f7502, 0x814cf012, 0x8d4697a3, 0x6bd3f9c6
+.word	0x038f5fe7, 0x15929c95, 0xbf6d7aeb, 0x955259da
+.word	0xd4be832d, 0x587421d3, 0x49e06929, 0x8ec9c844
+.word	0x75c2896a, 0xf48e7978, 0x99583e6b, 0x27b971dd
+.word	0xbee14fb6, 0xf088ad17, 0xc920ac66, 0x7dce3ab4
+.word	0x63df4a18, 0xe51a3182, 0x97513360, 0x62537f45
+.word	0xb16477e0, 0xbb6bae84, 0xfe81a01c, 0xf9082b94
+.word	0x70486858, 0x8f45fd19, 0x94de6c87, 0x527bf8b7
+.word	0xab73d323, 0x724b02e2, 0xe31f8f57, 0x6655ab2a
+.word	0xb2eb2807, 0x2fb5c203, 0x86c57b9a, 0xd33708a5
+.word	0x302887f2, 0x23bfa5b2, 0x02036aba, 0xed16825c
+.word	0x8acf1c2b, 0xa779b492, 0xf307f2f0, 0x4e69e2a1
+.word	0x65daf4cd, 0x0605bed5, 0xd134621f, 0xc4a6fe8a
+.word	0x342e539d, 0xa2f355a0, 0x058ae132, 0xa4f6eb75
+.word	0x0b83ec39, 0x4060efaa, 0x5e719f06, 0xbd6e1051
+.word	0x3e218af9, 0x96dd063d, 0xdd3e05ae, 0x4de6bd46
+.word	0x91548db5, 0x71c45d05, 0x0406d46f, 0x605015ff
+.word	0x1998fb24, 0xd6bde997, 0x894043cc, 0x67d99e77
+.word	0xb0e842bd, 0x07898b88, 0xe7195b38, 0x79c8eedb
+.word	0xa17c0a47, 0x7c420fe9, 0xf8841ec9, 0x00000000
+.word	0x09808683, 0x322bed48, 0x1e1170ac, 0x6c5a724e
+.word	0xfd0efffb, 0x0f853856, 0x3daed51e, 0x362d3927
+.word	0x0a0fd964, 0x685ca621, 0x9b5b54d1, 0x24362e3a
+.word	0x0c0a67b1, 0x9357e70f, 0xb4ee96d2, 0x1b9b919e
+.word	0x80c0c54f, 0x61dc20a2, 0x5a774b69, 0x1c121a16
+.word	0xe293ba0a, 0xc0a02ae5, 0x3c22e043, 0x121b171d
+.word	0x0e090d0b, 0xf28bc7ad, 0x2db6a8b9, 0x141ea9c8
+.word	0x57f11985, 0xaf75074c, 0xee99ddbb, 0xa37f60fd
+.word	0xf701269f, 0x5c72f5bc, 0x44663bc5, 0x5bfb7e34
+.word	0x8b432976, 0xcb23c6dc, 0xb6edfc68, 0xb8e4f163
+.word	0xd731dcca, 0x42638510, 0x13972240, 0x84c61120
+.word	0x854a247d, 0xd2bb3df8, 0xaef93211, 0xc729a16d
+.word	0x1d9e2f4b, 0xdcb230f3, 0x0d8652ec, 0x77c1e3d0
+.word	0x2bb3166c, 0xa970b999, 0x119448fa, 0x47e96422
+.word	0xa8fc8cc4, 0xa0f03f1a, 0x567d2cd8, 0x223390ef
+.word	0x87494ec7, 0xd938d1c1, 0x8ccaa2fe, 0x98d40b36
+.word	0xa6f581cf, 0xa57ade28, 0xdab78e26, 0x3fadbfa4
+.word	0x2c3a9de4, 0x5078920d, 0x6a5fcc9b, 0x547e4662
+.word	0xf68d13c2, 0x90d8b8e8, 0x2e39f75e, 0x82c3aff5
+.word	0x9f5d80be, 0x69d0937c, 0x6fd52da9, 0xcf2512b3
+.word	0xc8ac993b, 0x10187da7, 0xe89c636e, 0xdb3bbb7b
+.word	0xcd267809, 0x6e5918f4, 0xec9ab701, 0x834f9aa8
+.word	0xe6956e65, 0xaaffe67e, 0x21bccf08, 0xef15e8e6
+.word	0xbae79bd9, 0x4a6f36ce, 0xea9f09d4, 0x29b07cd6
+.word	0x31a4b2af, 0x2a3f2331, 0xc6a59430, 0x35a266c0
+.word	0x744ebc37, 0xfc82caa6, 0xe090d0b0, 0x33a7d815
+.word	0xf104984a, 0x41ecdaf7, 0x7fcd500e, 0x1791f62f
+.word	0x764dd68d, 0x43efb04d, 0xccaa4d54, 0xe49604df
+.word	0x9ed1b5e3, 0x4c6a881b, 0xc12c1fb8, 0x4665517f
+.word	0x9d5eea04, 0x018c355d, 0xfa877473, 0xfb0b412e
+.word	0xb3671d5a, 0x92dbd252, 0xe9105633, 0x6dd64713
+.word	0x9ad7618c, 0x37a10c7a, 0x59f8148e, 0xeb133c89
+.word	0xcea927ee, 0xb761c935, 0xe11ce5ed, 0x7a47b13c
+.word	0x9cd2df59, 0x55f2733f, 0x1814ce79, 0x73c737bf
+.word	0x53f7cdea, 0x5ffdaa5b, 0xdf3d6f14, 0x7844db86
+.word	0xcaaff381, 0xb968c43e, 0x3824342c, 0xc2a3405f
+.word	0x161dc372, 0xbce2250c, 0x283c498b, 0xff0d9541
+.word	0x39a80171, 0x080cb3de, 0xd8b4e49c, 0x6456c190
+.word	0x7bcb8461, 0xd532b670, 0x486c5c74, 0xd0b85742
+@ Td4[256]
+.byte	0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38
+.byte	0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb
+.byte	0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87
+.byte	0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb
+.byte	0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d
+.byte	0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e
+.byte	0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2
+.byte	0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25
+.byte	0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16
+.byte	0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92
+.byte	0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda
+.byte	0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84
+.byte	0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a
+.byte	0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06
+.byte	0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02
+.byte	0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b
+.byte	0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea
+.byte	0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73
+.byte	0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85
+.byte	0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e
+.byte	0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89
+.byte	0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b
+.byte	0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20
+.byte	0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4
+.byte	0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31
+.byte	0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f
+.byte	0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d
+.byte	0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef
+.byte	0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0
+.byte	0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61
+.byte	0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26
+.byte	0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d
+.size	AES_Td,.-AES_Td
+
+@ void AES_decrypt(const unsigned char *in, unsigned char *out,
+@ 		 const AES_KEY *key) {
+.global AES_decrypt
+.type   AES_decrypt,%function
+.align	5
+AES_decrypt:
+	sub	r3,pc,#8		@ AES_decrypt
+	stmdb   sp!,{r1,r4-r12,lr}
+	mov	r12,r0		@ inp
+	mov	r11,r2
+	sub	r10,r3,#AES_decrypt-AES_Td		@ Td
+
+	ldrb	r0,[r12,#3]	@ load input data in endian-neutral
+	ldrb	r4,[r12,#2]	@ manner...
+	ldrb	r5,[r12,#1]
+	ldrb	r6,[r12,#0]
+	orr	r0,r0,r4,lsl#8
+	ldrb	r1,[r12,#7]
+	orr	r0,r0,r5,lsl#16
+	ldrb	r4,[r12,#6]
+	orr	r0,r0,r6,lsl#24
+	ldrb	r5,[r12,#5]
+	ldrb	r6,[r12,#4]
+	orr	r1,r1,r4,lsl#8
+	ldrb	r2,[r12,#11]
+	orr	r1,r1,r5,lsl#16
+	ldrb	r4,[r12,#10]
+	orr	r1,r1,r6,lsl#24
+	ldrb	r5,[r12,#9]
+	ldrb	r6,[r12,#8]
+	orr	r2,r2,r4,lsl#8
+	ldrb	r3,[r12,#15]
+	orr	r2,r2,r5,lsl#16
+	ldrb	r4,[r12,#14]
+	orr	r2,r2,r6,lsl#24
+	ldrb	r5,[r12,#13]
+	ldrb	r6,[r12,#12]
+	orr	r3,r3,r4,lsl#8
+	orr	r3,r3,r5,lsl#16
+	orr	r3,r3,r6,lsl#24
+
+	bl	_armv4_AES_decrypt
+
+	ldr	r12,[sp],#4		@ pop out
+	mov	r4,r0,lsr#24		@ write output in endian-neutral
+	mov	r5,r0,lsr#16		@ manner...
+	mov	r6,r0,lsr#8
+	strb	r4,[r12,#0]
+	strb	r5,[r12,#1]
+	mov	r4,r1,lsr#24
+	strb	r6,[r12,#2]
+	mov	r5,r1,lsr#16
+	strb	r0,[r12,#3]
+	mov	r6,r1,lsr#8
+	strb	r4,[r12,#4]
+	strb	r5,[r12,#5]
+	mov	r4,r2,lsr#24
+	strb	r6,[r12,#6]
+	mov	r5,r2,lsr#16
+	strb	r1,[r12,#7]
+	mov	r6,r2,lsr#8
+	strb	r4,[r12,#8]
+	strb	r5,[r12,#9]
+	mov	r4,r3,lsr#24
+	strb	r6,[r12,#10]
+	mov	r5,r3,lsr#16
+	strb	r2,[r12,#11]
+	mov	r6,r3,lsr#8
+	strb	r4,[r12,#12]
+	strb	r5,[r12,#13]
+	strb	r6,[r12,#14]
+	strb	r3,[r12,#15]
+
+	ldmia   sp!,{r4-r12,lr}
+	tst	lr,#1
+	moveq	pc,lr			@ be binary compatible with V4, yet
+	.word	0xe12fff1e			@ interoperable with Thumb ISA:-)
+.size	AES_decrypt,.-AES_decrypt
+
+.type   _armv4_AES_decrypt,%function
+.align	2
+_armv4_AES_decrypt:
+	str	lr,[sp,#-4]!		@ push lr
+	ldmia	r11!,{r4-r7}
+	eor	r0,r0,r4
+	ldr	r12,[r11,#240-16]
+	eor	r1,r1,r5
+	eor	r2,r2,r6
+	eor	r3,r3,r7
+	sub	r12,r12,#1
+	mov	lr,#255
+
+	and	r7,lr,r0,lsr#16
+	and	r8,lr,r0,lsr#8
+	and	r9,lr,r0
+	mov	r0,r0,lsr#24
+.Ldec_loop:
+	ldr	r4,[r10,r7,lsl#2]	@ Td1[s0>>16]
+	and	r7,lr,r1		@ i0
+	ldr	r5,[r10,r8,lsl#2]	@ Td2[s0>>8]
+	and	r8,lr,r1,lsr#16
+	ldr	r6,[r10,r9,lsl#2]	@ Td3[s0>>0]
+	and	r9,lr,r1,lsr#8
+	ldr	r0,[r10,r0,lsl#2]	@ Td0[s0>>24]
+	mov	r1,r1,lsr#24
+
+	ldr	r7,[r10,r7,lsl#2]	@ Td3[s1>>0]
+	ldr	r8,[r10,r8,lsl#2]	@ Td1[s1>>16]
+	ldr	r9,[r10,r9,lsl#2]	@ Td2[s1>>8]
+	eor	r0,r0,r7,ror#24
+	ldr	r1,[r10,r1,lsl#2]	@ Td0[s1>>24]
+	and	r7,lr,r2,lsr#8	@ i0
+	eor	r5,r8,r5,ror#8
+	and	r8,lr,r2		@ i1
+	eor	r6,r9,r6,ror#8
+	and	r9,lr,r2,lsr#16
+	eor	r1,r1,r4,ror#8
+	ldr	r7,[r10,r7,lsl#2]	@ Td2[s2>>8]
+	mov	r2,r2,lsr#24
+
+	ldr	r8,[r10,r8,lsl#2]	@ Td3[s2>>0]
+	ldr	r9,[r10,r9,lsl#2]	@ Td1[s2>>16]
+	eor	r0,r0,r7,ror#16
+	ldr	r2,[r10,r2,lsl#2]	@ Td0[s2>>24]
+	and	r7,lr,r3,lsr#16	@ i0
+	eor	r1,r1,r8,ror#24
+	and	r8,lr,r3,lsr#8	@ i1
+	eor	r6,r9,r6,ror#8
+	and	r9,lr,r3		@ i2
+	eor	r2,r2,r5,ror#8
+	ldr	r7,[r10,r7,lsl#2]	@ Td1[s3>>16]
+	mov	r3,r3,lsr#24
+
+	ldr	r8,[r10,r8,lsl#2]	@ Td2[s3>>8]
+	ldr	r9,[r10,r9,lsl#2]	@ Td3[s3>>0]
+	eor	r0,r0,r7,ror#8
+	ldr	r3,[r10,r3,lsl#2]	@ Td0[s3>>24]
+	eor	r1,r1,r8,ror#16
+	eor	r2,r2,r9,ror#24
+	ldr	r7,[r11],#16
+	eor	r3,r3,r6,ror#8
+
+	ldr	r4,[r11,#-12]
+	ldr	r5,[r11,#-8]
+	eor	r0,r0,r7
+	ldr	r6,[r11,#-4]
+	and	r7,lr,r0,lsr#16
+	eor	r1,r1,r4
+	and	r8,lr,r0,lsr#8
+	eor	r2,r2,r5
+	and	r9,lr,r0
+	eor	r3,r3,r6
+	mov	r0,r0,lsr#24
+
+	subs	r12,r12,#1
+	bne	.Ldec_loop
+
+	add	r10,r10,#1024
+
+	ldr	r5,[r10,#0]		@ prefetch Td4
+	ldr	r6,[r10,#32]
+	ldr	r4,[r10,#64]
+	ldr	r5,[r10,#96]
+	ldr	r6,[r10,#128]
+	ldr	r4,[r10,#160]
+	ldr	r5,[r10,#192]
+	ldr	r6,[r10,#224]
+
+	ldrb	r0,[r10,r0]		@ Td4[s0>>24]
+	ldrb	r4,[r10,r7]		@ Td4[s0>>16]
+	and	r7,lr,r1		@ i0
+	ldrb	r5,[r10,r8]		@ Td4[s0>>8]
+	and	r8,lr,r1,lsr#16
+	ldrb	r6,[r10,r9]		@ Td4[s0>>0]
+	and	r9,lr,r1,lsr#8
+
+	ldrb	r7,[r10,r7]		@ Td4[s1>>0]
+	ldrb	r1,[r10,r1,lsr#24]	@ Td4[s1>>24]
+	ldrb	r8,[r10,r8]		@ Td4[s1>>16]
+	eor	r0,r7,r0,lsl#24
+	ldrb	r9,[r10,r9]		@ Td4[s1>>8]
+	eor	r1,r4,r1,lsl#8
+	and	r7,lr,r2,lsr#8	@ i0
+	eor	r5,r5,r8,lsl#8
+	and	r8,lr,r2		@ i1
+	eor	r6,r6,r9,lsl#8
+	ldrb	r7,[r10,r7]		@ Td4[s2>>8]
+	and	r9,lr,r2,lsr#16
+
+	ldrb	r8,[r10,r8]		@ Td4[s2>>0]
+	ldrb	r2,[r10,r2,lsr#24]	@ Td4[s2>>24]
+	eor	r0,r0,r7,lsl#8
+	ldrb	r9,[r10,r9]		@ Td4[s2>>16]
+	eor	r1,r8,r1,lsl#16
+	and	r7,lr,r3,lsr#16	@ i0
+	eor	r2,r5,r2,lsl#16
+	and	r8,lr,r3,lsr#8	@ i1
+	eor	r6,r6,r9,lsl#16
+	ldrb	r7,[r10,r7]		@ Td4[s3>>16]
+	and	r9,lr,r3		@ i2
+
+	ldrb	r8,[r10,r8]		@ Td4[s3>>8]
+	ldrb	r9,[r10,r9]		@ Td4[s3>>0]
+	ldrb	r3,[r10,r3,lsr#24]	@ Td4[s3>>24]
+	eor	r0,r0,r7,lsl#16
+	ldr	r7,[r11,#0]
+	eor	r1,r1,r8,lsl#8
+	ldr	r4,[r11,#4]
+	eor	r2,r9,r2,lsl#8
+	ldr	r5,[r11,#8]
+	eor	r3,r6,r3,lsl#24
+	ldr	r6,[r11,#12]
+
+	eor	r0,r0,r7
+	eor	r1,r1,r4
+	eor	r2,r2,r5
+	eor	r3,r3,r6
+
+	sub	r10,r10,#1024
+	ldr	pc,[sp],#4		@ pop and return
+.size	_armv4_AES_decrypt,.-_armv4_AES_decrypt
+.asciz	"AES for ARMv4, CRYPTOGAMS by <appro@openssl.org>"
+.align	2
diff --git a/openssl/crypto/aes/asm/aes-ia64.S b/openssl/crypto/aes/asm/aes-ia64.S
new file mode 100644
index 00000000..7f6c4c36
--- /dev/null
+++ b/openssl/crypto/aes/asm/aes-ia64.S
@@ -0,0 +1,1123 @@
+// ====================================================================
+// Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
+// project. Rights for redistribution and usage in source and binary
+// forms are granted according to the OpenSSL license.
+// ====================================================================
+//
+// What's wrong with compiler generated code? Compiler never uses
+// variable 'shr' which is pairable with 'extr'/'dep' instructions.
+// Then it uses 'zxt' which is an I-type, but can be replaced with
+// 'and' which in turn can be assigned to M-port [there're double as
+// much M-ports as there're I-ports on Itanium 2]. By sacrificing few
+// registers for small constants (255, 24 and 16) to be used with
+// 'shr' and 'and' instructions I can achieve better ILP, Intruction
+// Level Parallelism, and performance. This code outperforms GCC 3.3
+// generated code by over factor of 2 (two), GCC 3.4 - by 70% and
+// HP C - by 40%. Measured best-case scenario, i.e. aligned
+// big-endian input, ECB timing on Itanium 2 is (18 + 13*rounds)
+// ticks per block, or 9.25 CPU cycles per byte for 128 bit key.
+
+// Version 1.2 mitigates the hazard of cache-timing attacks by
+// a) compressing S-boxes from 8KB to 2KB+256B, b) scheduling
+// references to S-boxes for L2 cache latency, c) prefetching T[ed]4
+// prior last round. As result performance dropped to (26 + 15*rounds)
+// ticks per block or 11 cycles per byte processed with 128-bit key.
+// This is ~16% deterioration. For reference Itanium 2 L1 cache has
+// 64 bytes line size and L2 - 128 bytes...
+
+.ident	"aes-ia64.S, version 1.2"
+.ident	"IA-64 ISA artwork by Andy Polyakov <appro@fy.chalmers.se>"
+.explicit
+.text
+
+rk0=r8;     rk1=r9;
+
+pfssave=r2;
+lcsave=r10;
+prsave=r3;
+maskff=r11;
+twenty4=r14;
+sixteen=r15;
+
+te00=r16;   te11=r17;   te22=r18;   te33=r19;
+te01=r20;   te12=r21;   te23=r22;   te30=r23;
+te02=r24;   te13=r25;   te20=r26;   te31=r27;
+te03=r28;   te10=r29;   te21=r30;   te32=r31;
+
+// these are rotating...
+t0=r32;     s0=r33;
+t1=r34;     s1=r35;
+t2=r36;     s2=r37;
+t3=r38;     s3=r39;
+
+te0=r40;    te1=r41;    te2=r42;    te3=r43;
+
+#if defined(_HPUX_SOURCE) && !defined(_LP64)
+# define ADDP	addp4
+#else
+# define ADDP	add
+#endif
+
+// Offsets from Te0
+#define TE0	0
+#define TE2	2
+#if defined(_HPUX_SOURCE) || defined(B_ENDIAN)
+#define TE1	3
+#define TE3	1
+#else
+#define TE1	1
+#define TE3	3
+#endif
+
+// This implies that AES_KEY comprises 32-bit key schedule elements
+// even on LP64 platforms.
+#ifndef	KSZ
+# define KSZ	4
+# define LDKEY	ld4
+#endif
+
+.proc	_ia64_AES_encrypt#
+// Input:	rk0-rk1
+//		te0
+//		te3	as AES_KEY->rounds!!!
+//		s0-s3
+//		maskff,twenty4,sixteen
+// Output:	r16,r20,r24,r28 as s0-s3
+// Clobber:	r16-r31,rk0-rk1,r32-r43
+.align	32
+_ia64_AES_encrypt:
+	.prologue
+	.altrp	b6
+	.body
+{ .mmi;	alloc	r16=ar.pfs,12,0,0,8
+	LDKEY	t0=[rk0],2*KSZ
+	mov	pr.rot=1<<16	}
+{ .mmi;	LDKEY	t1=[rk1],2*KSZ
+	add	te1=TE1,te0
+	add	te3=-3,te3	};;
+{ .mib;	LDKEY	t2=[rk0],2*KSZ
+	mov	ar.ec=2		}
+{ .mib;	LDKEY	t3=[rk1],2*KSZ
+	add	te2=TE2,te0
+	brp.loop.imp	.Le_top,.Le_end-16	};;
+
+{ .mmi;	xor	s0=s0,t0
+	xor	s1=s1,t1
+	mov	ar.lc=te3	}
+{ .mmi;	xor	s2=s2,t2
+	xor	s3=s3,t3
+	add	te3=TE3,te0	};;
+
+.align	32
+.Le_top:
+{ .mmi;	(p0)	LDKEY	t0=[rk0],2*KSZ		// 0/0:rk[0]
+	(p0)	and	te33=s3,maskff		// 0/0:s3&0xff
+	(p0)	extr.u	te22=s2,8,8	}	// 0/0:s2>>8&0xff
+{ .mmi; (p0)	LDKEY	t1=[rk1],2*KSZ		// 0/1:rk[1]
+	(p0)	and	te30=s0,maskff		// 0/1:s0&0xff
+	(p0)	shr.u	te00=s0,twenty4	};;	// 0/0:s0>>24
+{ .mmi;	(p0)	LDKEY	t2=[rk0],2*KSZ		// 1/2:rk[2]
+	(p0)	shladd	te33=te33,3,te3		// 1/0:te0+s0>>24
+	(p0)	extr.u	te23=s3,8,8	}	// 1/1:s3>>8&0xff
+{ .mmi;	(p0)	LDKEY	t3=[rk1],2*KSZ		// 1/3:rk[3]
+	(p0)	shladd	te30=te30,3,te3		// 1/1:te3+s0
+	(p0)	shr.u	te01=s1,twenty4	};;	// 1/1:s1>>24
+{ .mmi;	(p0)	ld4	te33=[te33]		// 2/0:te3[s3&0xff]
+	(p0)	shladd	te22=te22,3,te2		// 2/0:te2+s2>>8&0xff
+	(p0)	extr.u	te20=s0,8,8	}	// 2/2:s0>>8&0xff
+{ .mmi;	(p0)	ld4	te30=[te30]		// 2/1:te3[s0]
+	(p0)	shladd	te23=te23,3,te2		// 2/1:te2+s3>>8
+	(p0)	shr.u	te02=s2,twenty4	};;	// 2/2:s2>>24
+{ .mmi;	(p0)	ld4	te22=[te22]		// 3/0:te2[s2>>8]
+	(p0)	shladd	te20=te20,3,te2		// 3/2:te2+s0>>8
+	(p0)	extr.u	te21=s1,8,8	}	// 3/3:s1>>8&0xff
+{ .mmi;	(p0)	ld4	te23=[te23]		// 3/1:te2[s3>>8]
+	(p0)	shladd	te00=te00,3,te0		// 3/0:te0+s0>>24
+	(p0)	shr.u	te03=s3,twenty4	};;	// 3/3:s3>>24
+{ .mmi;	(p0)	ld4	te20=[te20]		// 4/2:te2[s0>>8]
+	(p0)	shladd	te21=te21,3,te2		// 4/3:te3+s2
+	(p0)	extr.u	te11=s1,16,8	}	// 4/0:s1>>16&0xff
+{ .mmi;	(p0)	ld4	te00=[te00]		// 4/0:te0[s0>>24]
+	(p0)	shladd	te01=te01,3,te0		// 4/1:te0+s1>>24
+	(p0)	shr.u	te13=s3,sixteen	};;	// 4/2:s3>>16
+{ .mmi;	(p0)	ld4	te21=[te21]		// 5/3:te2[s1>>8]
+	(p0)	shladd	te11=te11,3,te1		// 5/0:te1+s1>>16
+	(p0)	extr.u	te12=s2,16,8	}	// 5/1:s2>>16&0xff
+{ .mmi;	(p0)	ld4	te01=[te01]		// 5/1:te0[s1>>24]
+	(p0)	shladd	te02=te02,3,te0		// 5/2:te0+s2>>24
+	(p0)	and	te31=s1,maskff	};;	// 5/2:s1&0xff
+{ .mmi;	(p0)	ld4	te11=[te11]		// 6/0:te1[s1>>16]
+	(p0)	shladd	te12=te12,3,te1		// 6/1:te1+s2>>16
+	(p0)	extr.u	te10=s0,16,8	}	// 6/3:s0>>16&0xff
+{ .mmi;	(p0)	ld4	te02=[te02]		// 6/2:te0[s2>>24]
+	(p0)	shladd	te03=te03,3,te0		// 6/3:te1+s0>>16
+	(p0)	and	te32=s2,maskff	};;	// 6/3:s2&0xff
+
+{ .mmi;	(p0)	ld4	te12=[te12]		// 7/1:te1[s2>>16]
+	(p0)	shladd	te31=te31,3,te3		// 7/2:te3+s1&0xff
+	(p0)	and	te13=te13,maskff}	// 7/2:s3>>16&0xff
+{ .mmi;	(p0)	ld4	te03=[te03]		// 7/3:te0[s3>>24]
+	(p0)	shladd	te32=te32,3,te3		// 7/3:te3+s2
+	(p0)	xor	t0=t0,te33	};;	// 7/0:
+{ .mmi;	(p0)	ld4	te31=[te31]		// 8/2:te3[s1]
+	(p0)	shladd	te13=te13,3,te1		// 8/2:te1+s3>>16
+	(p0)	xor	t0=t0,te22	}	// 8/0:
+{ .mmi;	(p0)	ld4	te32=[te32]		// 8/3:te3[s2]
+	(p0)	shladd	te10=te10,3,te1		// 8/3:te1+s0>>16
+	(p0)	xor	t1=t1,te30	};;	// 8/1:
+{ .mmi;	(p0)	ld4	te13=[te13]		// 9/2:te1[s3>>16]
+	(p0)	ld4	te10=[te10]		// 9/3:te1[s0>>16]
+	(p0)	xor	t0=t0,te00	};;	// 9/0:		!L2 scheduling
+{ .mmi;	(p0)	xor	t1=t1,te23		// 10[9]/1:	
+	(p0)	xor	t2=t2,te20		// 10[9]/2:
+	(p0)	xor	t3=t3,te21	};;	// 10[9]/3:
+{ .mmi;	(p0)	xor	t0=t0,te11		// 11[10]/0:done!
+	(p0)	xor	t1=t1,te01		// 11[10]/1:
+	(p0)	xor	t2=t2,te02	};;	// 11[10]/2:	!L2 scheduling
+{ .mmi;	(p0)	xor	t3=t3,te03		// 12[10]/3:
+	(p16)	cmp.eq	p0,p17=r0,r0 	};;	// 12[10]/clear (p17)
+{ .mmi;	(p0)	xor	t1=t1,te12		// 13[11]/1:done!
+	(p0)	xor	t2=t2,te31		// 13[11]/2:
+	(p0)	xor	t3=t3,te32	}	// 13[11]/3:
+{ .mmi;	(p17)	add	te0=2048,te0		// 13[11]/
+	(p17)	add	te1=2048+64-TE1,te1};;	// 13[11]/
+{ .mib;	(p0)	xor	t2=t2,te13		// 14[12]/2:done!
+	(p17)	add	te2=2048+128-TE2,te2}	// 14[12]/
+{ .mib;	(p0)	xor	t3=t3,te10		// 14[12]/3:done!
+	(p17)	add	te3=2048+192-TE3,te3	// 14[12]/
+	br.ctop.sptk	.Le_top		};;
+.Le_end:
+
+
+{ .mmi;	ld8	te12=[te0]		// prefetch Te4
+	ld8	te31=[te1]	}
+{ .mmi;	ld8	te10=[te2]
+	ld8	te32=[te3]	}
+
+{ .mmi;	LDKEY	t0=[rk0],2*KSZ		// 0/0:rk[0]
+	and	te33=s3,maskff		// 0/0:s3&0xff
+	extr.u	te22=s2,8,8	}	// 0/0:s2>>8&0xff
+{ .mmi; LDKEY	t1=[rk1],2*KSZ		// 0/1:rk[1]
+	and	te30=s0,maskff		// 0/1:s0&0xff
+	shr.u	te00=s0,twenty4	};;	// 0/0:s0>>24
+{ .mmi;	LDKEY	t2=[rk0],2*KSZ		// 1/2:rk[2]
+	add	te33=te33,te0		// 1/0:te0+s0>>24
+	extr.u	te23=s3,8,8	}	// 1/1:s3>>8&0xff
+{ .mmi;	LDKEY	t3=[rk1],2*KSZ		// 1/3:rk[3]
+	add	te30=te30,te0		// 1/1:te0+s0
+	shr.u	te01=s1,twenty4	};;	// 1/1:s1>>24
+{ .mmi;	ld1	te33=[te33]		// 2/0:te0[s3&0xff]
+	add	te22=te22,te0		// 2/0:te0+s2>>8&0xff
+	extr.u	te20=s0,8,8	}	// 2/2:s0>>8&0xff
+{ .mmi;	ld1	te30=[te30]		// 2/1:te0[s0]
+	add	te23=te23,te0		// 2/1:te0+s3>>8
+	shr.u	te02=s2,twenty4	};;	// 2/2:s2>>24
+{ .mmi;	ld1	te22=[te22]		// 3/0:te0[s2>>8]
+	add	te20=te20,te0		// 3/2:te0+s0>>8
+	extr.u	te21=s1,8,8	}	// 3/3:s1>>8&0xff
+{ .mmi;	ld1	te23=[te23]		// 3/1:te0[s3>>8]
+	add	te00=te00,te0		// 3/0:te0+s0>>24
+	shr.u	te03=s3,twenty4	};;	// 3/3:s3>>24
+{ .mmi;	ld1	te20=[te20]		// 4/2:te0[s0>>8]
+	add	te21=te21,te0		// 4/3:te0+s2
+	extr.u	te11=s1,16,8	}	// 4/0:s1>>16&0xff
+{ .mmi;	ld1	te00=[te00]		// 4/0:te0[s0>>24]
+	add	te01=te01,te0		// 4/1:te0+s1>>24
+	shr.u	te13=s3,sixteen	};;	// 4/2:s3>>16
+{ .mmi;	ld1	te21=[te21]		// 5/3:te0[s1>>8]
+	add	te11=te11,te0		// 5/0:te0+s1>>16
+	extr.u	te12=s2,16,8	}	// 5/1:s2>>16&0xff
+{ .mmi;	ld1	te01=[te01]		// 5/1:te0[s1>>24]
+	add	te02=te02,te0		// 5/2:te0+s2>>24
+	and	te31=s1,maskff	};;	// 5/2:s1&0xff
+{ .mmi;	ld1	te11=[te11]		// 6/0:te0[s1>>16]
+	add	te12=te12,te0		// 6/1:te0+s2>>16
+	extr.u	te10=s0,16,8	}	// 6/3:s0>>16&0xff
+{ .mmi;	ld1	te02=[te02]		// 6/2:te0[s2>>24]
+	add	te03=te03,te0		// 6/3:te0+s0>>16
+	and	te32=s2,maskff	};;	// 6/3:s2&0xff
+
+{ .mmi;	ld1	te12=[te12]		// 7/1:te0[s2>>16]
+	add	te31=te31,te0		// 7/2:te0+s1&0xff
+	dep	te33=te22,te33,8,8}	// 7/0:
+{ .mmi;	ld1	te03=[te03]		// 7/3:te0[s3>>24]
+	add	te32=te32,te0		// 7/3:te0+s2
+	and	te13=te13,maskff};;	// 7/2:s3>>16&0xff
+{ .mmi;	ld1	te31=[te31]		// 8/2:te0[s1]
+	add	te13=te13,te0		// 8/2:te0+s3>>16
+	dep	te30=te23,te30,8,8}	// 8/1:
+{ .mmi;	ld1	te32=[te32]		// 8/3:te0[s2]
+	add	te10=te10,te0		// 8/3:te0+s0>>16
+	shl	te00=te00,twenty4};;	// 8/0:
+{ .mii;	ld1	te13=[te13]		// 9/2:te0[s3>>16]
+	dep	te33=te11,te33,16,8	// 9/0:
+	shl	te01=te01,twenty4};;	// 9/1:
+{ .mii;	ld1	te10=[te10]		// 10/3:te0[s0>>16]
+	dep	te31=te20,te31,8,8	// 10/2:
+	shl	te02=te02,twenty4};;	// 10/2:
+{ .mii;	xor	t0=t0,te33		// 11/0:
+	dep	te32=te21,te32,8,8	// 11/3:
+	shl	te12=te12,sixteen};;	// 11/1:
+{ .mii;	xor	r16=t0,te00		// 12/0:done!
+	dep	te31=te13,te31,16,8	// 12/2:
+	shl	te03=te03,twenty4};;	// 12/3:
+{ .mmi;	xor	t1=t1,te01		// 13/1:
+	xor	t2=t2,te02		// 13/2:
+	dep	te32=te10,te32,16,8};;	// 13/3:
+{ .mmi;	xor	t1=t1,te30		// 14/1:
+	xor	r24=t2,te31		// 14/2:done!
+	xor	t3=t3,te32	};;	// 14/3:
+{ .mib;	xor	r20=t1,te12		// 15/1:done!
+	xor	r28=t3,te03		// 15/3:done!
+	br.ret.sptk	b6	};;
+.endp	_ia64_AES_encrypt#
+
+// void AES_encrypt (const void *in,void *out,const AES_KEY *key);
+.global	AES_encrypt#
+.proc	AES_encrypt#
+.align	32
+AES_encrypt:
+	.prologue
+	.save	ar.pfs,pfssave
+{ .mmi;	alloc	pfssave=ar.pfs,3,1,12,0
+	and	out0=3,in0
+	mov	r3=ip			}
+{ .mmi;	ADDP	in0=0,in0
+	mov	loc0=psr.um
+	ADDP	out11=KSZ*60,in2	};;	// &AES_KEY->rounds
+
+{ .mmi;	ld4	out11=[out11]			// AES_KEY->rounds
+	add	out8=(AES_Te#-AES_encrypt#),r3	// Te0
+	.save	pr,prsave
+	mov	prsave=pr		}
+{ .mmi;	rum	1<<3				// clear um.ac
+	.save	ar.lc,lcsave
+	mov	lcsave=ar.lc		};;
+
+	.body
+#if defined(_HPUX_SOURCE)	// HPUX is big-endian, cut 15+15 cycles...
+{ .mib; cmp.ne	p6,p0=out0,r0
+	add	out0=4,in0
+(p6)	br.dpnt.many	.Le_i_unaligned	};;
+
+{ .mmi;	ld4	out1=[in0],8		// s0
+	and	out9=3,in1
+	mov	twenty4=24		}
+{ .mmi;	ld4	out3=[out0],8		// s1
+	ADDP	rk0=0,in2
+	mov	sixteen=16		};;
+{ .mmi;	ld4	out5=[in0]		// s2
+	cmp.ne	p6,p0=out9,r0
+	mov	maskff=0xff		}
+{ .mmb;	ld4	out7=[out0]		// s3
+	ADDP	rk1=KSZ,in2
+	br.call.sptk.many	b6=_ia64_AES_encrypt	};;
+
+{ .mib;	ADDP	in0=4,in1
+	ADDP	in1=0,in1
+(p6)	br.spnt	.Le_o_unaligned		};;
+
+{ .mii;	mov	psr.um=loc0
+	mov	ar.pfs=pfssave
+	mov	ar.lc=lcsave		};;
+{ .mmi;	st4	[in1]=r16,8		// s0
+	st4	[in0]=r20,8		// s1
+	mov	pr=prsave,0x1ffff	};;
+{ .mmb;	st4	[in1]=r24		// s2
+	st4	[in0]=r28		// s3
+	br.ret.sptk.many	b0	};;
+#endif
+
+.align	32
+.Le_i_unaligned:
+{ .mmi;	add	out0=1,in0
+	add	out2=2,in0
+	add	out4=3,in0	};;
+{ .mmi;	ld1	r16=[in0],4
+	ld1	r17=[out0],4	}//;;
+{ .mmi;	ld1	r18=[out2],4
+	ld1	out1=[out4],4	};;	// s0
+{ .mmi;	ld1	r20=[in0],4
+	ld1	r21=[out0],4	}//;;
+{ .mmi;	ld1	r22=[out2],4
+	ld1	out3=[out4],4	};;	// s1
+{ .mmi;	ld1	r24=[in0],4
+	ld1	r25=[out0],4	}//;;
+{ .mmi;	ld1	r26=[out2],4
+	ld1	out5=[out4],4	};;	// s2
+{ .mmi;	ld1	r28=[in0]
+	ld1	r29=[out0]	}//;;
+{ .mmi;	ld1	r30=[out2]
+	ld1	out7=[out4]	};;	// s3
+
+{ .mii;
+	dep	out1=r16,out1,24,8	//;;
+	dep	out3=r20,out3,24,8	}//;;
+{ .mii;	ADDP	rk0=0,in2
+	dep	out5=r24,out5,24,8	//;;
+	dep	out7=r28,out7,24,8	};;
+{ .mii;	ADDP	rk1=KSZ,in2
+	dep	out1=r17,out1,16,8	//;;
+	dep	out3=r21,out3,16,8	}//;;
+{ .mii;	mov	twenty4=24
+	dep	out5=r25,out5,16,8	//;;
+	dep	out7=r29,out7,16,8	};;
+{ .mii;	mov	sixteen=16
+	dep	out1=r18,out1,8,8	//;;
+	dep	out3=r22,out3,8,8	}//;;
+{ .mii;	mov	maskff=0xff
+	dep	out5=r26,out5,8,8	//;;
+	dep	out7=r30,out7,8,8	};;
+
+{ .mib;	br.call.sptk.many	b6=_ia64_AES_encrypt	};;
+
+.Le_o_unaligned:
+{ .mii;	ADDP	out0=0,in1
+	extr.u	r17=r16,8,8			// s0
+	shr.u	r19=r16,twenty4		}//;;
+{ .mii;	ADDP	out1=1,in1
+	extr.u	r18=r16,16,8
+	shr.u	r23=r20,twenty4		}//;;	// s1
+{ .mii;	ADDP	out2=2,in1
+	extr.u	r21=r20,8,8
+	shr.u	r22=r20,sixteen		}//;;
+{ .mii;	ADDP	out3=3,in1
+	extr.u	r25=r24,8,8			// s2
+	shr.u	r27=r24,twenty4		};;
+{ .mii;	st1	[out3]=r16,4
+	extr.u	r26=r24,16,8
+	shr.u	r31=r28,twenty4		}//;;	// s3
+{ .mii;	st1	[out2]=r17,4
+	extr.u	r29=r28,8,8
+	shr.u	r30=r28,sixteen		}//;;
+
+{ .mmi;	st1	[out1]=r18,4
+	st1	[out0]=r19,4		};;
+{ .mmi;	st1	[out3]=r20,4
+	st1	[out2]=r21,4		}//;;
+{ .mmi;	st1	[out1]=r22,4
+	st1	[out0]=r23,4		};;
+{ .mmi;	st1	[out3]=r24,4
+	st1	[out2]=r25,4
+	mov	pr=prsave,0x1ffff	}//;;
+{ .mmi;	st1	[out1]=r26,4
+	st1	[out0]=r27,4
+	mov	ar.pfs=pfssave		};;
+{ .mmi;	st1	[out3]=r28
+	st1	[out2]=r29
+	mov	ar.lc=lcsave		}//;;
+{ .mmi;	st1	[out1]=r30
+	st1	[out0]=r31		}
+{ .mfb;	mov	psr.um=loc0			// restore user mask
+	br.ret.sptk.many	b0	};;
+.endp	AES_encrypt#
+
+// *AES_decrypt are autogenerated by the following script:
+#if 0
+#!/usr/bin/env perl
+print "// *AES_decrypt are autogenerated by the following script:\n#if 0\n";
+open(PROG,'<'.$0); while(<PROG>) { print; } close(PROG);
+print "#endif\n";
+while(<>) {
+	$process=1	if (/\.proc\s+_ia64_AES_encrypt/);
+	next		if (!$process);
+
+	#s/te00=s0/td00=s0/;	s/te00/td00/g;
+	s/te11=s1/td13=s3/;	s/te11/td13/g;
+	#s/te22=s2/td22=s2/;	s/te22/td22/g;
+	s/te33=s3/td31=s1/;	s/te33/td31/g;
+
+	#s/te01=s1/td01=s1/;	s/te01/td01/g;
+	s/te12=s2/td10=s0/;	s/te12/td10/g;
+	#s/te23=s3/td23=s3/;	s/te23/td23/g;
+	s/te30=s0/td32=s2/;	s/te30/td32/g;
+
+	#s/te02=s2/td02=s2/;	s/te02/td02/g;
+	s/te13=s3/td11=s1/;	s/te13/td11/g;
+	#s/te20=s0/td20=s0/;	s/te20/td20/g;
+	s/te31=s1/td33=s3/;	s/te31/td33/g;
+
+	#s/te03=s3/td03=s3/;	s/te03/td03/g;
+	s/te10=s0/td12=s2/;	s/te10/td12/g;
+	#s/te21=s1/td21=s1/;	s/te21/td21/g;
+	s/te32=s2/td30=s0/;	s/te32/td30/g;
+
+	s/td/te/g;
+
+	s/AES_encrypt/AES_decrypt/g;
+	s/\.Le_/.Ld_/g;
+	s/AES_Te#/AES_Td#/g;
+
+	print;
+
+	exit		if (/\.endp\s+AES_decrypt/);
+}
+#endif
+.proc	_ia64_AES_decrypt#
+// Input:	rk0-rk1
+//		te0
+//		te3	as AES_KEY->rounds!!!
+//		s0-s3
+//		maskff,twenty4,sixteen
+// Output:	r16,r20,r24,r28 as s0-s3
+// Clobber:	r16-r31,rk0-rk1,r32-r43
+.align	32
+_ia64_AES_decrypt:
+	.prologue
+	.altrp	b6
+	.body
+{ .mmi;	alloc	r16=ar.pfs,12,0,0,8
+	LDKEY	t0=[rk0],2*KSZ
+	mov	pr.rot=1<<16	}
+{ .mmi;	LDKEY	t1=[rk1],2*KSZ
+	add	te1=TE1,te0
+	add	te3=-3,te3	};;
+{ .mib;	LDKEY	t2=[rk0],2*KSZ
+	mov	ar.ec=2		}
+{ .mib;	LDKEY	t3=[rk1],2*KSZ
+	add	te2=TE2,te0
+	brp.loop.imp	.Ld_top,.Ld_end-16	};;
+
+{ .mmi;	xor	s0=s0,t0
+	xor	s1=s1,t1
+	mov	ar.lc=te3	}
+{ .mmi;	xor	s2=s2,t2
+	xor	s3=s3,t3
+	add	te3=TE3,te0	};;
+
+.align	32
+.Ld_top:
+{ .mmi;	(p0)	LDKEY	t0=[rk0],2*KSZ		// 0/0:rk[0]
+	(p0)	and	te31=s1,maskff		// 0/0:s3&0xff
+	(p0)	extr.u	te22=s2,8,8	}	// 0/0:s2>>8&0xff
+{ .mmi; (p0)	LDKEY	t1=[rk1],2*KSZ		// 0/1:rk[1]
+	(p0)	and	te32=s2,maskff		// 0/1:s0&0xff
+	(p0)	shr.u	te00=s0,twenty4	};;	// 0/0:s0>>24
+{ .mmi;	(p0)	LDKEY	t2=[rk0],2*KSZ		// 1/2:rk[2]
+	(p0)	shladd	te31=te31,3,te3		// 1/0:te0+s0>>24
+	(p0)	extr.u	te23=s3,8,8	}	// 1/1:s3>>8&0xff
+{ .mmi;	(p0)	LDKEY	t3=[rk1],2*KSZ		// 1/3:rk[3]
+	(p0)	shladd	te32=te32,3,te3		// 1/1:te3+s0
+	(p0)	shr.u	te01=s1,twenty4	};;	// 1/1:s1>>24
+{ .mmi;	(p0)	ld4	te31=[te31]		// 2/0:te3[s3&0xff]
+	(p0)	shladd	te22=te22,3,te2		// 2/0:te2+s2>>8&0xff
+	(p0)	extr.u	te20=s0,8,8	}	// 2/2:s0>>8&0xff
+{ .mmi;	(p0)	ld4	te32=[te32]		// 2/1:te3[s0]
+	(p0)	shladd	te23=te23,3,te2		// 2/1:te2+s3>>8
+	(p0)	shr.u	te02=s2,twenty4	};;	// 2/2:s2>>24
+{ .mmi;	(p0)	ld4	te22=[te22]		// 3/0:te2[s2>>8]
+	(p0)	shladd	te20=te20,3,te2		// 3/2:te2+s0>>8
+	(p0)	extr.u	te21=s1,8,8	}	// 3/3:s1>>8&0xff
+{ .mmi;	(p0)	ld4	te23=[te23]		// 3/1:te2[s3>>8]
+	(p0)	shladd	te00=te00,3,te0		// 3/0:te0+s0>>24
+	(p0)	shr.u	te03=s3,twenty4	};;	// 3/3:s3>>24
+{ .mmi;	(p0)	ld4	te20=[te20]		// 4/2:te2[s0>>8]
+	(p0)	shladd	te21=te21,3,te2		// 4/3:te3+s2
+	(p0)	extr.u	te13=s3,16,8	}	// 4/0:s1>>16&0xff
+{ .mmi;	(p0)	ld4	te00=[te00]		// 4/0:te0[s0>>24]
+	(p0)	shladd	te01=te01,3,te0		// 4/1:te0+s1>>24
+	(p0)	shr.u	te11=s1,sixteen	};;	// 4/2:s3>>16
+{ .mmi;	(p0)	ld4	te21=[te21]		// 5/3:te2[s1>>8]
+	(p0)	shladd	te13=te13,3,te1		// 5/0:te1+s1>>16
+	(p0)	extr.u	te10=s0,16,8	}	// 5/1:s2>>16&0xff
+{ .mmi;	(p0)	ld4	te01=[te01]		// 5/1:te0[s1>>24]
+	(p0)	shladd	te02=te02,3,te0		// 5/2:te0+s2>>24
+	(p0)	and	te33=s3,maskff	};;	// 5/2:s1&0xff
+{ .mmi;	(p0)	ld4	te13=[te13]		// 6/0:te1[s1>>16]
+	(p0)	shladd	te10=te10,3,te1		// 6/1:te1+s2>>16
+	(p0)	extr.u	te12=s2,16,8	}	// 6/3:s0>>16&0xff
+{ .mmi;	(p0)	ld4	te02=[te02]		// 6/2:te0[s2>>24]
+	(p0)	shladd	te03=te03,3,te0		// 6/3:te1+s0>>16
+	(p0)	and	te30=s0,maskff	};;	// 6/3:s2&0xff
+
+{ .mmi;	(p0)	ld4	te10=[te10]		// 7/1:te1[s2>>16]
+	(p0)	shladd	te33=te33,3,te3		// 7/2:te3+s1&0xff
+	(p0)	and	te11=te11,maskff}	// 7/2:s3>>16&0xff
+{ .mmi;	(p0)	ld4	te03=[te03]		// 7/3:te0[s3>>24]
+	(p0)	shladd	te30=te30,3,te3		// 7/3:te3+s2
+	(p0)	xor	t0=t0,te31	};;	// 7/0:
+{ .mmi;	(p0)	ld4	te33=[te33]		// 8/2:te3[s1]
+	(p0)	shladd	te11=te11,3,te1		// 8/2:te1+s3>>16
+	(p0)	xor	t0=t0,te22	}	// 8/0:
+{ .mmi;	(p0)	ld4	te30=[te30]		// 8/3:te3[s2]
+	(p0)	shladd	te12=te12,3,te1		// 8/3:te1+s0>>16
+	(p0)	xor	t1=t1,te32	};;	// 8/1:
+{ .mmi;	(p0)	ld4	te11=[te11]		// 9/2:te1[s3>>16]
+	(p0)	ld4	te12=[te12]		// 9/3:te1[s0>>16]
+	(p0)	xor	t0=t0,te00	};;	// 9/0:		!L2 scheduling
+{ .mmi;	(p0)	xor	t1=t1,te23		// 10[9]/1:	
+	(p0)	xor	t2=t2,te20		// 10[9]/2:
+	(p0)	xor	t3=t3,te21	};;	// 10[9]/3:
+{ .mmi;	(p0)	xor	t0=t0,te13		// 11[10]/0:done!
+	(p0)	xor	t1=t1,te01		// 11[10]/1:
+	(p0)	xor	t2=t2,te02	};;	// 11[10]/2:	!L2 scheduling
+{ .mmi;	(p0)	xor	t3=t3,te03		// 12[10]/3:
+	(p16)	cmp.eq	p0,p17=r0,r0 	};;	// 12[10]/clear (p17)
+{ .mmi;	(p0)	xor	t1=t1,te10		// 13[11]/1:done!
+	(p0)	xor	t2=t2,te33		// 13[11]/2:
+	(p0)	xor	t3=t3,te30	}	// 13[11]/3:
+{ .mmi;	(p17)	add	te0=2048,te0		// 13[11]/
+	(p17)	add	te1=2048+64-TE1,te1};;	// 13[11]/
+{ .mib;	(p0)	xor	t2=t2,te11		// 14[12]/2:done!
+	(p17)	add	te2=2048+128-TE2,te2}	// 14[12]/
+{ .mib;	(p0)	xor	t3=t3,te12		// 14[12]/3:done!
+	(p17)	add	te3=2048+192-TE3,te3	// 14[12]/
+	br.ctop.sptk	.Ld_top		};;
+.Ld_end:
+
+
+{ .mmi;	ld8	te10=[te0]		// prefetch Td4
+	ld8	te33=[te1]	}
+{ .mmi;	ld8	te12=[te2]
+	ld8	te30=[te3]	}
+
+{ .mmi;	LDKEY	t0=[rk0],2*KSZ		// 0/0:rk[0]
+	and	te31=s1,maskff		// 0/0:s3&0xff
+	extr.u	te22=s2,8,8	}	// 0/0:s2>>8&0xff
+{ .mmi; LDKEY	t1=[rk1],2*KSZ		// 0/1:rk[1]
+	and	te32=s2,maskff		// 0/1:s0&0xff
+	shr.u	te00=s0,twenty4	};;	// 0/0:s0>>24
+{ .mmi;	LDKEY	t2=[rk0],2*KSZ		// 1/2:rk[2]
+	add	te31=te31,te0		// 1/0:te0+s0>>24
+	extr.u	te23=s3,8,8	}	// 1/1:s3>>8&0xff
+{ .mmi;	LDKEY	t3=[rk1],2*KSZ		// 1/3:rk[3]
+	add	te32=te32,te0		// 1/1:te0+s0
+	shr.u	te01=s1,twenty4	};;	// 1/1:s1>>24
+{ .mmi;	ld1	te31=[te31]		// 2/0:te0[s3&0xff]
+	add	te22=te22,te0		// 2/0:te0+s2>>8&0xff
+	extr.u	te20=s0,8,8	}	// 2/2:s0>>8&0xff
+{ .mmi;	ld1	te32=[te32]		// 2/1:te0[s0]
+	add	te23=te23,te0		// 2/1:te0+s3>>8
+	shr.u	te02=s2,twenty4	};;	// 2/2:s2>>24
+{ .mmi;	ld1	te22=[te22]		// 3/0:te0[s2>>8]
+	add	te20=te20,te0		// 3/2:te0+s0>>8
+	extr.u	te21=s1,8,8	}	// 3/3:s1>>8&0xff
+{ .mmi;	ld1	te23=[te23]		// 3/1:te0[s3>>8]
+	add	te00=te00,te0		// 3/0:te0+s0>>24
+	shr.u	te03=s3,twenty4	};;	// 3/3:s3>>24
+{ .mmi;	ld1	te20=[te20]		// 4/2:te0[s0>>8]
+	add	te21=te21,te0		// 4/3:te0+s2
+	extr.u	te13=s3,16,8	}	// 4/0:s1>>16&0xff
+{ .mmi;	ld1	te00=[te00]		// 4/0:te0[s0>>24]
+	add	te01=te01,te0		// 4/1:te0+s1>>24
+	shr.u	te11=s1,sixteen	};;	// 4/2:s3>>16
+{ .mmi;	ld1	te21=[te21]		// 5/3:te0[s1>>8]
+	add	te13=te13,te0		// 5/0:te0+s1>>16
+	extr.u	te10=s0,16,8	}	// 5/1:s2>>16&0xff
+{ .mmi;	ld1	te01=[te01]		// 5/1:te0[s1>>24]
+	add	te02=te02,te0		// 5/2:te0+s2>>24
+	and	te33=s3,maskff	};;	// 5/2:s1&0xff
+{ .mmi;	ld1	te13=[te13]		// 6/0:te0[s1>>16]
+	add	te10=te10,te0		// 6/1:te0+s2>>16
+	extr.u	te12=s2,16,8	}	// 6/3:s0>>16&0xff
+{ .mmi;	ld1	te02=[te02]		// 6/2:te0[s2>>24]
+	add	te03=te03,te0		// 6/3:te0+s0>>16
+	and	te30=s0,maskff	};;	// 6/3:s2&0xff
+
+{ .mmi;	ld1	te10=[te10]		// 7/1:te0[s2>>16]
+	add	te33=te33,te0		// 7/2:te0+s1&0xff
+	dep	te31=te22,te31,8,8}	// 7/0:
+{ .mmi;	ld1	te03=[te03]		// 7/3:te0[s3>>24]
+	add	te30=te30,te0		// 7/3:te0+s2
+	and	te11=te11,maskff};;	// 7/2:s3>>16&0xff
+{ .mmi;	ld1	te33=[te33]		// 8/2:te0[s1]
+	add	te11=te11,te0		// 8/2:te0+s3>>16
+	dep	te32=te23,te32,8,8}	// 8/1:
+{ .mmi;	ld1	te30=[te30]		// 8/3:te0[s2]
+	add	te12=te12,te0		// 8/3:te0+s0>>16
+	shl	te00=te00,twenty4};;	// 8/0:
+{ .mii;	ld1	te11=[te11]		// 9/2:te0[s3>>16]
+	dep	te31=te13,te31,16,8	// 9/0:
+	shl	te01=te01,twenty4};;	// 9/1:
+{ .mii;	ld1	te12=[te12]		// 10/3:te0[s0>>16]
+	dep	te33=te20,te33,8,8	// 10/2:
+	shl	te02=te02,twenty4};;	// 10/2:
+{ .mii;	xor	t0=t0,te31		// 11/0:
+	dep	te30=te21,te30,8,8	// 11/3:
+	shl	te10=te10,sixteen};;	// 11/1:
+{ .mii;	xor	r16=t0,te00		// 12/0:done!
+	dep	te33=te11,te33,16,8	// 12/2:
+	shl	te03=te03,twenty4};;	// 12/3:
+{ .mmi;	xor	t1=t1,te01		// 13/1:
+	xor	t2=t2,te02		// 13/2:
+	dep	te30=te12,te30,16,8};;	// 13/3:
+{ .mmi;	xor	t1=t1,te32		// 14/1:
+	xor	r24=t2,te33		// 14/2:done!
+	xor	t3=t3,te30	};;	// 14/3:
+{ .mib;	xor	r20=t1,te10		// 15/1:done!
+	xor	r28=t3,te03		// 15/3:done!
+	br.ret.sptk	b6	};;
+.endp	_ia64_AES_decrypt#
+
+// void AES_decrypt (const void *in,void *out,const AES_KEY *key);
+.global	AES_decrypt#
+.proc	AES_decrypt#
+.align	32
+AES_decrypt:
+	.prologue
+	.save	ar.pfs,pfssave
+{ .mmi;	alloc	pfssave=ar.pfs,3,1,12,0
+	and	out0=3,in0
+	mov	r3=ip			}
+{ .mmi;	ADDP	in0=0,in0
+	mov	loc0=psr.um
+	ADDP	out11=KSZ*60,in2	};;	// &AES_KEY->rounds
+
+{ .mmi;	ld4	out11=[out11]			// AES_KEY->rounds
+	add	out8=(AES_Td#-AES_decrypt#),r3	// Te0
+	.save	pr,prsave
+	mov	prsave=pr		}
+{ .mmi;	rum	1<<3				// clear um.ac
+	.save	ar.lc,lcsave
+	mov	lcsave=ar.lc		};;
+
+	.body
+#if defined(_HPUX_SOURCE)	// HPUX is big-endian, cut 15+15 cycles...
+{ .mib; cmp.ne	p6,p0=out0,r0
+	add	out0=4,in0
+(p6)	br.dpnt.many	.Ld_i_unaligned	};;
+
+{ .mmi;	ld4	out1=[in0],8		// s0
+	and	out9=3,in1
+	mov	twenty4=24		}
+{ .mmi;	ld4	out3=[out0],8		// s1
+	ADDP	rk0=0,in2
+	mov	sixteen=16		};;
+{ .mmi;	ld4	out5=[in0]		// s2
+	cmp.ne	p6,p0=out9,r0
+	mov	maskff=0xff		}
+{ .mmb;	ld4	out7=[out0]		// s3
+	ADDP	rk1=KSZ,in2
+	br.call.sptk.many	b6=_ia64_AES_decrypt	};;
+
+{ .mib;	ADDP	in0=4,in1
+	ADDP	in1=0,in1
+(p6)	br.spnt	.Ld_o_unaligned		};;
+
+{ .mii;	mov	psr.um=loc0
+	mov	ar.pfs=pfssave
+	mov	ar.lc=lcsave		};;
+{ .mmi;	st4	[in1]=r16,8		// s0
+	st4	[in0]=r20,8		// s1
+	mov	pr=prsave,0x1ffff	};;
+{ .mmb;	st4	[in1]=r24		// s2
+	st4	[in0]=r28		// s3
+	br.ret.sptk.many	b0	};;
+#endif
+
+.align	32
+.Ld_i_unaligned:
+{ .mmi;	add	out0=1,in0
+	add	out2=2,in0
+	add	out4=3,in0	};;
+{ .mmi;	ld1	r16=[in0],4
+	ld1	r17=[out0],4	}//;;
+{ .mmi;	ld1	r18=[out2],4
+	ld1	out1=[out4],4	};;	// s0
+{ .mmi;	ld1	r20=[in0],4
+	ld1	r21=[out0],4	}//;;
+{ .mmi;	ld1	r22=[out2],4
+	ld1	out3=[out4],4	};;	// s1
+{ .mmi;	ld1	r24=[in0],4
+	ld1	r25=[out0],4	}//;;
+{ .mmi;	ld1	r26=[out2],4
+	ld1	out5=[out4],4	};;	// s2
+{ .mmi;	ld1	r28=[in0]
+	ld1	r29=[out0]	}//;;
+{ .mmi;	ld1	r30=[out2]
+	ld1	out7=[out4]	};;	// s3
+
+{ .mii;
+	dep	out1=r16,out1,24,8	//;;
+	dep	out3=r20,out3,24,8	}//;;
+{ .mii;	ADDP	rk0=0,in2
+	dep	out5=r24,out5,24,8	//;;
+	dep	out7=r28,out7,24,8	};;
+{ .mii;	ADDP	rk1=KSZ,in2
+	dep	out1=r17,out1,16,8	//;;
+	dep	out3=r21,out3,16,8	}//;;
+{ .mii;	mov	twenty4=24
+	dep	out5=r25,out5,16,8	//;;
+	dep	out7=r29,out7,16,8	};;
+{ .mii;	mov	sixteen=16
+	dep	out1=r18,out1,8,8	//;;
+	dep	out3=r22,out3,8,8	}//;;
+{ .mii;	mov	maskff=0xff
+	dep	out5=r26,out5,8,8	//;;
+	dep	out7=r30,out7,8,8	};;
+
+{ .mib;	br.call.sptk.many	b6=_ia64_AES_decrypt	};;
+
+.Ld_o_unaligned:
+{ .mii;	ADDP	out0=0,in1
+	extr.u	r17=r16,8,8			// s0
+	shr.u	r19=r16,twenty4		}//;;
+{ .mii;	ADDP	out1=1,in1
+	extr.u	r18=r16,16,8
+	shr.u	r23=r20,twenty4		}//;;	// s1
+{ .mii;	ADDP	out2=2,in1
+	extr.u	r21=r20,8,8
+	shr.u	r22=r20,sixteen		}//;;
+{ .mii;	ADDP	out3=3,in1
+	extr.u	r25=r24,8,8			// s2
+	shr.u	r27=r24,twenty4		};;
+{ .mii;	st1	[out3]=r16,4
+	extr.u	r26=r24,16,8
+	shr.u	r31=r28,twenty4		}//;;	// s3
+{ .mii;	st1	[out2]=r17,4
+	extr.u	r29=r28,8,8
+	shr.u	r30=r28,sixteen		}//;;
+
+{ .mmi;	st1	[out1]=r18,4
+	st1	[out0]=r19,4		};;
+{ .mmi;	st1	[out3]=r20,4
+	st1	[out2]=r21,4		}//;;
+{ .mmi;	st1	[out1]=r22,4
+	st1	[out0]=r23,4		};;
+{ .mmi;	st1	[out3]=r24,4
+	st1	[out2]=r25,4
+	mov	pr=prsave,0x1ffff	}//;;
+{ .mmi;	st1	[out1]=r26,4
+	st1	[out0]=r27,4
+	mov	ar.pfs=pfssave		};;
+{ .mmi;	st1	[out3]=r28
+	st1	[out2]=r29
+	mov	ar.lc=lcsave		}//;;
+{ .mmi;	st1	[out1]=r30
+	st1	[out0]=r31		}
+{ .mfb;	mov	psr.um=loc0			// restore user mask
+	br.ret.sptk.many	b0	};;
+.endp	AES_decrypt#
+
+// leave it in .text segment...
+.align	64
+.global	AES_Te#
+.type	AES_Te#,@object
+AES_Te:	data4	0xc66363a5,0xc66363a5, 0xf87c7c84,0xf87c7c84
+	data4	0xee777799,0xee777799, 0xf67b7b8d,0xf67b7b8d
+	data4	0xfff2f20d,0xfff2f20d, 0xd66b6bbd,0xd66b6bbd
+	data4	0xde6f6fb1,0xde6f6fb1, 0x91c5c554,0x91c5c554
+	data4	0x60303050,0x60303050, 0x02010103,0x02010103
+	data4	0xce6767a9,0xce6767a9, 0x562b2b7d,0x562b2b7d
+	data4	0xe7fefe19,0xe7fefe19, 0xb5d7d762,0xb5d7d762
+	data4	0x4dababe6,0x4dababe6, 0xec76769a,0xec76769a
+	data4	0x8fcaca45,0x8fcaca45, 0x1f82829d,0x1f82829d
+	data4	0x89c9c940,0x89c9c940, 0xfa7d7d87,0xfa7d7d87
+	data4	0xeffafa15,0xeffafa15, 0xb25959eb,0xb25959eb
+	data4	0x8e4747c9,0x8e4747c9, 0xfbf0f00b,0xfbf0f00b
+	data4	0x41adadec,0x41adadec, 0xb3d4d467,0xb3d4d467
+	data4	0x5fa2a2fd,0x5fa2a2fd, 0x45afafea,0x45afafea
+	data4	0x239c9cbf,0x239c9cbf, 0x53a4a4f7,0x53a4a4f7
+	data4	0xe4727296,0xe4727296, 0x9bc0c05b,0x9bc0c05b
+	data4	0x75b7b7c2,0x75b7b7c2, 0xe1fdfd1c,0xe1fdfd1c
+	data4	0x3d9393ae,0x3d9393ae, 0x4c26266a,0x4c26266a
+	data4	0x6c36365a,0x6c36365a, 0x7e3f3f41,0x7e3f3f41
+	data4	0xf5f7f702,0xf5f7f702, 0x83cccc4f,0x83cccc4f
+	data4	0x6834345c,0x6834345c, 0x51a5a5f4,0x51a5a5f4
+	data4	0xd1e5e534,0xd1e5e534, 0xf9f1f108,0xf9f1f108
+	data4	0xe2717193,0xe2717193, 0xabd8d873,0xabd8d873
+	data4	0x62313153,0x62313153, 0x2a15153f,0x2a15153f
+	data4	0x0804040c,0x0804040c, 0x95c7c752,0x95c7c752
+	data4	0x46232365,0x46232365, 0x9dc3c35e,0x9dc3c35e
+	data4	0x30181828,0x30181828, 0x379696a1,0x379696a1
+	data4	0x0a05050f,0x0a05050f, 0x2f9a9ab5,0x2f9a9ab5
+	data4	0x0e070709,0x0e070709, 0x24121236,0x24121236
+	data4	0x1b80809b,0x1b80809b, 0xdfe2e23d,0xdfe2e23d
+	data4	0xcdebeb26,0xcdebeb26, 0x4e272769,0x4e272769
+	data4	0x7fb2b2cd,0x7fb2b2cd, 0xea75759f,0xea75759f
+	data4	0x1209091b,0x1209091b, 0x1d83839e,0x1d83839e
+	data4	0x582c2c74,0x582c2c74, 0x341a1a2e,0x341a1a2e
+	data4	0x361b1b2d,0x361b1b2d, 0xdc6e6eb2,0xdc6e6eb2
+	data4	0xb45a5aee,0xb45a5aee, 0x5ba0a0fb,0x5ba0a0fb
+	data4	0xa45252f6,0xa45252f6, 0x763b3b4d,0x763b3b4d
+	data4	0xb7d6d661,0xb7d6d661, 0x7db3b3ce,0x7db3b3ce
+	data4	0x5229297b,0x5229297b, 0xdde3e33e,0xdde3e33e
+	data4	0x5e2f2f71,0x5e2f2f71, 0x13848497,0x13848497
+	data4	0xa65353f5,0xa65353f5, 0xb9d1d168,0xb9d1d168
+	data4	0x00000000,0x00000000, 0xc1eded2c,0xc1eded2c
+	data4	0x40202060,0x40202060, 0xe3fcfc1f,0xe3fcfc1f
+	data4	0x79b1b1c8,0x79b1b1c8, 0xb65b5bed,0xb65b5bed
+	data4	0xd46a6abe,0xd46a6abe, 0x8dcbcb46,0x8dcbcb46
+	data4	0x67bebed9,0x67bebed9, 0x7239394b,0x7239394b
+	data4	0x944a4ade,0x944a4ade, 0x984c4cd4,0x984c4cd4
+	data4	0xb05858e8,0xb05858e8, 0x85cfcf4a,0x85cfcf4a
+	data4	0xbbd0d06b,0xbbd0d06b, 0xc5efef2a,0xc5efef2a
+	data4	0x4faaaae5,0x4faaaae5, 0xedfbfb16,0xedfbfb16
+	data4	0x864343c5,0x864343c5, 0x9a4d4dd7,0x9a4d4dd7
+	data4	0x66333355,0x66333355, 0x11858594,0x11858594
+	data4	0x8a4545cf,0x8a4545cf, 0xe9f9f910,0xe9f9f910
+	data4	0x04020206,0x04020206, 0xfe7f7f81,0xfe7f7f81
+	data4	0xa05050f0,0xa05050f0, 0x783c3c44,0x783c3c44
+	data4	0x259f9fba,0x259f9fba, 0x4ba8a8e3,0x4ba8a8e3
+	data4	0xa25151f3,0xa25151f3, 0x5da3a3fe,0x5da3a3fe
+	data4	0x804040c0,0x804040c0, 0x058f8f8a,0x058f8f8a
+	data4	0x3f9292ad,0x3f9292ad, 0x219d9dbc,0x219d9dbc
+	data4	0x70383848,0x70383848, 0xf1f5f504,0xf1f5f504
+	data4	0x63bcbcdf,0x63bcbcdf, 0x77b6b6c1,0x77b6b6c1
+	data4	0xafdada75,0xafdada75, 0x42212163,0x42212163
+	data4	0x20101030,0x20101030, 0xe5ffff1a,0xe5ffff1a
+	data4	0xfdf3f30e,0xfdf3f30e, 0xbfd2d26d,0xbfd2d26d
+	data4	0x81cdcd4c,0x81cdcd4c, 0x180c0c14,0x180c0c14
+	data4	0x26131335,0x26131335, 0xc3ecec2f,0xc3ecec2f
+	data4	0xbe5f5fe1,0xbe5f5fe1, 0x359797a2,0x359797a2
+	data4	0x884444cc,0x884444cc, 0x2e171739,0x2e171739
+	data4	0x93c4c457,0x93c4c457, 0x55a7a7f2,0x55a7a7f2
+	data4	0xfc7e7e82,0xfc7e7e82, 0x7a3d3d47,0x7a3d3d47
+	data4	0xc86464ac,0xc86464ac, 0xba5d5de7,0xba5d5de7
+	data4	0x3219192b,0x3219192b, 0xe6737395,0xe6737395
+	data4	0xc06060a0,0xc06060a0, 0x19818198,0x19818198
+	data4	0x9e4f4fd1,0x9e4f4fd1, 0xa3dcdc7f,0xa3dcdc7f
+	data4	0x44222266,0x44222266, 0x542a2a7e,0x542a2a7e
+	data4	0x3b9090ab,0x3b9090ab, 0x0b888883,0x0b888883
+	data4	0x8c4646ca,0x8c4646ca, 0xc7eeee29,0xc7eeee29
+	data4	0x6bb8b8d3,0x6bb8b8d3, 0x2814143c,0x2814143c
+	data4	0xa7dede79,0xa7dede79, 0xbc5e5ee2,0xbc5e5ee2
+	data4	0x160b0b1d,0x160b0b1d, 0xaddbdb76,0xaddbdb76
+	data4	0xdbe0e03b,0xdbe0e03b, 0x64323256,0x64323256
+	data4	0x743a3a4e,0x743a3a4e, 0x140a0a1e,0x140a0a1e
+	data4	0x924949db,0x924949db, 0x0c06060a,0x0c06060a
+	data4	0x4824246c,0x4824246c, 0xb85c5ce4,0xb85c5ce4
+	data4	0x9fc2c25d,0x9fc2c25d, 0xbdd3d36e,0xbdd3d36e
+	data4	0x43acacef,0x43acacef, 0xc46262a6,0xc46262a6
+	data4	0x399191a8,0x399191a8, 0x319595a4,0x319595a4
+	data4	0xd3e4e437,0xd3e4e437, 0xf279798b,0xf279798b
+	data4	0xd5e7e732,0xd5e7e732, 0x8bc8c843,0x8bc8c843
+	data4	0x6e373759,0x6e373759, 0xda6d6db7,0xda6d6db7
+	data4	0x018d8d8c,0x018d8d8c, 0xb1d5d564,0xb1d5d564
+	data4	0x9c4e4ed2,0x9c4e4ed2, 0x49a9a9e0,0x49a9a9e0
+	data4	0xd86c6cb4,0xd86c6cb4, 0xac5656fa,0xac5656fa
+	data4	0xf3f4f407,0xf3f4f407, 0xcfeaea25,0xcfeaea25
+	data4	0xca6565af,0xca6565af, 0xf47a7a8e,0xf47a7a8e
+	data4	0x47aeaee9,0x47aeaee9, 0x10080818,0x10080818
+	data4	0x6fbabad5,0x6fbabad5, 0xf0787888,0xf0787888
+	data4	0x4a25256f,0x4a25256f, 0x5c2e2e72,0x5c2e2e72
+	data4	0x381c1c24,0x381c1c24, 0x57a6a6f1,0x57a6a6f1
+	data4	0x73b4b4c7,0x73b4b4c7, 0x97c6c651,0x97c6c651
+	data4	0xcbe8e823,0xcbe8e823, 0xa1dddd7c,0xa1dddd7c
+	data4	0xe874749c,0xe874749c, 0x3e1f1f21,0x3e1f1f21
+	data4	0x964b4bdd,0x964b4bdd, 0x61bdbddc,0x61bdbddc
+	data4	0x0d8b8b86,0x0d8b8b86, 0x0f8a8a85,0x0f8a8a85
+	data4	0xe0707090,0xe0707090, 0x7c3e3e42,0x7c3e3e42
+	data4	0x71b5b5c4,0x71b5b5c4, 0xcc6666aa,0xcc6666aa
+	data4	0x904848d8,0x904848d8, 0x06030305,0x06030305
+	data4	0xf7f6f601,0xf7f6f601, 0x1c0e0e12,0x1c0e0e12
+	data4	0xc26161a3,0xc26161a3, 0x6a35355f,0x6a35355f
+	data4	0xae5757f9,0xae5757f9, 0x69b9b9d0,0x69b9b9d0
+	data4	0x17868691,0x17868691, 0x99c1c158,0x99c1c158
+	data4	0x3a1d1d27,0x3a1d1d27, 0x279e9eb9,0x279e9eb9
+	data4	0xd9e1e138,0xd9e1e138, 0xebf8f813,0xebf8f813
+	data4	0x2b9898b3,0x2b9898b3, 0x22111133,0x22111133
+	data4	0xd26969bb,0xd26969bb, 0xa9d9d970,0xa9d9d970
+	data4	0x078e8e89,0x078e8e89, 0x339494a7,0x339494a7
+	data4	0x2d9b9bb6,0x2d9b9bb6, 0x3c1e1e22,0x3c1e1e22
+	data4	0x15878792,0x15878792, 0xc9e9e920,0xc9e9e920
+	data4	0x87cece49,0x87cece49, 0xaa5555ff,0xaa5555ff
+	data4	0x50282878,0x50282878, 0xa5dfdf7a,0xa5dfdf7a
+	data4	0x038c8c8f,0x038c8c8f, 0x59a1a1f8,0x59a1a1f8
+	data4	0x09898980,0x09898980, 0x1a0d0d17,0x1a0d0d17
+	data4	0x65bfbfda,0x65bfbfda, 0xd7e6e631,0xd7e6e631
+	data4	0x844242c6,0x844242c6, 0xd06868b8,0xd06868b8
+	data4	0x824141c3,0x824141c3, 0x299999b0,0x299999b0
+	data4	0x5a2d2d77,0x5a2d2d77, 0x1e0f0f11,0x1e0f0f11
+	data4	0x7bb0b0cb,0x7bb0b0cb, 0xa85454fc,0xa85454fc
+	data4	0x6dbbbbd6,0x6dbbbbd6, 0x2c16163a,0x2c16163a
+// Te4:
+	data1	0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5
+	data1	0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76
+	data1	0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0
+	data1	0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0
+	data1	0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc
+	data1	0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15
+	data1	0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a
+	data1	0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75
+	data1	0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0
+	data1	0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84
+	data1	0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b
+	data1	0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf
+	data1	0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85
+	data1	0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8
+	data1	0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5
+	data1	0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2
+	data1	0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17
+	data1	0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73
+	data1	0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88
+	data1	0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb
+	data1	0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c
+	data1	0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79
+	data1	0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9
+	data1	0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08
+	data1	0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6
+	data1	0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a
+	data1	0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e
+	data1	0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e
+	data1	0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94
+	data1	0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf
+	data1	0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68
+	data1	0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16
+.size	AES_Te#,2048+256	// HP-UX assembler fails to ".-AES_Te#"
+
+.align	64
+.global	AES_Td#
+.type	AES_Td#,@object
+AES_Td:	data4	0x51f4a750,0x51f4a750, 0x7e416553,0x7e416553
+	data4	0x1a17a4c3,0x1a17a4c3, 0x3a275e96,0x3a275e96
+	data4	0x3bab6bcb,0x3bab6bcb, 0x1f9d45f1,0x1f9d45f1
+	data4	0xacfa58ab,0xacfa58ab, 0x4be30393,0x4be30393
+	data4	0x2030fa55,0x2030fa55, 0xad766df6,0xad766df6
+	data4	0x88cc7691,0x88cc7691, 0xf5024c25,0xf5024c25
+	data4	0x4fe5d7fc,0x4fe5d7fc, 0xc52acbd7,0xc52acbd7
+	data4	0x26354480,0x26354480, 0xb562a38f,0xb562a38f
+	data4	0xdeb15a49,0xdeb15a49, 0x25ba1b67,0x25ba1b67
+	data4	0x45ea0e98,0x45ea0e98, 0x5dfec0e1,0x5dfec0e1
+	data4	0xc32f7502,0xc32f7502, 0x814cf012,0x814cf012
+	data4	0x8d4697a3,0x8d4697a3, 0x6bd3f9c6,0x6bd3f9c6
+	data4	0x038f5fe7,0x038f5fe7, 0x15929c95,0x15929c95
+	data4	0xbf6d7aeb,0xbf6d7aeb, 0x955259da,0x955259da
+	data4	0xd4be832d,0xd4be832d, 0x587421d3,0x587421d3
+	data4	0x49e06929,0x49e06929, 0x8ec9c844,0x8ec9c844
+	data4	0x75c2896a,0x75c2896a, 0xf48e7978,0xf48e7978
+	data4	0x99583e6b,0x99583e6b, 0x27b971dd,0x27b971dd
+	data4	0xbee14fb6,0xbee14fb6, 0xf088ad17,0xf088ad17
+	data4	0xc920ac66,0xc920ac66, 0x7dce3ab4,0x7dce3ab4
+	data4	0x63df4a18,0x63df4a18, 0xe51a3182,0xe51a3182
+	data4	0x97513360,0x97513360, 0x62537f45,0x62537f45
+	data4	0xb16477e0,0xb16477e0, 0xbb6bae84,0xbb6bae84
+	data4	0xfe81a01c,0xfe81a01c, 0xf9082b94,0xf9082b94
+	data4	0x70486858,0x70486858, 0x8f45fd19,0x8f45fd19
+	data4	0x94de6c87,0x94de6c87, 0x527bf8b7,0x527bf8b7
+	data4	0xab73d323,0xab73d323, 0x724b02e2,0x724b02e2
+	data4	0xe31f8f57,0xe31f8f57, 0x6655ab2a,0x6655ab2a
+	data4	0xb2eb2807,0xb2eb2807, 0x2fb5c203,0x2fb5c203
+	data4	0x86c57b9a,0x86c57b9a, 0xd33708a5,0xd33708a5
+	data4	0x302887f2,0x302887f2, 0x23bfa5b2,0x23bfa5b2
+	data4	0x02036aba,0x02036aba, 0xed16825c,0xed16825c
+	data4	0x8acf1c2b,0x8acf1c2b, 0xa779b492,0xa779b492
+	data4	0xf307f2f0,0xf307f2f0, 0x4e69e2a1,0x4e69e2a1
+	data4	0x65daf4cd,0x65daf4cd, 0x0605bed5,0x0605bed5
+	data4	0xd134621f,0xd134621f, 0xc4a6fe8a,0xc4a6fe8a
+	data4	0x342e539d,0x342e539d, 0xa2f355a0,0xa2f355a0
+	data4	0x058ae132,0x058ae132, 0xa4f6eb75,0xa4f6eb75
+	data4	0x0b83ec39,0x0b83ec39, 0x4060efaa,0x4060efaa
+	data4	0x5e719f06,0x5e719f06, 0xbd6e1051,0xbd6e1051
+	data4	0x3e218af9,0x3e218af9, 0x96dd063d,0x96dd063d
+	data4	0xdd3e05ae,0xdd3e05ae, 0x4de6bd46,0x4de6bd46
+	data4	0x91548db5,0x91548db5, 0x71c45d05,0x71c45d05
+	data4	0x0406d46f,0x0406d46f, 0x605015ff,0x605015ff
+	data4	0x1998fb24,0x1998fb24, 0xd6bde997,0xd6bde997
+	data4	0x894043cc,0x894043cc, 0x67d99e77,0x67d99e77
+	data4	0xb0e842bd,0xb0e842bd, 0x07898b88,0x07898b88
+	data4	0xe7195b38,0xe7195b38, 0x79c8eedb,0x79c8eedb
+	data4	0xa17c0a47,0xa17c0a47, 0x7c420fe9,0x7c420fe9
+	data4	0xf8841ec9,0xf8841ec9, 0x00000000,0x00000000
+	data4	0x09808683,0x09808683, 0x322bed48,0x322bed48
+	data4	0x1e1170ac,0x1e1170ac, 0x6c5a724e,0x6c5a724e
+	data4	0xfd0efffb,0xfd0efffb, 0x0f853856,0x0f853856
+	data4	0x3daed51e,0x3daed51e, 0x362d3927,0x362d3927
+	data4	0x0a0fd964,0x0a0fd964, 0x685ca621,0x685ca621
+	data4	0x9b5b54d1,0x9b5b54d1, 0x24362e3a,0x24362e3a
+	data4	0x0c0a67b1,0x0c0a67b1, 0x9357e70f,0x9357e70f
+	data4	0xb4ee96d2,0xb4ee96d2, 0x1b9b919e,0x1b9b919e
+	data4	0x80c0c54f,0x80c0c54f, 0x61dc20a2,0x61dc20a2
+	data4	0x5a774b69,0x5a774b69, 0x1c121a16,0x1c121a16
+	data4	0xe293ba0a,0xe293ba0a, 0xc0a02ae5,0xc0a02ae5
+	data4	0x3c22e043,0x3c22e043, 0x121b171d,0x121b171d
+	data4	0x0e090d0b,0x0e090d0b, 0xf28bc7ad,0xf28bc7ad
+	data4	0x2db6a8b9,0x2db6a8b9, 0x141ea9c8,0x141ea9c8
+	data4	0x57f11985,0x57f11985, 0xaf75074c,0xaf75074c
+	data4	0xee99ddbb,0xee99ddbb, 0xa37f60fd,0xa37f60fd
+	data4	0xf701269f,0xf701269f, 0x5c72f5bc,0x5c72f5bc
+	data4	0x44663bc5,0x44663bc5, 0x5bfb7e34,0x5bfb7e34
+	data4	0x8b432976,0x8b432976, 0xcb23c6dc,0xcb23c6dc
+	data4	0xb6edfc68,0xb6edfc68, 0xb8e4f163,0xb8e4f163
+	data4	0xd731dcca,0xd731dcca, 0x42638510,0x42638510
+	data4	0x13972240,0x13972240, 0x84c61120,0x84c61120
+	data4	0x854a247d,0x854a247d, 0xd2bb3df8,0xd2bb3df8
+	data4	0xaef93211,0xaef93211, 0xc729a16d,0xc729a16d
+	data4	0x1d9e2f4b,0x1d9e2f4b, 0xdcb230f3,0xdcb230f3
+	data4	0x0d8652ec,0x0d8652ec, 0x77c1e3d0,0x77c1e3d0
+	data4	0x2bb3166c,0x2bb3166c, 0xa970b999,0xa970b999
+	data4	0x119448fa,0x119448fa, 0x47e96422,0x47e96422
+	data4	0xa8fc8cc4,0xa8fc8cc4, 0xa0f03f1a,0xa0f03f1a
+	data4	0x567d2cd8,0x567d2cd8, 0x223390ef,0x223390ef
+	data4	0x87494ec7,0x87494ec7, 0xd938d1c1,0xd938d1c1
+	data4	0x8ccaa2fe,0x8ccaa2fe, 0x98d40b36,0x98d40b36
+	data4	0xa6f581cf,0xa6f581cf, 0xa57ade28,0xa57ade28
+	data4	0xdab78e26,0xdab78e26, 0x3fadbfa4,0x3fadbfa4
+	data4	0x2c3a9de4,0x2c3a9de4, 0x5078920d,0x5078920d
+	data4	0x6a5fcc9b,0x6a5fcc9b, 0x547e4662,0x547e4662
+	data4	0xf68d13c2,0xf68d13c2, 0x90d8b8e8,0x90d8b8e8
+	data4	0x2e39f75e,0x2e39f75e, 0x82c3aff5,0x82c3aff5
+	data4	0x9f5d80be,0x9f5d80be, 0x69d0937c,0x69d0937c
+	data4	0x6fd52da9,0x6fd52da9, 0xcf2512b3,0xcf2512b3
+	data4	0xc8ac993b,0xc8ac993b, 0x10187da7,0x10187da7
+	data4	0xe89c636e,0xe89c636e, 0xdb3bbb7b,0xdb3bbb7b
+	data4	0xcd267809,0xcd267809, 0x6e5918f4,0x6e5918f4
+	data4	0xec9ab701,0xec9ab701, 0x834f9aa8,0x834f9aa8
+	data4	0xe6956e65,0xe6956e65, 0xaaffe67e,0xaaffe67e
+	data4	0x21bccf08,0x21bccf08, 0xef15e8e6,0xef15e8e6
+	data4	0xbae79bd9,0xbae79bd9, 0x4a6f36ce,0x4a6f36ce
+	data4	0xea9f09d4,0xea9f09d4, 0x29b07cd6,0x29b07cd6
+	data4	0x31a4b2af,0x31a4b2af, 0x2a3f2331,0x2a3f2331
+	data4	0xc6a59430,0xc6a59430, 0x35a266c0,0x35a266c0
+	data4	0x744ebc37,0x744ebc37, 0xfc82caa6,0xfc82caa6
+	data4	0xe090d0b0,0xe090d0b0, 0x33a7d815,0x33a7d815
+	data4	0xf104984a,0xf104984a, 0x41ecdaf7,0x41ecdaf7
+	data4	0x7fcd500e,0x7fcd500e, 0x1791f62f,0x1791f62f
+	data4	0x764dd68d,0x764dd68d, 0x43efb04d,0x43efb04d
+	data4	0xccaa4d54,0xccaa4d54, 0xe49604df,0xe49604df
+	data4	0x9ed1b5e3,0x9ed1b5e3, 0x4c6a881b,0x4c6a881b
+	data4	0xc12c1fb8,0xc12c1fb8, 0x4665517f,0x4665517f
+	data4	0x9d5eea04,0x9d5eea04, 0x018c355d,0x018c355d
+	data4	0xfa877473,0xfa877473, 0xfb0b412e,0xfb0b412e
+	data4	0xb3671d5a,0xb3671d5a, 0x92dbd252,0x92dbd252
+	data4	0xe9105633,0xe9105633, 0x6dd64713,0x6dd64713
+	data4	0x9ad7618c,0x9ad7618c, 0x37a10c7a,0x37a10c7a
+	data4	0x59f8148e,0x59f8148e, 0xeb133c89,0xeb133c89
+	data4	0xcea927ee,0xcea927ee, 0xb761c935,0xb761c935
+	data4	0xe11ce5ed,0xe11ce5ed, 0x7a47b13c,0x7a47b13c
+	data4	0x9cd2df59,0x9cd2df59, 0x55f2733f,0x55f2733f
+	data4	0x1814ce79,0x1814ce79, 0x73c737bf,0x73c737bf
+	data4	0x53f7cdea,0x53f7cdea, 0x5ffdaa5b,0x5ffdaa5b
+	data4	0xdf3d6f14,0xdf3d6f14, 0x7844db86,0x7844db86
+	data4	0xcaaff381,0xcaaff381, 0xb968c43e,0xb968c43e
+	data4	0x3824342c,0x3824342c, 0xc2a3405f,0xc2a3405f
+	data4	0x161dc372,0x161dc372, 0xbce2250c,0xbce2250c
+	data4	0x283c498b,0x283c498b, 0xff0d9541,0xff0d9541
+	data4	0x39a80171,0x39a80171, 0x080cb3de,0x080cb3de
+	data4	0xd8b4e49c,0xd8b4e49c, 0x6456c190,0x6456c190
+	data4	0x7bcb8461,0x7bcb8461, 0xd532b670,0xd532b670
+	data4	0x486c5c74,0x486c5c74, 0xd0b85742,0xd0b85742
+// Td4:
+	data1	0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38
+	data1	0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb
+	data1	0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87
+	data1	0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb
+	data1	0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d
+	data1	0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e
+	data1	0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2
+	data1	0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25
+	data1	0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16
+	data1	0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92
+	data1	0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda
+	data1	0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84
+	data1	0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a
+	data1	0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06
+	data1	0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02
+	data1	0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b
+	data1	0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea
+	data1	0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73
+	data1	0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85
+	data1	0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e
+	data1	0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89
+	data1	0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b
+	data1	0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20
+	data1	0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4
+	data1	0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31
+	data1	0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f
+	data1	0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d
+	data1	0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef
+	data1	0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0
+	data1	0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61
+	data1	0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26
+	data1	0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d
+.size	AES_Td#,2048+256	// HP-UX assembler fails to ".-AES_Td#"
diff --git a/openssl/crypto/aes/asm/aes-ppc.pl b/openssl/crypto/aes/asm/aes-ppc.pl
new file mode 100644
index 00000000..f82c5e18
--- /dev/null
+++ b/openssl/crypto/aes/asm/aes-ppc.pl
@@ -0,0 +1,1189 @@
+#!/usr/bin/env perl
+
+# ====================================================================
+# Written by Andy Polyakov <appro@fy.chalmers.se> 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/.
+# ====================================================================
+
+# 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
+# 4.0. But these are not the ones currently used! Their "compact"
+# counterparts are, for security reason. ppc_AES_encrypt_compact runs
+# 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 gives 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;
+	$STU	="stdu";
+	$POP	="ld";
+	$PUSH	="std";
+} elsif ($flavour =~ /32/) {
+	$SIZE_T	=4;
+	$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 ".shift || die "can't call $xlate: $!";
+
+$FRAME=32*$SIZE_T;
+
+sub _data_word()
+{ my $i;
+    while(defined($i=shift)) { $code.=sprintf"\t.long\t0x%08x,0x%08x\n",$i,$i; }
+}
+
+$sp="r1";
+$toc="r2";
+$inp="r3";
+$out="r4";
+$key="r5";
+
+$Tbl0="r3";
+$Tbl1="r6";
+$Tbl2="r7";
+$Tbl3="r2";
+
+$s0="r8";
+$s1="r9";
+$s2="r10";
+$s3="r11";
+
+$t0="r12";
+$t1="r13";
+$t2="r14";
+$t3="r15";
+
+$acc00="r16";
+$acc01="r17";
+$acc02="r18";
+$acc03="r19";
+
+$acc04="r20";
+$acc05="r21";
+$acc06="r22";
+$acc07="r23";
+
+$acc08="r24";
+$acc09="r25";
+$acc10="r26";
+$acc11="r27";
+
+$acc12="r28";
+$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=$Tbl3;
+
+$code.=<<___;
+.machine	"any"
+.text
+
+.align	7
+LAES_Te:
+	mflr	r0
+	bcl	20,31,\$+4
+	mflr	$Tbl0	;    vvvvv "distance" between . and 1st data entry
+	addi	$Tbl0,$Tbl0,`128-8`
+	mtlr	r0
+	blr
+	.space	`32-24`
+LAES_Td:
+	mflr	r0
+	bcl	20,31,\$+4
+	mflr	$Tbl0	;    vvvvvvvv "distance" between . and 1st data entry
+	addi	$Tbl0,$Tbl0,`128-8-32+2048+256`
+	mtlr	r0
+	blr
+	.space	`128-32-24`
+___
+&_data_word(
+	0xc66363a5, 0xf87c7c84, 0xee777799, 0xf67b7b8d,
+	0xfff2f20d, 0xd66b6bbd, 0xde6f6fb1, 0x91c5c554,
+	0x60303050, 0x02010103, 0xce6767a9, 0x562b2b7d,
+	0xe7fefe19, 0xb5d7d762, 0x4dababe6, 0xec76769a,
+	0x8fcaca45, 0x1f82829d, 0x89c9c940, 0xfa7d7d87,
+	0xeffafa15, 0xb25959eb, 0x8e4747c9, 0xfbf0f00b,
+	0x41adadec, 0xb3d4d467, 0x5fa2a2fd, 0x45afafea,
+	0x239c9cbf, 0x53a4a4f7, 0xe4727296, 0x9bc0c05b,
+	0x75b7b7c2, 0xe1fdfd1c, 0x3d9393ae, 0x4c26266a,
+	0x6c36365a, 0x7e3f3f41, 0xf5f7f702, 0x83cccc4f,
+	0x6834345c, 0x51a5a5f4, 0xd1e5e534, 0xf9f1f108,
+	0xe2717193, 0xabd8d873, 0x62313153, 0x2a15153f,
+	0x0804040c, 0x95c7c752, 0x46232365, 0x9dc3c35e,
+	0x30181828, 0x379696a1, 0x0a05050f, 0x2f9a9ab5,
+	0x0e070709, 0x24121236, 0x1b80809b, 0xdfe2e23d,
+	0xcdebeb26, 0x4e272769, 0x7fb2b2cd, 0xea75759f,
+	0x1209091b, 0x1d83839e, 0x582c2c74, 0x341a1a2e,
+	0x361b1b2d, 0xdc6e6eb2, 0xb45a5aee, 0x5ba0a0fb,
+	0xa45252f6, 0x763b3b4d, 0xb7d6d661, 0x7db3b3ce,
+	0x5229297b, 0xdde3e33e, 0x5e2f2f71, 0x13848497,
+	0xa65353f5, 0xb9d1d168, 0x00000000, 0xc1eded2c,
+	0x40202060, 0xe3fcfc1f, 0x79b1b1c8, 0xb65b5bed,
+	0xd46a6abe, 0x8dcbcb46, 0x67bebed9, 0x7239394b,
+	0x944a4ade, 0x984c4cd4, 0xb05858e8, 0x85cfcf4a,
+	0xbbd0d06b, 0xc5efef2a, 0x4faaaae5, 0xedfbfb16,
+	0x864343c5, 0x9a4d4dd7, 0x66333355, 0x11858594,
+	0x8a4545cf, 0xe9f9f910, 0x04020206, 0xfe7f7f81,
+	0xa05050f0, 0x783c3c44, 0x259f9fba, 0x4ba8a8e3,
+	0xa25151f3, 0x5da3a3fe, 0x804040c0, 0x058f8f8a,
+	0x3f9292ad, 0x219d9dbc, 0x70383848, 0xf1f5f504,
+	0x63bcbcdf, 0x77b6b6c1, 0xafdada75, 0x42212163,
+	0x20101030, 0xe5ffff1a, 0xfdf3f30e, 0xbfd2d26d,
+	0x81cdcd4c, 0x180c0c14, 0x26131335, 0xc3ecec2f,
+	0xbe5f5fe1, 0x359797a2, 0x884444cc, 0x2e171739,
+	0x93c4c457, 0x55a7a7f2, 0xfc7e7e82, 0x7a3d3d47,
+	0xc86464ac, 0xba5d5de7, 0x3219192b, 0xe6737395,
+	0xc06060a0, 0x19818198, 0x9e4f4fd1, 0xa3dcdc7f,
+	0x44222266, 0x542a2a7e, 0x3b9090ab, 0x0b888883,
+	0x8c4646ca, 0xc7eeee29, 0x6bb8b8d3, 0x2814143c,
+	0xa7dede79, 0xbc5e5ee2, 0x160b0b1d, 0xaddbdb76,
+	0xdbe0e03b, 0x64323256, 0x743a3a4e, 0x140a0a1e,
+	0x924949db, 0x0c06060a, 0x4824246c, 0xb85c5ce4,
+	0x9fc2c25d, 0xbdd3d36e, 0x43acacef, 0xc46262a6,
+	0x399191a8, 0x319595a4, 0xd3e4e437, 0xf279798b,
+	0xd5e7e732, 0x8bc8c843, 0x6e373759, 0xda6d6db7,
+	0x018d8d8c, 0xb1d5d564, 0x9c4e4ed2, 0x49a9a9e0,
+	0xd86c6cb4, 0xac5656fa, 0xf3f4f407, 0xcfeaea25,
+	0xca6565af, 0xf47a7a8e, 0x47aeaee9, 0x10080818,
+	0x6fbabad5, 0xf0787888, 0x4a25256f, 0x5c2e2e72,
+	0x381c1c24, 0x57a6a6f1, 0x73b4b4c7, 0x97c6c651,
+	0xcbe8e823, 0xa1dddd7c, 0xe874749c, 0x3e1f1f21,
+	0x964b4bdd, 0x61bdbddc, 0x0d8b8b86, 0x0f8a8a85,
+	0xe0707090, 0x7c3e3e42, 0x71b5b5c4, 0xcc6666aa,
+	0x904848d8, 0x06030305, 0xf7f6f601, 0x1c0e0e12,
+	0xc26161a3, 0x6a35355f, 0xae5757f9, 0x69b9b9d0,
+	0x17868691, 0x99c1c158, 0x3a1d1d27, 0x279e9eb9,
+	0xd9e1e138, 0xebf8f813, 0x2b9898b3, 0x22111133,
+	0xd26969bb, 0xa9d9d970, 0x078e8e89, 0x339494a7,
+	0x2d9b9bb6, 0x3c1e1e22, 0x15878792, 0xc9e9e920,
+	0x87cece49, 0xaa5555ff, 0x50282878, 0xa5dfdf7a,
+	0x038c8c8f, 0x59a1a1f8, 0x09898980, 0x1a0d0d17,
+	0x65bfbfda, 0xd7e6e631, 0x844242c6, 0xd06868b8,
+	0x824141c3, 0x299999b0, 0x5a2d2d77, 0x1e0f0f11,
+	0x7bb0b0cb, 0xa85454fc, 0x6dbbbbd6, 0x2c16163a);
+$code.=<<___;
+.byte	0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5
+.byte	0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76
+.byte	0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0
+.byte	0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0
+.byte	0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc
+.byte	0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15
+.byte	0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a
+.byte	0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75
+.byte	0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0
+.byte	0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84
+.byte	0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b
+.byte	0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf
+.byte	0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85
+.byte	0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8
+.byte	0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5
+.byte	0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2
+.byte	0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17
+.byte	0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73
+.byte	0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88
+.byte	0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb
+.byte	0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c
+.byte	0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79
+.byte	0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9
+.byte	0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08
+.byte	0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6
+.byte	0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a
+.byte	0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e
+.byte	0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e
+.byte	0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94
+.byte	0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf
+.byte	0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68
+.byte	0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16
+___
+&_data_word(
+	0x51f4a750, 0x7e416553, 0x1a17a4c3, 0x3a275e96,
+	0x3bab6bcb, 0x1f9d45f1, 0xacfa58ab, 0x4be30393,
+	0x2030fa55, 0xad766df6, 0x88cc7691, 0xf5024c25,
+	0x4fe5d7fc, 0xc52acbd7, 0x26354480, 0xb562a38f,
+	0xdeb15a49, 0x25ba1b67, 0x45ea0e98, 0x5dfec0e1,
+	0xc32f7502, 0x814cf012, 0x8d4697a3, 0x6bd3f9c6,
+	0x038f5fe7, 0x15929c95, 0xbf6d7aeb, 0x955259da,
+	0xd4be832d, 0x587421d3, 0x49e06929, 0x8ec9c844,
+	0x75c2896a, 0xf48e7978, 0x99583e6b, 0x27b971dd,
+	0xbee14fb6, 0xf088ad17, 0xc920ac66, 0x7dce3ab4,
+	0x63df4a18, 0xe51a3182, 0x97513360, 0x62537f45,
+	0xb16477e0, 0xbb6bae84, 0xfe81a01c, 0xf9082b94,
+	0x70486858, 0x8f45fd19, 0x94de6c87, 0x527bf8b7,
+	0xab73d323, 0x724b02e2, 0xe31f8f57, 0x6655ab2a,
+	0xb2eb2807, 0x2fb5c203, 0x86c57b9a, 0xd33708a5,
+	0x302887f2, 0x23bfa5b2, 0x02036aba, 0xed16825c,
+	0x8acf1c2b, 0xa779b492, 0xf307f2f0, 0x4e69e2a1,
+	0x65daf4cd, 0x0605bed5, 0xd134621f, 0xc4a6fe8a,
+	0x342e539d, 0xa2f355a0, 0x058ae132, 0xa4f6eb75,
+	0x0b83ec39, 0x4060efaa, 0x5e719f06, 0xbd6e1051,
+	0x3e218af9, 0x96dd063d, 0xdd3e05ae, 0x4de6bd46,
+	0x91548db5, 0x71c45d05, 0x0406d46f, 0x605015ff,
+	0x1998fb24, 0xd6bde997, 0x894043cc, 0x67d99e77,
+	0xb0e842bd, 0x07898b88, 0xe7195b38, 0x79c8eedb,
+	0xa17c0a47, 0x7c420fe9, 0xf8841ec9, 0x00000000,
+	0x09808683, 0x322bed48, 0x1e1170ac, 0x6c5a724e,
+	0xfd0efffb, 0x0f853856, 0x3daed51e, 0x362d3927,
+	0x0a0fd964, 0x685ca621, 0x9b5b54d1, 0x24362e3a,
+	0x0c0a67b1, 0x9357e70f, 0xb4ee96d2, 0x1b9b919e,
+	0x80c0c54f, 0x61dc20a2, 0x5a774b69, 0x1c121a16,
+	0xe293ba0a, 0xc0a02ae5, 0x3c22e043, 0x121b171d,
+	0x0e090d0b, 0xf28bc7ad, 0x2db6a8b9, 0x141ea9c8,
+	0x57f11985, 0xaf75074c, 0xee99ddbb, 0xa37f60fd,
+	0xf701269f, 0x5c72f5bc, 0x44663bc5, 0x5bfb7e34,
+	0x8b432976, 0xcb23c6dc, 0xb6edfc68, 0xb8e4f163,
+	0xd731dcca, 0x42638510, 0x13972240, 0x84c61120,
+	0x854a247d, 0xd2bb3df8, 0xaef93211, 0xc729a16d,
+	0x1d9e2f4b, 0xdcb230f3, 0x0d8652ec, 0x77c1e3d0,
+	0x2bb3166c, 0xa970b999, 0x119448fa, 0x47e96422,
+	0xa8fc8cc4, 0xa0f03f1a, 0x567d2cd8, 0x223390ef,
+	0x87494ec7, 0xd938d1c1, 0x8ccaa2fe, 0x98d40b36,
+	0xa6f581cf, 0xa57ade28, 0xdab78e26, 0x3fadbfa4,
+	0x2c3a9de4, 0x5078920d, 0x6a5fcc9b, 0x547e4662,
+	0xf68d13c2, 0x90d8b8e8, 0x2e39f75e, 0x82c3aff5,
+	0x9f5d80be, 0x69d0937c, 0x6fd52da9, 0xcf2512b3,
+	0xc8ac993b, 0x10187da7, 0xe89c636e, 0xdb3bbb7b,
+	0xcd267809, 0x6e5918f4, 0xec9ab701, 0x834f9aa8,
+	0xe6956e65, 0xaaffe67e, 0x21bccf08, 0xef15e8e6,
+	0xbae79bd9, 0x4a6f36ce, 0xea9f09d4, 0x29b07cd6,
+	0x31a4b2af, 0x2a3f2331, 0xc6a59430, 0x35a266c0,
+	0x744ebc37, 0xfc82caa6, 0xe090d0b0, 0x33a7d815,
+	0xf104984a, 0x41ecdaf7, 0x7fcd500e, 0x1791f62f,
+	0x764dd68d, 0x43efb04d, 0xccaa4d54, 0xe49604df,
+	0x9ed1b5e3, 0x4c6a881b, 0xc12c1fb8, 0x4665517f,
+	0x9d5eea04, 0x018c355d, 0xfa877473, 0xfb0b412e,
+	0xb3671d5a, 0x92dbd252, 0xe9105633, 0x6dd64713,
+	0x9ad7618c, 0x37a10c7a, 0x59f8148e, 0xeb133c89,
+	0xcea927ee, 0xb761c935, 0xe11ce5ed, 0x7a47b13c,
+	0x9cd2df59, 0x55f2733f, 0x1814ce79, 0x73c737bf,
+	0x53f7cdea, 0x5ffdaa5b, 0xdf3d6f14, 0x7844db86,
+	0xcaaff381, 0xb968c43e, 0x3824342c, 0xc2a3405f,
+	0x161dc372, 0xbce2250c, 0x283c498b, 0xff0d9541,
+	0x39a80171, 0x080cb3de, 0xd8b4e49c, 0x6456c190,
+	0x7bcb8461, 0xd532b670, 0x486c5c74, 0xd0b85742);
+$code.=<<___;
+.byte	0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38
+.byte	0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb
+.byte	0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87
+.byte	0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb
+.byte	0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d
+.byte	0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e
+.byte	0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2
+.byte	0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25
+.byte	0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16
+.byte	0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92
+.byte	0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda
+.byte	0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84
+.byte	0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a
+.byte	0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06
+.byte	0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02
+.byte	0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b
+.byte	0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea
+.byte	0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73
+.byte	0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85
+.byte	0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e
+.byte	0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89
+.byte	0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b
+.byte	0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20
+.byte	0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4
+.byte	0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31
+.byte	0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f
+.byte	0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d
+.byte	0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef
+.byte	0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0
+.byte	0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61
+.byte	0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26
+.byte	0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d
+
+
+.globl	.AES_encrypt
+.align	7
+.AES_encrypt:
+	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)
+	$PUSH	r15,`$FRAME-$SIZE_T*17`($sp)
+	$PUSH	r16,`$FRAME-$SIZE_T*16`($sp)
+	$PUSH	r17,`$FRAME-$SIZE_T*15`($sp)
+	$PUSH	r18,`$FRAME-$SIZE_T*14`($sp)
+	$PUSH	r19,`$FRAME-$SIZE_T*13`($sp)
+	$PUSH	r20,`$FRAME-$SIZE_T*12`($sp)
+	$PUSH	r21,`$FRAME-$SIZE_T*11`($sp)
+	$PUSH	r22,`$FRAME-$SIZE_T*10`($sp)
+	$PUSH	r23,`$FRAME-$SIZE_T*9`($sp)
+	$PUSH	r24,`$FRAME-$SIZE_T*8`($sp)
+	$PUSH	r25,`$FRAME-$SIZE_T*7`($sp)
+	$PUSH	r26,`$FRAME-$SIZE_T*6`($sp)
+	$PUSH	r27,`$FRAME-$SIZE_T*5`($sp)
+	$PUSH	r28,`$FRAME-$SIZE_T*4`($sp)
+	$PUSH	r29,`$FRAME-$SIZE_T*3`($sp)
+	$PUSH	r30,`$FRAME-$SIZE_T*2`($sp)
+	$PUSH	r31,`$FRAME-$SIZE_T*1`($sp)
+
+	lwz	$s0,0($inp)
+	lwz	$s1,4($inp)
+	lwz	$s2,8($inp)
+	lwz	$s3,12($inp)
+	bl	LAES_Te
+	bl	Lppc_AES_encrypt_compact
+	stw	$s0,0($out)
+	stw	$s1,4($out)
+	stw	$s2,8($out)
+	stw	$s3,12($out)
+
+	$POP	r0,`$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)
+	$POP	r15,`$FRAME-$SIZE_T*17`($sp)
+	$POP	r16,`$FRAME-$SIZE_T*16`($sp)
+	$POP	r17,`$FRAME-$SIZE_T*15`($sp)
+	$POP	r18,`$FRAME-$SIZE_T*14`($sp)
+	$POP	r19,`$FRAME-$SIZE_T*13`($sp)
+	$POP	r20,`$FRAME-$SIZE_T*12`($sp)
+	$POP	r21,`$FRAME-$SIZE_T*11`($sp)
+	$POP	r22,`$FRAME-$SIZE_T*10`($sp)
+	$POP	r23,`$FRAME-$SIZE_T*9`($sp)
+	$POP	r24,`$FRAME-$SIZE_T*8`($sp)
+	$POP	r25,`$FRAME-$SIZE_T*7`($sp)
+	$POP	r26,`$FRAME-$SIZE_T*6`($sp)
+	$POP	r27,`$FRAME-$SIZE_T*5`($sp)
+	$POP	r28,`$FRAME-$SIZE_T*4`($sp)
+	$POP	r29,`$FRAME-$SIZE_T*3`($sp)
+	$POP	r30,`$FRAME-$SIZE_T*2`($sp)
+	$POP	r31,`$FRAME-$SIZE_T*1`($sp)
+	mtlr	r0
+	addi	$sp,$sp,$FRAME
+	blr
+
+.align	5
+Lppc_AES_encrypt:
+	lwz	$acc00,240($key)
+	lwz	$t0,0($key)
+	lwz	$t1,4($key)
+	lwz	$t2,8($key)
+	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
+	xor	$s2,$s2,$t2
+	xor	$s3,$s3,$t3
+	mtctr	$acc00
+.align	4
+Lenc_loop:
+	rlwinm	$acc00,$s0,`32-24+3`,21,28
+	rlwinm	$acc01,$s1,`32-24+3`,21,28
+	rlwinm	$acc02,$s2,`32-24+3`,21,28
+	rlwinm	$acc03,$s3,`32-24+3`,21,28
+	lwz	$t0,0($key)
+	lwz	$t1,4($key)
+	rlwinm	$acc04,$s1,`32-16+3`,21,28
+	rlwinm	$acc05,$s2,`32-16+3`,21,28
+	lwz	$t2,8($key)
+	lwz	$t3,12($key)
+	rlwinm	$acc06,$s3,`32-16+3`,21,28
+	rlwinm	$acc07,$s0,`32-16+3`,21,28
+	lwzx	$acc00,$Tbl0,$acc00
+	lwzx	$acc01,$Tbl0,$acc01
+	rlwinm	$acc08,$s2,`32-8+3`,21,28
+	rlwinm	$acc09,$s3,`32-8+3`,21,28
+	lwzx	$acc02,$Tbl0,$acc02
+	lwzx	$acc03,$Tbl0,$acc03
+	rlwinm	$acc10,$s0,`32-8+3`,21,28
+	rlwinm	$acc11,$s1,`32-8+3`,21,28
+	lwzx	$acc04,$Tbl1,$acc04
+	lwzx	$acc05,$Tbl1,$acc05
+	rlwinm	$acc12,$s3,`0+3`,21,28
+	rlwinm	$acc13,$s0,`0+3`,21,28
+	lwzx	$acc06,$Tbl1,$acc06
+	lwzx	$acc07,$Tbl1,$acc07
+	rlwinm	$acc14,$s1,`0+3`,21,28
+	rlwinm	$acc15,$s2,`0+3`,21,28
+	lwzx	$acc08,$Tbl2,$acc08
+	lwzx	$acc09,$Tbl2,$acc09
+	xor	$t0,$t0,$acc00
+	xor	$t1,$t1,$acc01
+	lwzx	$acc10,$Tbl2,$acc10
+	lwzx	$acc11,$Tbl2,$acc11
+	xor	$t2,$t2,$acc02
+	xor	$t3,$t3,$acc03
+	lwzx	$acc12,$Tbl3,$acc12
+	lwzx	$acc13,$Tbl3,$acc13
+	xor	$t0,$t0,$acc04
+	xor	$t1,$t1,$acc05
+	lwzx	$acc14,$Tbl3,$acc14
+	lwzx	$acc15,$Tbl3,$acc15
+	xor	$t2,$t2,$acc06
+	xor	$t3,$t3,$acc07
+	xor	$t0,$t0,$acc08
+	xor	$t1,$t1,$acc09
+	xor	$t2,$t2,$acc10
+	xor	$t3,$t3,$acc11
+	xor	$s0,$t0,$acc12
+	xor	$s1,$t1,$acc13
+	xor	$s2,$t2,$acc14
+	xor	$s3,$t3,$acc15
+	addi	$key,$key,16
+	bdnz-	Lenc_loop
+
+	addi	$Tbl2,$Tbl0,2048
+	nop
+	lwz	$t0,0($key)
+	lwz	$t1,4($key)
+	rlwinm	$acc00,$s0,`32-24`,24,31
+	rlwinm	$acc01,$s1,`32-24`,24,31
+	lwz	$t2,8($key)
+	lwz	$t3,12($key)
+	rlwinm	$acc02,$s2,`32-24`,24,31
+	rlwinm	$acc03,$s3,`32-24`,24,31
+	lwz	$acc08,`2048+0`($Tbl0)	! prefetch Te4
+	lwz	$acc09,`2048+32`($Tbl0)
+	rlwinm	$acc04,$s1,`32-16`,24,31
+	rlwinm	$acc05,$s2,`32-16`,24,31
+	lwz	$acc10,`2048+64`($Tbl0)
+	lwz	$acc11,`2048+96`($Tbl0)
+	rlwinm	$acc06,$s3,`32-16`,24,31
+	rlwinm	$acc07,$s0,`32-16`,24,31
+	lwz	$acc12,`2048+128`($Tbl0)
+	lwz	$acc13,`2048+160`($Tbl0)
+	rlwinm	$acc08,$s2,`32-8`,24,31
+	rlwinm	$acc09,$s3,`32-8`,24,31
+	lwz	$acc14,`2048+192`($Tbl0)
+	lwz	$acc15,`2048+224`($Tbl0)
+	rlwinm	$acc10,$s0,`32-8`,24,31
+	rlwinm	$acc11,$s1,`32-8`,24,31
+	lbzx	$acc00,$Tbl2,$acc00
+	lbzx	$acc01,$Tbl2,$acc01
+	rlwinm	$acc12,$s3,`0`,24,31
+	rlwinm	$acc13,$s0,`0`,24,31
+	lbzx	$acc02,$Tbl2,$acc02
+	lbzx	$acc03,$Tbl2,$acc03
+	rlwinm	$acc14,$s1,`0`,24,31
+	rlwinm	$acc15,$s2,`0`,24,31
+	lbzx	$acc04,$Tbl2,$acc04
+	lbzx	$acc05,$Tbl2,$acc05
+	rlwinm	$s0,$acc00,24,0,7
+	rlwinm	$s1,$acc01,24,0,7
+	lbzx	$acc06,$Tbl2,$acc06
+	lbzx	$acc07,$Tbl2,$acc07
+	rlwinm	$s2,$acc02,24,0,7
+	rlwinm	$s3,$acc03,24,0,7
+	lbzx	$acc08,$Tbl2,$acc08
+	lbzx	$acc09,$Tbl2,$acc09
+	rlwimi	$s0,$acc04,16,8,15
+	rlwimi	$s1,$acc05,16,8,15
+	lbzx	$acc10,$Tbl2,$acc10
+	lbzx	$acc11,$Tbl2,$acc11
+	rlwimi	$s2,$acc06,16,8,15
+	rlwimi	$s3,$acc07,16,8,15
+	lbzx	$acc12,$Tbl2,$acc12
+	lbzx	$acc13,$Tbl2,$acc13
+	rlwimi	$s0,$acc08,8,16,23
+	rlwimi	$s1,$acc09,8,16,23
+	lbzx	$acc14,$Tbl2,$acc14
+	lbzx	$acc15,$Tbl2,$acc15
+	rlwimi	$s2,$acc10,8,16,23
+	rlwimi	$s3,$acc11,8,16,23
+	or	$s0,$s0,$acc12
+	or	$s1,$s1,$acc13
+	or	$s2,$s2,$acc14
+	or	$s3,$s3,$acc15
+	xor	$s0,$s0,$t0
+	xor	$s1,$s1,$t1
+	xor	$s2,$s2,$t2
+	xor	$s3,$s3,$t3
+	blr
+
+.align	4
+Lppc_AES_encrypt_compact:
+	lwz	$acc00,240($key)
+	lwz	$t0,0($key)
+	lwz	$t1,4($key)
+	lwz	$t2,8($key)
+	lwz	$t3,12($key)
+	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
+	xor	$s2,$s2,$t2
+	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
+	rlwinm	$acc04,$s1,`32-16`,24,31
+	rlwinm	$acc05,$s2,`32-16`,24,31
+	rlwinm	$acc06,$s3,`32-16`,24,31
+	rlwinm	$acc07,$s0,`32-16`,24,31
+	lbzx	$acc00,$Tbl1,$acc00
+	lbzx	$acc01,$Tbl1,$acc01
+	rlwinm	$acc08,$s2,`32-8`,24,31
+	rlwinm	$acc09,$s3,`32-8`,24,31
+	lbzx	$acc02,$Tbl1,$acc02
+	lbzx	$acc03,$Tbl1,$acc03
+	rlwinm	$acc10,$s0,`32-8`,24,31
+	rlwinm	$acc11,$s1,`32-8`,24,31
+	lbzx	$acc04,$Tbl1,$acc04
+	lbzx	$acc05,$Tbl1,$acc05
+	rlwinm	$acc12,$s3,`0`,24,31
+	rlwinm	$acc13,$s0,`0`,24,31
+	lbzx	$acc06,$Tbl1,$acc06
+	lbzx	$acc07,$Tbl1,$acc07
+	rlwinm	$acc14,$s1,`0`,24,31
+	rlwinm	$acc15,$s2,`0`,24,31
+	lbzx	$acc08,$Tbl1,$acc08
+	lbzx	$acc09,$Tbl1,$acc09
+	rlwinm	$s0,$acc00,24,0,7
+	rlwinm	$s1,$acc01,24,0,7
+	lbzx	$acc10,$Tbl1,$acc10
+	lbzx	$acc11,$Tbl1,$acc11
+	rlwinm	$s2,$acc02,24,0,7
+	rlwinm	$s3,$acc03,24,0,7
+	lbzx	$acc12,$Tbl1,$acc12
+	lbzx	$acc13,$Tbl1,$acc13
+	rlwimi	$s0,$acc04,16,8,15
+	rlwimi	$s1,$acc05,16,8,15
+	lbzx	$acc14,$Tbl1,$acc14
+	lbzx	$acc15,$Tbl1,$acc15
+	rlwimi	$s2,$acc06,16,8,15
+	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)
+	lwz	$t1,4($key)
+	or	$s0,$s0,$acc12
+	or	$s1,$s1,$acc13
+	lwz	$t2,8($key)
+	lwz	$t3,12($key)
+	or	$s2,$s2,$acc14
+	or	$s3,$s3,$acc15
+
+	addi	$key,$key,16
+	bdz	Lenc_compact_done
+
+	and	$acc00,$s0,$mask80	# r1=r0&0x80808080
+	and	$acc01,$s1,$mask80
+	and	$acc02,$s2,$mask80
+	and	$acc03,$s3,$mask80
+	srwi	$acc04,$acc00,7		# r1>>7
+	srwi	$acc05,$acc01,7
+	srwi	$acc06,$acc02,7
+	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
+	sub	$acc02,$acc02,$acc06
+	sub	$acc03,$acc03,$acc07
+	add	$acc08,$acc08,$acc08	# (r0&0x7f7f7f7f)<<1
+	add	$acc09,$acc09,$acc09
+	add	$acc10,$acc10,$acc10
+	add	$acc11,$acc11,$acc11
+	and	$acc00,$acc00,$mask1b	# (r1-(r1>>7))&0x1b1b1b1b
+	and	$acc01,$acc01,$mask1b
+	and	$acc02,$acc02,$mask1b
+	and	$acc03,$acc03,$mask1b
+	xor	$acc00,$acc00,$acc08	# r2
+	xor	$acc01,$acc01,$acc09
+	xor	$acc02,$acc02,$acc10
+	xor	$acc03,$acc03,$acc11
+
+	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
+	xor	$s2,$s2,$acc02
+	xor	$s3,$s3,$acc03
+	rotrwi	$s0,$s0,24		# ROTATE(r2^r0,24)
+	rotrwi	$s1,$s1,24
+	rotrwi	$s2,$s2,24
+	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)
+	rotlwi	$acc09,$acc13,8
+	rotlwi	$acc10,$acc14,8
+	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
+	xor	$s2,$s2,$acc10
+	xor	$s3,$s3,$acc11
+
+	b	Lenc_compact_loop
+.align	4
+Lenc_compact_done:
+	xor	$s0,$s0,$t0
+	xor	$s1,$s1,$t1
+	xor	$s2,$s2,$t2
+	xor	$s3,$s3,$t3
+	blr
+
+.globl	.AES_decrypt
+.align	7
+.AES_decrypt:
+	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)
+	$PUSH	r15,`$FRAME-$SIZE_T*17`($sp)
+	$PUSH	r16,`$FRAME-$SIZE_T*16`($sp)
+	$PUSH	r17,`$FRAME-$SIZE_T*15`($sp)
+	$PUSH	r18,`$FRAME-$SIZE_T*14`($sp)
+	$PUSH	r19,`$FRAME-$SIZE_T*13`($sp)
+	$PUSH	r20,`$FRAME-$SIZE_T*12`($sp)
+	$PUSH	r21,`$FRAME-$SIZE_T*11`($sp)
+	$PUSH	r22,`$FRAME-$SIZE_T*10`($sp)
+	$PUSH	r23,`$FRAME-$SIZE_T*9`($sp)
+	$PUSH	r24,`$FRAME-$SIZE_T*8`($sp)
+	$PUSH	r25,`$FRAME-$SIZE_T*7`($sp)
+	$PUSH	r26,`$FRAME-$SIZE_T*6`($sp)
+	$PUSH	r27,`$FRAME-$SIZE_T*5`($sp)
+	$PUSH	r28,`$FRAME-$SIZE_T*4`($sp)
+	$PUSH	r29,`$FRAME-$SIZE_T*3`($sp)
+	$PUSH	r30,`$FRAME-$SIZE_T*2`($sp)
+	$PUSH	r31,`$FRAME-$SIZE_T*1`($sp)
+
+	lwz	$s0,0($inp)
+	lwz	$s1,4($inp)
+	lwz	$s2,8($inp)
+	lwz	$s3,12($inp)
+	bl	LAES_Td
+	bl	Lppc_AES_decrypt_compact
+	stw	$s0,0($out)
+	stw	$s1,4($out)
+	stw	$s2,8($out)
+	stw	$s3,12($out)
+
+	$POP	r0,`$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)
+	$POP	r15,`$FRAME-$SIZE_T*17`($sp)
+	$POP	r16,`$FRAME-$SIZE_T*16`($sp)
+	$POP	r17,`$FRAME-$SIZE_T*15`($sp)
+	$POP	r18,`$FRAME-$SIZE_T*14`($sp)
+	$POP	r19,`$FRAME-$SIZE_T*13`($sp)
+	$POP	r20,`$FRAME-$SIZE_T*12`($sp)
+	$POP	r21,`$FRAME-$SIZE_T*11`($sp)
+	$POP	r22,`$FRAME-$SIZE_T*10`($sp)
+	$POP	r23,`$FRAME-$SIZE_T*9`($sp)
+	$POP	r24,`$FRAME-$SIZE_T*8`($sp)
+	$POP	r25,`$FRAME-$SIZE_T*7`($sp)
+	$POP	r26,`$FRAME-$SIZE_T*6`($sp)
+	$POP	r27,`$FRAME-$SIZE_T*5`($sp)
+	$POP	r28,`$FRAME-$SIZE_T*4`($sp)
+	$POP	r29,`$FRAME-$SIZE_T*3`($sp)
+	$POP	r30,`$FRAME-$SIZE_T*2`($sp)
+	$POP	r31,`$FRAME-$SIZE_T*1`($sp)
+	mtlr	r0
+	addi	$sp,$sp,$FRAME
+	blr
+
+.align	5
+Lppc_AES_decrypt:
+	lwz	$acc00,240($key)
+	lwz	$t0,0($key)
+	lwz	$t1,4($key)
+	lwz	$t2,8($key)
+	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
+	xor	$s2,$s2,$t2
+	xor	$s3,$s3,$t3
+	mtctr	$acc00
+.align	4
+Ldec_loop:
+	rlwinm	$acc00,$s0,`32-24+3`,21,28
+	rlwinm	$acc01,$s1,`32-24+3`,21,28
+	rlwinm	$acc02,$s2,`32-24+3`,21,28
+	rlwinm	$acc03,$s3,`32-24+3`,21,28
+	lwz	$t0,0($key)
+	lwz	$t1,4($key)
+	rlwinm	$acc04,$s3,`32-16+3`,21,28
+	rlwinm	$acc05,$s0,`32-16+3`,21,28
+	lwz	$t2,8($key)
+	lwz	$t3,12($key)
+	rlwinm	$acc06,$s1,`32-16+3`,21,28
+	rlwinm	$acc07,$s2,`32-16+3`,21,28
+	lwzx	$acc00,$Tbl0,$acc00
+	lwzx	$acc01,$Tbl0,$acc01
+	rlwinm	$acc08,$s2,`32-8+3`,21,28
+	rlwinm	$acc09,$s3,`32-8+3`,21,28
+	lwzx	$acc02,$Tbl0,$acc02
+	lwzx	$acc03,$Tbl0,$acc03
+	rlwinm	$acc10,$s0,`32-8+3`,21,28
+	rlwinm	$acc11,$s1,`32-8+3`,21,28
+	lwzx	$acc04,$Tbl1,$acc04
+	lwzx	$acc05,$Tbl1,$acc05
+	rlwinm	$acc12,$s1,`0+3`,21,28
+	rlwinm	$acc13,$s2,`0+3`,21,28
+	lwzx	$acc06,$Tbl1,$acc06
+	lwzx	$acc07,$Tbl1,$acc07
+	rlwinm	$acc14,$s3,`0+3`,21,28
+	rlwinm	$acc15,$s0,`0+3`,21,28
+	lwzx	$acc08,$Tbl2,$acc08
+	lwzx	$acc09,$Tbl2,$acc09
+	xor	$t0,$t0,$acc00
+	xor	$t1,$t1,$acc01
+	lwzx	$acc10,$Tbl2,$acc10
+	lwzx	$acc11,$Tbl2,$acc11
+	xor	$t2,$t2,$acc02
+	xor	$t3,$t3,$acc03
+	lwzx	$acc12,$Tbl3,$acc12
+	lwzx	$acc13,$Tbl3,$acc13
+	xor	$t0,$t0,$acc04
+	xor	$t1,$t1,$acc05
+	lwzx	$acc14,$Tbl3,$acc14
+	lwzx	$acc15,$Tbl3,$acc15
+	xor	$t2,$t2,$acc06
+	xor	$t3,$t3,$acc07
+	xor	$t0,$t0,$acc08
+	xor	$t1,$t1,$acc09
+	xor	$t2,$t2,$acc10
+	xor	$t3,$t3,$acc11
+	xor	$s0,$t0,$acc12
+	xor	$s1,$t1,$acc13
+	xor	$s2,$t2,$acc14
+	xor	$s3,$t3,$acc15
+	addi	$key,$key,16
+	bdnz-	Ldec_loop
+
+	addi	$Tbl2,$Tbl0,2048
+	nop
+	lwz	$t0,0($key)
+	lwz	$t1,4($key)
+	rlwinm	$acc00,$s0,`32-24`,24,31
+	rlwinm	$acc01,$s1,`32-24`,24,31
+	lwz	$t2,8($key)
+	lwz	$t3,12($key)
+	rlwinm	$acc02,$s2,`32-24`,24,31
+	rlwinm	$acc03,$s3,`32-24`,24,31
+	lwz	$acc08,`2048+0`($Tbl0)	! prefetch Td4
+	lwz	$acc09,`2048+32`($Tbl0)
+	rlwinm	$acc04,$s3,`32-16`,24,31
+	rlwinm	$acc05,$s0,`32-16`,24,31
+	lwz	$acc10,`2048+64`($Tbl0)
+	lwz	$acc11,`2048+96`($Tbl0)
+	lbzx	$acc00,$Tbl2,$acc00
+	lbzx	$acc01,$Tbl2,$acc01
+	lwz	$acc12,`2048+128`($Tbl0)
+	lwz	$acc13,`2048+160`($Tbl0)
+	rlwinm	$acc06,$s1,`32-16`,24,31
+	rlwinm	$acc07,$s2,`32-16`,24,31
+	lwz	$acc14,`2048+192`($Tbl0)
+	lwz	$acc15,`2048+224`($Tbl0)
+	rlwinm	$acc08,$s2,`32-8`,24,31
+	rlwinm	$acc09,$s3,`32-8`,24,31
+	lbzx	$acc02,$Tbl2,$acc02
+	lbzx	$acc03,$Tbl2,$acc03
+	rlwinm	$acc10,$s0,`32-8`,24,31
+	rlwinm	$acc11,$s1,`32-8`,24,31
+	lbzx	$acc04,$Tbl2,$acc04
+	lbzx	$acc05,$Tbl2,$acc05
+	rlwinm	$acc12,$s1,`0`,24,31
+	rlwinm	$acc13,$s2,`0`,24,31
+	lbzx	$acc06,$Tbl2,$acc06
+	lbzx	$acc07,$Tbl2,$acc07
+	rlwinm	$acc14,$s3,`0`,24,31
+	rlwinm	$acc15,$s0,`0`,24,31
+	lbzx	$acc08,$Tbl2,$acc08
+	lbzx	$acc09,$Tbl2,$acc09
+	rlwinm	$s0,$acc00,24,0,7
+	rlwinm	$s1,$acc01,24,0,7
+	lbzx	$acc10,$Tbl2,$acc10
+	lbzx	$acc11,$Tbl2,$acc11
+	rlwinm	$s2,$acc02,24,0,7
+	rlwinm	$s3,$acc03,24,0,7
+	lbzx	$acc12,$Tbl2,$acc12
+	lbzx	$acc13,$Tbl2,$acc13
+	rlwimi	$s0,$acc04,16,8,15
+	rlwimi	$s1,$acc05,16,8,15
+	lbzx	$acc14,$Tbl2,$acc14
+	lbzx	$acc15,$Tbl2,$acc15
+	rlwimi	$s2,$acc06,16,8,15
+	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
+	or	$s0,$s0,$acc12
+	or	$s1,$s1,$acc13
+	or	$s2,$s2,$acc14
+	or	$s3,$s3,$acc15
+	xor	$s0,$s0,$t0
+	xor	$s1,$s1,$t1
+	xor	$s2,$s2,$t2
+	xor	$s3,$s3,$t3
+	blr
+
+.align	4
+Lppc_AES_decrypt_compact:
+	lwz	$acc00,240($key)
+	lwz	$t0,0($key)
+	lwz	$t1,4($key)
+	lwz	$t2,8($key)
+	lwz	$t3,12($key)
+	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
+	insrdi	$mask1b,$mask1b,32,0
+___
+$code.=<<___;
+	mtctr	$acc00
+.align	4
+Ldec_compact_loop:
+	xor	$s0,$s0,$t0
+	xor	$s1,$s1,$t1
+	xor	$s2,$s2,$t2
+	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
+	rlwinm	$acc04,$s3,`32-16`,24,31
+	rlwinm	$acc05,$s0,`32-16`,24,31
+	rlwinm	$acc06,$s1,`32-16`,24,31
+	rlwinm	$acc07,$s2,`32-16`,24,31
+	lbzx	$acc00,$Tbl1,$acc00
+	lbzx	$acc01,$Tbl1,$acc01
+	rlwinm	$acc08,$s2,`32-8`,24,31
+	rlwinm	$acc09,$s3,`32-8`,24,31
+	lbzx	$acc02,$Tbl1,$acc02
+	lbzx	$acc03,$Tbl1,$acc03
+	rlwinm	$acc10,$s0,`32-8`,24,31
+	rlwinm	$acc11,$s1,`32-8`,24,31
+	lbzx	$acc04,$Tbl1,$acc04
+	lbzx	$acc05,$Tbl1,$acc05
+	rlwinm	$acc12,$s1,`0`,24,31
+	rlwinm	$acc13,$s2,`0`,24,31
+	lbzx	$acc06,$Tbl1,$acc06
+	lbzx	$acc07,$Tbl1,$acc07
+	rlwinm	$acc14,$s3,`0`,24,31
+	rlwinm	$acc15,$s0,`0`,24,31
+	lbzx	$acc08,$Tbl1,$acc08
+	lbzx	$acc09,$Tbl1,$acc09
+	rlwinm	$s0,$acc00,24,0,7
+	rlwinm	$s1,$acc01,24,0,7
+	lbzx	$acc10,$Tbl1,$acc10
+	lbzx	$acc11,$Tbl1,$acc11
+	rlwinm	$s2,$acc02,24,0,7
+	rlwinm	$s3,$acc03,24,0,7
+	lbzx	$acc12,$Tbl1,$acc12
+	lbzx	$acc13,$Tbl1,$acc13
+	rlwimi	$s0,$acc04,16,8,15
+	rlwimi	$s1,$acc05,16,8,15
+	lbzx	$acc14,$Tbl1,$acc14
+	lbzx	$acc15,$Tbl1,$acc15
+	rlwimi	$s2,$acc06,16,8,15
+	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)
+	lwz	$t1,4($key)
+	or	$s0,$s0,$acc12
+	or	$s1,$s1,$acc13
+	lwz	$t2,8($key)
+	lwz	$t3,12($key)
+	or	$s2,$s2,$acc14
+	or	$s3,$s3,$acc15
+
+	addi	$key,$key,16
+	bdz	Ldec_compact_done
+___
+$code.=<<___ if ($SIZE_T==8);
+	# vectorized permutation improves decrypt performance by 10%
+	insrdi	$s0,$s1,32,0
+	insrdi	$s2,$s3,32,0
+
+	and	$acc00,$s0,$mask80	# r1=r0&0x80808080
+	and	$acc02,$s2,$mask80
+	srdi	$acc04,$acc00,7		# r1>>7
+	srdi	$acc06,$acc02,7
+	andc	$acc08,$s0,$mask80	# r0&0x7f7f7f7f
+	andc	$acc10,$s2,$mask80
+	sub	$acc00,$acc00,$acc04	# r1-(r1>>7)
+	sub	$acc02,$acc02,$acc06
+	add	$acc08,$acc08,$acc08	# (r0&0x7f7f7f7f)<<1
+	add	$acc10,$acc10,$acc10
+	and	$acc00,$acc00,$mask1b	# (r1-(r1>>7))&0x1b1b1b1b
+	and	$acc02,$acc02,$mask1b
+	xor	$acc00,$acc00,$acc08	# r2
+	xor	$acc02,$acc02,$acc10
+
+	and	$acc04,$acc00,$mask80	# r1=r2&0x80808080
+	and	$acc06,$acc02,$mask80
+	srdi	$acc08,$acc04,7		# r1>>7
+	srdi	$acc10,$acc06,7
+	andc	$acc12,$acc00,$mask80	# r2&0x7f7f7f7f
+	andc	$acc14,$acc02,$mask80
+	sub	$acc04,$acc04,$acc08	# r1-(r1>>7)
+	sub	$acc06,$acc06,$acc10
+	add	$acc12,$acc12,$acc12	# (r2&0x7f7f7f7f)<<1
+	add	$acc14,$acc14,$acc14
+	and	$acc04,$acc04,$mask1b	# (r1-(r1>>7))&0x1b1b1b1b
+	and	$acc06,$acc06,$mask1b
+	xor	$acc04,$acc04,$acc12	# r4
+	xor	$acc06,$acc06,$acc14
+
+	and	$acc08,$acc04,$mask80	# r1=r4&0x80808080
+	and	$acc10,$acc06,$mask80
+	srdi	$acc12,$acc08,7		# r1>>7
+	srdi	$acc14,$acc10,7
+	sub	$acc08,$acc08,$acc12	# r1-(r1>>7)
+	sub	$acc10,$acc10,$acc14
+	andc	$acc12,$acc04,$mask80	# r4&0x7f7f7f7f
+	andc	$acc14,$acc06,$mask80
+	add	$acc12,$acc12,$acc12	# (r4&0x7f7f7f7f)<<1
+	add	$acc14,$acc14,$acc14
+	and	$acc08,$acc08,$mask1b	# (r1-(r1>>7))&0x1b1b1b1b
+	and	$acc10,$acc10,$mask1b
+	xor	$acc08,$acc08,$acc12	# r8
+	xor	$acc10,$acc10,$acc14
+
+	xor	$acc00,$acc00,$s0	# r2^r0
+	xor	$acc02,$acc02,$s2
+	xor	$acc04,$acc04,$s0	# r4^r0
+	xor	$acc06,$acc06,$s2
+
+	extrdi	$acc01,$acc00,32,0
+	extrdi	$acc03,$acc02,32,0
+	extrdi	$acc05,$acc04,32,0
+	extrdi	$acc07,$acc06,32,0
+	extrdi	$acc09,$acc08,32,0
+	extrdi	$acc11,$acc10,32,0
+___
+$code.=<<___ if ($SIZE_T==4);
+	and	$acc00,$s0,$mask80	# r1=r0&0x80808080
+	and	$acc01,$s1,$mask80
+	and	$acc02,$s2,$mask80
+	and	$acc03,$s3,$mask80
+	srwi	$acc04,$acc00,7		# r1>>7
+	srwi	$acc05,$acc01,7
+	srwi	$acc06,$acc02,7
+	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
+	sub	$acc02,$acc02,$acc06
+	sub	$acc03,$acc03,$acc07
+	add	$acc08,$acc08,$acc08	# (r0&0x7f7f7f7f)<<1
+	add	$acc09,$acc09,$acc09
+	add	$acc10,$acc10,$acc10
+	add	$acc11,$acc11,$acc11
+	and	$acc00,$acc00,$mask1b	# (r1-(r1>>7))&0x1b1b1b1b
+	and	$acc01,$acc01,$mask1b
+	and	$acc02,$acc02,$mask1b
+	and	$acc03,$acc03,$mask1b
+	xor	$acc00,$acc00,$acc08	# r2
+	xor	$acc01,$acc01,$acc09
+	xor	$acc02,$acc02,$acc10
+	xor	$acc03,$acc03,$acc11
+
+	and	$acc04,$acc00,$mask80	# r1=r2&0x80808080
+	and	$acc05,$acc01,$mask80
+	and	$acc06,$acc02,$mask80
+	and	$acc07,$acc03,$mask80
+	srwi	$acc08,$acc04,7		# r1>>7
+	srwi	$acc09,$acc05,7
+	srwi	$acc10,$acc06,7
+	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
+	sub	$acc06,$acc06,$acc10
+	sub	$acc07,$acc07,$acc11
+	add	$acc12,$acc12,$acc12	# (r2&0x7f7f7f7f)<<1
+	add	$acc13,$acc13,$acc13
+	add	$acc14,$acc14,$acc14
+	add	$acc15,$acc15,$acc15
+	and	$acc04,$acc04,$mask1b	# (r1-(r1>>7))&0x1b1b1b1b
+	and	$acc05,$acc05,$mask1b
+	and	$acc06,$acc06,$mask1b
+	and	$acc07,$acc07,$mask1b
+	xor	$acc04,$acc04,$acc12	# r4
+	xor	$acc05,$acc05,$acc13
+	xor	$acc06,$acc06,$acc14
+	xor	$acc07,$acc07,$acc15
+
+	and	$acc08,$acc04,$mask80	# r1=r4&0x80808080
+	and	$acc09,$acc05,$mask80
+	and	$acc10,$acc06,$mask80
+	and	$acc11,$acc07,$mask80
+	srwi	$acc12,$acc08,7		# r1>>7
+	srwi	$acc13,$acc09,7
+	srwi	$acc14,$acc10,7
+	srwi	$acc15,$acc11,7
+	sub	$acc08,$acc08,$acc12	# r1-(r1>>7)
+	sub	$acc09,$acc09,$acc13
+	sub	$acc10,$acc10,$acc14
+	sub	$acc11,$acc11,$acc15
+	andc	$acc12,$acc04,$mask80	# r4&0x7f7f7f7f
+	andc	$acc13,$acc05,$mask80
+	andc	$acc14,$acc06,$mask80
+	andc	$acc15,$acc07,$mask80
+	add	$acc12,$acc12,$acc12	# (r4&0x7f7f7f7f)<<1
+	add	$acc13,$acc13,$acc13
+	add	$acc14,$acc14,$acc14
+	add	$acc15,$acc15,$acc15
+	and	$acc08,$acc08,$mask1b	# (r1-(r1>>7))&0x1b1b1b1b
+	and	$acc09,$acc09,$mask1b
+	and	$acc10,$acc10,$mask1b
+	and	$acc11,$acc11,$mask1b
+	xor	$acc08,$acc08,$acc12	# r8
+	xor	$acc09,$acc09,$acc13
+	xor	$acc10,$acc10,$acc14
+	xor	$acc11,$acc11,$acc15
+
+	xor	$acc00,$acc00,$s0	# r2^r0
+	xor	$acc01,$acc01,$s1
+	xor	$acc02,$acc02,$s2
+	xor	$acc03,$acc03,$s3
+	xor	$acc04,$acc04,$s0	# r4^r0
+	xor	$acc05,$acc05,$s1
+	xor	$acc06,$acc06,$s2
+	xor	$acc07,$acc07,$s3
+___
+$code.=<<___;
+	rotrwi	$s0,$s0,8		# = ROTATE(r0,8)
+	rotrwi	$s1,$s1,8
+	rotrwi	$s2,$s2,8
+	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
+	xor	$acc01,$acc01,$acc09
+	xor	$acc02,$acc02,$acc10
+	xor	$acc03,$acc03,$acc11
+	xor	$s0,$s0,$acc04		# ^= r4^r0
+	xor	$s1,$s1,$acc05
+	xor	$s2,$s2,$acc06
+	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)]
+	xor	$s1,$s1,$acc09
+	xor	$s2,$s2,$acc10
+	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)
+	xor	$s1,$s1,$acc01
+	xor	$s2,$s2,$acc02
+	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
+	xor	$s2,$s2,$acc06
+	xor	$s3,$s3,$acc07
+	xor	$s0,$s0,$acc08		# ^= ROTATE(r8,8)	
+	xor	$s1,$s1,$acc09	
+	xor	$s2,$s2,$acc10	
+	xor	$s3,$s3,$acc11	
+
+	b	Ldec_compact_loop
+.align	4
+Ldec_compact_done:
+	xor	$s0,$s0,$t0
+	xor	$s1,$s1,$t1
+	xor	$s2,$s2,$t2
+	xor	$s3,$s3,$t3
+	blr
+.long	0
+.asciz	"AES for PPC, CRYPTOGAMS by <appro\@openssl.org>"
+.align	7
+___
+
+$code =~ s/\`([^\`]*)\`/eval $1/gem;
+print $code;
+close STDOUT;
diff --git a/openssl/crypto/aes/asm/aes-s390x.pl b/openssl/crypto/aes/asm/aes-s390x.pl
new file mode 100644
index 00000000..7e018892
--- /dev/null
+++ b/openssl/crypto/aes/asm/aes-s390x.pl
@@ -0,0 +1,1339 @@
+#!/usr/bin/env perl
+
+# ====================================================================
+# Written by Andy Polyakov <appro@fy.chalmers.se> 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/.
+# ====================================================================
+
+# AES for s390x.
+
+# April 2007.
+#
+# Software performance improvement over gcc-generated code is ~70% and
+# in absolute terms is ~73 cycles per byte processed with 128-bit key.
+# You're likely to exclaim "why so slow?" Keep in mind that z-CPUs are
+# *strictly* in-order execution and issued instruction [in this case
+# load value from memory is critical] has to complete before execution
+# flow proceeds. S-boxes are compressed to 2KB[+256B].
+#
+# As for hardware acceleration support. It's basically a "teaser," as
+# it can and should be improved in several ways. Most notably support
+# for CBC is not utilized, nor multiple blocks are ever processed.
+# Then software key schedule can be postponed till hardware support
+# detection... Performance improvement over assembler is reportedly
+# ~2.5x, but can reach >8x [naturally on larger chunks] if proper
+# support is implemented.
+
+# May 2007.
+#
+# Implement AES_set_[en|de]crypt_key. Key schedule setup is avoided
+# for 128-bit keys, if hardware support is detected.
+
+# Januray 2009.
+#
+# Add support for hardware AES192/256 and reschedule instructions to
+# minimize/avoid Address Generation Interlock hazard and to favour
+# dual-issue z10 pipeline. This gave ~25% improvement on z10 and
+# almost 50% on z9. The gain is smaller on z10, because being dual-
+# issue z10 makes it improssible to eliminate the interlock condition:
+# critial path is not long enough. Yet it spends ~24 cycles per byte
+# processed with 128-bit key.
+#
+# Unlike previous version hardware support detection takes place only
+# at the moment of key schedule setup, which is denoted in key->rounds.
+# This is done, because deferred key setup can't be made MT-safe, not
+# for key lengthes longer than 128 bits.
+#
+# Add AES_cbc_encrypt, which gives incredible performance improvement,
+# it was measured to be ~6.6x. It's less than previously mentioned 8x,
+# because software implementation was optimized.
+
+$softonly=0;	# allow hardware support
+
+$t0="%r0";	$mask="%r0";
+$t1="%r1";
+$t2="%r2";	$inp="%r2";
+$t3="%r3";	$out="%r3";	$bits="%r3";
+$key="%r4";
+$i1="%r5";
+$i2="%r6";
+$i3="%r7";
+$s0="%r8";
+$s1="%r9";
+$s2="%r10";
+$s3="%r11";
+$tbl="%r12";
+$rounds="%r13";
+$ra="%r14";
+$sp="%r15";
+
+sub _data_word()
+{ my $i;
+    while(defined($i=shift)) { $code.=sprintf".long\t0x%08x,0x%08x\n",$i,$i; }
+}
+
+$code=<<___;
+.text
+
+.type	AES_Te,\@object
+.align	256
+AES_Te:
+___
+&_data_word(
+	0xc66363a5, 0xf87c7c84, 0xee777799, 0xf67b7b8d,
+	0xfff2f20d, 0xd66b6bbd, 0xde6f6fb1, 0x91c5c554,
+	0x60303050, 0x02010103, 0xce6767a9, 0x562b2b7d,
+	0xe7fefe19, 0xb5d7d762, 0x4dababe6, 0xec76769a,
+	0x8fcaca45, 0x1f82829d, 0x89c9c940, 0xfa7d7d87,
+	0xeffafa15, 0xb25959eb, 0x8e4747c9, 0xfbf0f00b,
+	0x41adadec, 0xb3d4d467, 0x5fa2a2fd, 0x45afafea,
+	0x239c9cbf, 0x53a4a4f7, 0xe4727296, 0x9bc0c05b,
+	0x75b7b7c2, 0xe1fdfd1c, 0x3d9393ae, 0x4c26266a,
+	0x6c36365a, 0x7e3f3f41, 0xf5f7f702, 0x83cccc4f,
+	0x6834345c, 0x51a5a5f4, 0xd1e5e534, 0xf9f1f108,
+	0xe2717193, 0xabd8d873, 0x62313153, 0x2a15153f,
+	0x0804040c, 0x95c7c752, 0x46232365, 0x9dc3c35e,
+	0x30181828, 0x379696a1, 0x0a05050f, 0x2f9a9ab5,
+	0x0e070709, 0x24121236, 0x1b80809b, 0xdfe2e23d,
+	0xcdebeb26, 0x4e272769, 0x7fb2b2cd, 0xea75759f,
+	0x1209091b, 0x1d83839e, 0x582c2c74, 0x341a1a2e,
+	0x361b1b2d, 0xdc6e6eb2, 0xb45a5aee, 0x5ba0a0fb,
+	0xa45252f6, 0x763b3b4d, 0xb7d6d661, 0x7db3b3ce,
+	0x5229297b, 0xdde3e33e, 0x5e2f2f71, 0x13848497,
+	0xa65353f5, 0xb9d1d168, 0x00000000, 0xc1eded2c,
+	0x40202060, 0xe3fcfc1f, 0x79b1b1c8, 0xb65b5bed,
+	0xd46a6abe, 0x8dcbcb46, 0x67bebed9, 0x7239394b,
+	0x944a4ade, 0x984c4cd4, 0xb05858e8, 0x85cfcf4a,
+	0xbbd0d06b, 0xc5efef2a, 0x4faaaae5, 0xedfbfb16,
+	0x864343c5, 0x9a4d4dd7, 0x66333355, 0x11858594,
+	0x8a4545cf, 0xe9f9f910, 0x04020206, 0xfe7f7f81,
+	0xa05050f0, 0x783c3c44, 0x259f9fba, 0x4ba8a8e3,
+	0xa25151f3, 0x5da3a3fe, 0x804040c0, 0x058f8f8a,
+	0x3f9292ad, 0x219d9dbc, 0x70383848, 0xf1f5f504,
+	0x63bcbcdf, 0x77b6b6c1, 0xafdada75, 0x42212163,
+	0x20101030, 0xe5ffff1a, 0xfdf3f30e, 0xbfd2d26d,
+	0x81cdcd4c, 0x180c0c14, 0x26131335, 0xc3ecec2f,
+	0xbe5f5fe1, 0x359797a2, 0x884444cc, 0x2e171739,
+	0x93c4c457, 0x55a7a7f2, 0xfc7e7e82, 0x7a3d3d47,
+	0xc86464ac, 0xba5d5de7, 0x3219192b, 0xe6737395,
+	0xc06060a0, 0x19818198, 0x9e4f4fd1, 0xa3dcdc7f,
+	0x44222266, 0x542a2a7e, 0x3b9090ab, 0x0b888883,
+	0x8c4646ca, 0xc7eeee29, 0x6bb8b8d3, 0x2814143c,
+	0xa7dede79, 0xbc5e5ee2, 0x160b0b1d, 0xaddbdb76,
+	0xdbe0e03b, 0x64323256, 0x743a3a4e, 0x140a0a1e,
+	0x924949db, 0x0c06060a, 0x4824246c, 0xb85c5ce4,
+	0x9fc2c25d, 0xbdd3d36e, 0x43acacef, 0xc46262a6,
+	0x399191a8, 0x319595a4, 0xd3e4e437, 0xf279798b,
+	0xd5e7e732, 0x8bc8c843, 0x6e373759, 0xda6d6db7,
+	0x018d8d8c, 0xb1d5d564, 0x9c4e4ed2, 0x49a9a9e0,
+	0xd86c6cb4, 0xac5656fa, 0xf3f4f407, 0xcfeaea25,
+	0xca6565af, 0xf47a7a8e, 0x47aeaee9, 0x10080818,
+	0x6fbabad5, 0xf0787888, 0x4a25256f, 0x5c2e2e72,
+	0x381c1c24, 0x57a6a6f1, 0x73b4b4c7, 0x97c6c651,
+	0xcbe8e823, 0xa1dddd7c, 0xe874749c, 0x3e1f1f21,
+	0x964b4bdd, 0x61bdbddc, 0x0d8b8b86, 0x0f8a8a85,
+	0xe0707090, 0x7c3e3e42, 0x71b5b5c4, 0xcc6666aa,
+	0x904848d8, 0x06030305, 0xf7f6f601, 0x1c0e0e12,
+	0xc26161a3, 0x6a35355f, 0xae5757f9, 0x69b9b9d0,
+	0x17868691, 0x99c1c158, 0x3a1d1d27, 0x279e9eb9,
+	0xd9e1e138, 0xebf8f813, 0x2b9898b3, 0x22111133,
+	0xd26969bb, 0xa9d9d970, 0x078e8e89, 0x339494a7,
+	0x2d9b9bb6, 0x3c1e1e22, 0x15878792, 0xc9e9e920,
+	0x87cece49, 0xaa5555ff, 0x50282878, 0xa5dfdf7a,
+	0x038c8c8f, 0x59a1a1f8, 0x09898980, 0x1a0d0d17,
+	0x65bfbfda, 0xd7e6e631, 0x844242c6, 0xd06868b8,
+	0x824141c3, 0x299999b0, 0x5a2d2d77, 0x1e0f0f11,
+	0x7bb0b0cb, 0xa85454fc, 0x6dbbbbd6, 0x2c16163a);
+$code.=<<___;
+# Te4[256]
+.byte	0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5
+.byte	0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76
+.byte	0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0
+.byte	0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0
+.byte	0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc
+.byte	0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15
+.byte	0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a
+.byte	0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75
+.byte	0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0
+.byte	0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84
+.byte	0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b
+.byte	0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf
+.byte	0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85
+.byte	0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8
+.byte	0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5
+.byte	0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2
+.byte	0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17
+.byte	0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73
+.byte	0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88
+.byte	0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb
+.byte	0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c
+.byte	0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79
+.byte	0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9
+.byte	0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08
+.byte	0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6
+.byte	0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a
+.byte	0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e
+.byte	0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e
+.byte	0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94
+.byte	0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf
+.byte	0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68
+.byte	0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16
+# rcon[]
+.long	0x01000000, 0x02000000, 0x04000000, 0x08000000
+.long	0x10000000, 0x20000000, 0x40000000, 0x80000000
+.long	0x1B000000, 0x36000000, 0, 0, 0, 0, 0, 0
+.align	256
+.size	AES_Te,.-AES_Te
+
+# void AES_encrypt(const unsigned char *inp, unsigned char *out,
+# 		 const AES_KEY *key) {
+.globl	AES_encrypt
+.type	AES_encrypt,\@function
+AES_encrypt:
+___
+$code.=<<___ if (!$softonly);
+	l	%r0,240($key)
+	lhi	%r1,16
+	clr	%r0,%r1
+	jl	.Lesoft
+
+	la	%r1,0($key)
+	#la	%r2,0($inp)
+	la	%r4,0($out)
+	lghi	%r3,16		# single block length
+	.long	0xb92e0042	# km %r4,%r2
+	brc	1,.-4		# can this happen?
+	br	%r14
+.align	64
+.Lesoft:
+___
+$code.=<<___;
+	stmg	%r3,$ra,24($sp)
+
+	llgf	$s0,0($inp)
+	llgf	$s1,4($inp)
+	llgf	$s2,8($inp)
+	llgf	$s3,12($inp)
+
+	larl	$tbl,AES_Te
+	bras	$ra,_s390x_AES_encrypt
+
+	lg	$out,24($sp)
+	st	$s0,0($out)
+	st	$s1,4($out)
+	st	$s2,8($out)
+	st	$s3,12($out)
+
+	lmg	%r6,$ra,48($sp)
+	br	$ra
+.size	AES_encrypt,.-AES_encrypt
+
+.type   _s390x_AES_encrypt,\@function
+.align	16
+_s390x_AES_encrypt:
+	stg	$ra,152($sp)
+	x	$s0,0($key)
+	x	$s1,4($key)
+	x	$s2,8($key)
+	x	$s3,12($key)
+	l	$rounds,240($key)
+	llill	$mask,`0xff<<3`
+	aghi	$rounds,-1
+	j	.Lenc_loop
+.align	16
+.Lenc_loop:
+	sllg	$t1,$s0,`0+3`
+	srlg	$t2,$s0,`8-3`
+	srlg	$t3,$s0,`16-3`
+	srl	$s0,`24-3`
+	nr	$s0,$mask
+	ngr	$t1,$mask
+	nr	$t2,$mask
+	nr	$t3,$mask
+
+	srlg	$i1,$s1,`16-3`	# i0
+	sllg	$i2,$s1,`0+3`
+	srlg	$i3,$s1,`8-3`
+	srl	$s1,`24-3`
+	nr	$i1,$mask
+	nr	$s1,$mask
+	ngr	$i2,$mask
+	nr	$i3,$mask
+
+	l	$s0,0($s0,$tbl)	# Te0[s0>>24]
+	l	$t1,1($t1,$tbl)	# Te3[s0>>0]
+	l	$t2,2($t2,$tbl) # Te2[s0>>8]
+	l	$t3,3($t3,$tbl)	# Te1[s0>>16]
+
+	x	$s0,3($i1,$tbl)	# Te1[s1>>16]
+	l	$s1,0($s1,$tbl)	# Te0[s1>>24]
+	x	$t2,1($i2,$tbl)	# Te3[s1>>0]
+	x	$t3,2($i3,$tbl)	# Te2[s1>>8]
+
+	srlg	$i1,$s2,`8-3`	# i0
+	srlg	$i2,$s2,`16-3`	# i1
+	nr	$i1,$mask
+	nr	$i2,$mask
+	sllg	$i3,$s2,`0+3`
+	srl	$s2,`24-3`
+	nr	$s2,$mask
+	ngr	$i3,$mask
+
+	xr	$s1,$t1
+	srlg	$ra,$s3,`8-3`	# i1
+	sllg	$t1,$s3,`0+3`	# i0
+	nr	$ra,$mask
+	la	$key,16($key)
+	ngr	$t1,$mask
+
+	x	$s0,2($i1,$tbl)	# Te2[s2>>8]
+	x	$s1,3($i2,$tbl)	# Te1[s2>>16]
+	l	$s2,0($s2,$tbl)	# Te0[s2>>24]
+	x	$t3,1($i3,$tbl)	# Te3[s2>>0]
+
+	srlg	$i3,$s3,`16-3`	# i2
+	xr	$s2,$t2
+	srl	$s3,`24-3`
+	nr	$i3,$mask
+	nr	$s3,$mask
+
+	x	$s0,0($key)
+	x	$s1,4($key)
+	x	$s2,8($key)
+	x	$t3,12($key)
+
+	x	$s0,1($t1,$tbl)	# Te3[s3>>0]
+	x	$s1,2($ra,$tbl)	# Te2[s3>>8]
+	x	$s2,3($i3,$tbl)	# Te1[s3>>16]
+	l	$s3,0($s3,$tbl)	# Te0[s3>>24]
+	xr	$s3,$t3
+
+	brct	$rounds,.Lenc_loop
+	.align	16
+
+	sllg	$t1,$s0,`0+3`
+	srlg	$t2,$s0,`8-3`
+	ngr	$t1,$mask
+	srlg	$t3,$s0,`16-3`
+	srl	$s0,`24-3`
+	nr	$s0,$mask
+	nr	$t2,$mask
+	nr	$t3,$mask
+
+	srlg	$i1,$s1,`16-3`	# i0
+	sllg	$i2,$s1,`0+3`
+	ngr	$i2,$mask
+	srlg	$i3,$s1,`8-3`
+	srl	$s1,`24-3`
+	nr	$i1,$mask
+	nr	$s1,$mask
+	nr	$i3,$mask
+
+	llgc	$s0,2($s0,$tbl)	# Te4[s0>>24]
+	llgc	$t1,2($t1,$tbl)	# Te4[s0>>0]
+	sll	$s0,24
+	llgc	$t2,2($t2,$tbl)	# Te4[s0>>8]
+	llgc	$t3,2($t3,$tbl)	# Te4[s0>>16]
+	sll	$t2,8
+	sll	$t3,16
+
+	llgc	$i1,2($i1,$tbl)	# Te4[s1>>16]
+	llgc	$s1,2($s1,$tbl)	# Te4[s1>>24]
+	llgc	$i2,2($i2,$tbl)	# Te4[s1>>0]
+	llgc	$i3,2($i3,$tbl)	# Te4[s1>>8]
+	sll	$i1,16
+	sll	$s1,24
+	sll	$i3,8
+	or	$s0,$i1
+	or	$s1,$t1
+	or	$t2,$i2
+	or	$t3,$i3
+	
+	srlg	$i1,$s2,`8-3`	# i0
+	srlg	$i2,$s2,`16-3`	# i1
+	nr	$i1,$mask
+	nr	$i2,$mask
+	sllg	$i3,$s2,`0+3`
+	srl	$s2,`24-3`
+	ngr	$i3,$mask
+	nr	$s2,$mask
+
+	sllg	$t1,$s3,`0+3`	# i0
+	srlg	$ra,$s3,`8-3`	# i1
+	ngr	$t1,$mask
+
+	llgc	$i1,2($i1,$tbl)	# Te4[s2>>8]
+	llgc	$i2,2($i2,$tbl)	# Te4[s2>>16]
+	sll	$i1,8
+	llgc	$s2,2($s2,$tbl)	# Te4[s2>>24]
+	llgc	$i3,2($i3,$tbl)	# Te4[s2>>0]
+	sll	$i2,16
+	nr	$ra,$mask
+	sll	$s2,24
+	or	$s0,$i1
+	or	$s1,$i2
+	or	$s2,$t2
+	or	$t3,$i3
+
+	srlg	$i3,$s3,`16-3`	# i2
+	srl	$s3,`24-3`
+	nr	$i3,$mask
+	nr	$s3,$mask
+
+	l	$t0,16($key)
+	l	$t2,20($key)
+
+	llgc	$i1,2($t1,$tbl)	# Te4[s3>>0]
+	llgc	$i2,2($ra,$tbl)	# Te4[s3>>8]
+	llgc	$i3,2($i3,$tbl)	# Te4[s3>>16]
+	llgc	$s3,2($s3,$tbl)	# Te4[s3>>24]
+	sll	$i2,8
+	sll	$i3,16
+	sll	$s3,24
+	or	$s0,$i1
+	or	$s1,$i2
+	or	$s2,$i3
+	or	$s3,$t3
+
+	lg	$ra,152($sp)
+	xr	$s0,$t0
+	xr	$s1,$t2
+	x	$s2,24($key)
+	x	$s3,28($key)
+
+	br	$ra	
+.size	_s390x_AES_encrypt,.-_s390x_AES_encrypt
+___
+
+$code.=<<___;
+.type	AES_Td,\@object
+.align	256
+AES_Td:
+___
+&_data_word(
+	0x51f4a750, 0x7e416553, 0x1a17a4c3, 0x3a275e96,
+	0x3bab6bcb, 0x1f9d45f1, 0xacfa58ab, 0x4be30393,
+	0x2030fa55, 0xad766df6, 0x88cc7691, 0xf5024c25,
+	0x4fe5d7fc, 0xc52acbd7, 0x26354480, 0xb562a38f,
+	0xdeb15a49, 0x25ba1b67, 0x45ea0e98, 0x5dfec0e1,
+	0xc32f7502, 0x814cf012, 0x8d4697a3, 0x6bd3f9c6,
+	0x038f5fe7, 0x15929c95, 0xbf6d7aeb, 0x955259da,
+	0xd4be832d, 0x587421d3, 0x49e06929, 0x8ec9c844,
+	0x75c2896a, 0xf48e7978, 0x99583e6b, 0x27b971dd,
+	0xbee14fb6, 0xf088ad17, 0xc920ac66, 0x7dce3ab4,
+	0x63df4a18, 0xe51a3182, 0x97513360, 0x62537f45,
+	0xb16477e0, 0xbb6bae84, 0xfe81a01c, 0xf9082b94,
+	0x70486858, 0x8f45fd19, 0x94de6c87, 0x527bf8b7,
+	0xab73d323, 0x724b02e2, 0xe31f8f57, 0x6655ab2a,
+	0xb2eb2807, 0x2fb5c203, 0x86c57b9a, 0xd33708a5,
+	0x302887f2, 0x23bfa5b2, 0x02036aba, 0xed16825c,
+	0x8acf1c2b, 0xa779b492, 0xf307f2f0, 0x4e69e2a1,
+	0x65daf4cd, 0x0605bed5, 0xd134621f, 0xc4a6fe8a,
+	0x342e539d, 0xa2f355a0, 0x058ae132, 0xa4f6eb75,
+	0x0b83ec39, 0x4060efaa, 0x5e719f06, 0xbd6e1051,
+	0x3e218af9, 0x96dd063d, 0xdd3e05ae, 0x4de6bd46,
+	0x91548db5, 0x71c45d05, 0x0406d46f, 0x605015ff,
+	0x1998fb24, 0xd6bde997, 0x894043cc, 0x67d99e77,
+	0xb0e842bd, 0x07898b88, 0xe7195b38, 0x79c8eedb,
+	0xa17c0a47, 0x7c420fe9, 0xf8841ec9, 0x00000000,
+	0x09808683, 0x322bed48, 0x1e1170ac, 0x6c5a724e,
+	0xfd0efffb, 0x0f853856, 0x3daed51e, 0x362d3927,
+	0x0a0fd964, 0x685ca621, 0x9b5b54d1, 0x24362e3a,
+	0x0c0a67b1, 0x9357e70f, 0xb4ee96d2, 0x1b9b919e,
+	0x80c0c54f, 0x61dc20a2, 0x5a774b69, 0x1c121a16,
+	0xe293ba0a, 0xc0a02ae5, 0x3c22e043, 0x121b171d,
+	0x0e090d0b, 0xf28bc7ad, 0x2db6a8b9, 0x141ea9c8,
+	0x57f11985, 0xaf75074c, 0xee99ddbb, 0xa37f60fd,
+	0xf701269f, 0x5c72f5bc, 0x44663bc5, 0x5bfb7e34,
+	0x8b432976, 0xcb23c6dc, 0xb6edfc68, 0xb8e4f163,
+	0xd731dcca, 0x42638510, 0x13972240, 0x84c61120,
+	0x854a247d, 0xd2bb3df8, 0xaef93211, 0xc729a16d,
+	0x1d9e2f4b, 0xdcb230f3, 0x0d8652ec, 0x77c1e3d0,
+	0x2bb3166c, 0xa970b999, 0x119448fa, 0x47e96422,
+	0xa8fc8cc4, 0xa0f03f1a, 0x567d2cd8, 0x223390ef,
+	0x87494ec7, 0xd938d1c1, 0x8ccaa2fe, 0x98d40b36,
+	0xa6f581cf, 0xa57ade28, 0xdab78e26, 0x3fadbfa4,
+	0x2c3a9de4, 0x5078920d, 0x6a5fcc9b, 0x547e4662,
+	0xf68d13c2, 0x90d8b8e8, 0x2e39f75e, 0x82c3aff5,
+	0x9f5d80be, 0x69d0937c, 0x6fd52da9, 0xcf2512b3,
+	0xc8ac993b, 0x10187da7, 0xe89c636e, 0xdb3bbb7b,
+	0xcd267809, 0x6e5918f4, 0xec9ab701, 0x834f9aa8,
+	0xe6956e65, 0xaaffe67e, 0x21bccf08, 0xef15e8e6,
+	0xbae79bd9, 0x4a6f36ce, 0xea9f09d4, 0x29b07cd6,
+	0x31a4b2af, 0x2a3f2331, 0xc6a59430, 0x35a266c0,
+	0x744ebc37, 0xfc82caa6, 0xe090d0b0, 0x33a7d815,
+	0xf104984a, 0x41ecdaf7, 0x7fcd500e, 0x1791f62f,
+	0x764dd68d, 0x43efb04d, 0xccaa4d54, 0xe49604df,
+	0x9ed1b5e3, 0x4c6a881b, 0xc12c1fb8, 0x4665517f,
+	0x9d5eea04, 0x018c355d, 0xfa877473, 0xfb0b412e,
+	0xb3671d5a, 0x92dbd252, 0xe9105633, 0x6dd64713,
+	0x9ad7618c, 0x37a10c7a, 0x59f8148e, 0xeb133c89,
+	0xcea927ee, 0xb761c935, 0xe11ce5ed, 0x7a47b13c,
+	0x9cd2df59, 0x55f2733f, 0x1814ce79, 0x73c737bf,
+	0x53f7cdea, 0x5ffdaa5b, 0xdf3d6f14, 0x7844db86,
+	0xcaaff381, 0xb968c43e, 0x3824342c, 0xc2a3405f,
+	0x161dc372, 0xbce2250c, 0x283c498b, 0xff0d9541,
+	0x39a80171, 0x080cb3de, 0xd8b4e49c, 0x6456c190,
+	0x7bcb8461, 0xd532b670, 0x486c5c74, 0xd0b85742);
+$code.=<<___;
+# Td4[256]
+.byte	0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38
+.byte	0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb
+.byte	0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87
+.byte	0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb
+.byte	0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d
+.byte	0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e
+.byte	0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2
+.byte	0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25
+.byte	0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16
+.byte	0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92
+.byte	0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda
+.byte	0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84
+.byte	0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a
+.byte	0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06
+.byte	0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02
+.byte	0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b
+.byte	0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea
+.byte	0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73
+.byte	0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85
+.byte	0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e
+.byte	0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89
+.byte	0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b
+.byte	0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20
+.byte	0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4
+.byte	0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31
+.byte	0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f
+.byte	0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d
+.byte	0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef
+.byte	0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0
+.byte	0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61
+.byte	0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26
+.byte	0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d
+.size	AES_Td,.-AES_Td
+
+# void AES_decrypt(const unsigned char *inp, unsigned char *out,
+# 		 const AES_KEY *key) {
+.globl	AES_decrypt
+.type	AES_decrypt,\@function
+AES_decrypt:
+___
+$code.=<<___ if (!$softonly);
+	l	%r0,240($key)
+	lhi	%r1,16
+	clr	%r0,%r1
+	jl	.Ldsoft
+
+	la	%r1,0($key)
+	#la	%r2,0($inp)
+	la	%r4,0($out)
+	lghi	%r3,16		# single block length
+	.long	0xb92e0042	# km %r4,%r2
+	brc	1,.-4		# can this happen?
+	br	%r14
+.align	64
+.Ldsoft:
+___
+$code.=<<___;
+	stmg	%r3,$ra,24($sp)
+
+	llgf	$s0,0($inp)
+	llgf	$s1,4($inp)
+	llgf	$s2,8($inp)
+	llgf	$s3,12($inp)
+
+	larl	$tbl,AES_Td
+	bras	$ra,_s390x_AES_decrypt
+
+	lg	$out,24($sp)
+	st	$s0,0($out)
+	st	$s1,4($out)
+	st	$s2,8($out)
+	st	$s3,12($out)
+
+	lmg	%r6,$ra,48($sp)
+	br	$ra
+.size	AES_decrypt,.-AES_decrypt
+
+.type   _s390x_AES_decrypt,\@function
+.align	16
+_s390x_AES_decrypt:
+	stg	$ra,152($sp)
+	x	$s0,0($key)
+	x	$s1,4($key)
+	x	$s2,8($key)
+	x	$s3,12($key)
+	l	$rounds,240($key)
+	llill	$mask,`0xff<<3`
+	aghi	$rounds,-1
+	j	.Ldec_loop
+.align	16
+.Ldec_loop:
+	srlg	$t1,$s0,`16-3`
+	srlg	$t2,$s0,`8-3`
+	sllg	$t3,$s0,`0+3`
+	srl	$s0,`24-3`
+	nr	$s0,$mask
+	nr	$t1,$mask
+	nr	$t2,$mask
+	ngr	$t3,$mask
+
+	sllg	$i1,$s1,`0+3`	# i0
+	srlg	$i2,$s1,`16-3`
+	srlg	$i3,$s1,`8-3`
+	srl	$s1,`24-3`
+	ngr	$i1,$mask
+	nr	$s1,$mask
+	nr	$i2,$mask
+	nr	$i3,$mask
+
+	l	$s0,0($s0,$tbl)	# Td0[s0>>24]
+	l	$t1,3($t1,$tbl)	# Td1[s0>>16]
+	l	$t2,2($t2,$tbl)	# Td2[s0>>8]
+	l	$t3,1($t3,$tbl)	# Td3[s0>>0]
+
+	x	$s0,1($i1,$tbl)	# Td3[s1>>0]
+	l	$s1,0($s1,$tbl)	# Td0[s1>>24]
+	x	$t2,3($i2,$tbl)	# Td1[s1>>16]
+	x	$t3,2($i3,$tbl)	# Td2[s1>>8]
+
+	srlg	$i1,$s2,`8-3`	# i0
+	sllg	$i2,$s2,`0+3`	# i1
+	srlg	$i3,$s2,`16-3`
+	srl	$s2,`24-3`
+	nr	$i1,$mask
+	ngr	$i2,$mask
+	nr	$s2,$mask
+	nr	$i3,$mask
+
+	xr	$s1,$t1
+	srlg	$ra,$s3,`8-3`	# i1
+	srlg	$t1,$s3,`16-3`	# i0
+	nr	$ra,$mask
+	la	$key,16($key)
+	nr	$t1,$mask
+
+	x	$s0,2($i1,$tbl)	# Td2[s2>>8]
+	x	$s1,1($i2,$tbl)	# Td3[s2>>0]
+	l	$s2,0($s2,$tbl)	# Td0[s2>>24]
+	x	$t3,3($i3,$tbl)	# Td1[s2>>16]
+
+	sllg	$i3,$s3,`0+3`	# i2
+	srl	$s3,`24-3`
+	ngr	$i3,$mask
+	nr	$s3,$mask
+
+	xr	$s2,$t2
+	x	$s0,0($key)
+	x	$s1,4($key)
+	x	$s2,8($key)
+	x	$t3,12($key)
+
+	x	$s0,3($t1,$tbl)	# Td1[s3>>16]
+	x	$s1,2($ra,$tbl)	# Td2[s3>>8]
+	x	$s2,1($i3,$tbl)	# Td3[s3>>0]
+	l	$s3,0($s3,$tbl)	# Td0[s3>>24]
+	xr	$s3,$t3
+
+	brct	$rounds,.Ldec_loop
+	.align	16
+
+	l	$t1,`2048+0`($tbl)	# prefetch Td4
+	l	$t2,`2048+64`($tbl)
+	l	$t3,`2048+128`($tbl)
+	l	$i1,`2048+192`($tbl)
+	llill	$mask,0xff
+
+	srlg	$i3,$s0,24	# i0
+	srlg	$t1,$s0,16
+	srlg	$t2,$s0,8
+	nr	$s0,$mask	# i3
+	nr	$t1,$mask
+
+	srlg	$i1,$s1,24
+	nr	$t2,$mask
+	srlg	$i2,$s1,16
+	srlg	$ra,$s1,8
+	nr	$s1,$mask	# i0
+	nr	$i2,$mask
+	nr	$ra,$mask
+
+	llgc	$i3,2048($i3,$tbl)	# Td4[s0>>24]
+	llgc	$t1,2048($t1,$tbl)	# Td4[s0>>16]
+	llgc	$t2,2048($t2,$tbl)	# Td4[s0>>8]
+	sll	$t1,16
+	llgc	$t3,2048($s0,$tbl)	# Td4[s0>>0]
+	sllg	$s0,$i3,24
+	sll	$t2,8
+
+	llgc	$s1,2048($s1,$tbl)	# Td4[s1>>0]
+	llgc	$i1,2048($i1,$tbl)	# Td4[s1>>24]
+	llgc	$i2,2048($i2,$tbl)	# Td4[s1>>16]
+	sll	$i1,24
+	llgc	$i3,2048($ra,$tbl)	# Td4[s1>>8]
+	sll	$i2,16
+	sll	$i3,8
+	or	$s0,$s1
+	or	$t1,$i1
+	or	$t2,$i2
+	or	$t3,$i3
+
+	srlg	$i1,$s2,8	# i0
+	srlg	$i2,$s2,24
+	srlg	$i3,$s2,16
+	nr	$s2,$mask	# i1
+	nr	$i1,$mask
+	nr	$i3,$mask
+	llgc	$i1,2048($i1,$tbl)	# Td4[s2>>8]
+	llgc	$s1,2048($s2,$tbl)	# Td4[s2>>0]
+	llgc	$i2,2048($i2,$tbl)	# Td4[s2>>24]
+	llgc	$i3,2048($i3,$tbl)	# Td4[s2>>16]
+	sll	$i1,8
+	sll	$i2,24
+	or	$s0,$i1
+	sll	$i3,16
+	or	$t2,$i2
+	or	$t3,$i3
+
+	srlg	$i1,$s3,16	# i0
+	srlg	$i2,$s3,8	# i1
+	srlg	$i3,$s3,24
+	nr	$s3,$mask	# i2
+	nr	$i1,$mask
+	nr	$i2,$mask
+
+	lg	$ra,152($sp)
+	or	$s1,$t1
+	l	$t0,16($key)
+	l	$t1,20($key)
+
+	llgc	$i1,2048($i1,$tbl)	# Td4[s3>>16]
+	llgc	$i2,2048($i2,$tbl)	# Td4[s3>>8]
+	sll	$i1,16
+	llgc	$s2,2048($s3,$tbl)	# Td4[s3>>0]
+	llgc	$s3,2048($i3,$tbl)	# Td4[s3>>24]
+	sll	$i2,8
+	sll	$s3,24
+	or	$s0,$i1
+	or	$s1,$i2
+	or	$s2,$t2
+	or	$s3,$t3
+
+	xr	$s0,$t0
+	xr	$s1,$t1
+	x	$s2,24($key)
+	x	$s3,28($key)
+
+	br	$ra	
+.size	_s390x_AES_decrypt,.-_s390x_AES_decrypt
+___
+
+$code.=<<___;
+# void AES_set_encrypt_key(const unsigned char *in, int bits,
+# 		 AES_KEY *key) {
+.globl	AES_set_encrypt_key
+.type	AES_set_encrypt_key,\@function
+.align	16
+AES_set_encrypt_key:
+	lghi	$t0,0
+	clgr	$inp,$t0
+	je	.Lminus1
+	clgr	$key,$t0
+	je	.Lminus1
+
+	lghi	$t0,128
+	clr	$bits,$t0
+	je	.Lproceed
+	lghi	$t0,192
+	clr	$bits,$t0
+	je	.Lproceed
+	lghi	$t0,256
+	clr	$bits,$t0
+	je	.Lproceed
+	lghi	%r2,-2
+	br	%r14
+
+.align	16
+.Lproceed:
+___
+$code.=<<___ if (!$softonly);
+	# convert bits to km code, [128,192,256]->[18,19,20]
+	lhi	%r5,-128
+	lhi	%r0,18
+	ar	%r5,$bits
+	srl	%r5,6
+	ar	%r5,%r0
+
+	larl	%r1,OPENSSL_s390xcap_P
+	lg	%r0,0(%r1)
+	tmhl	%r0,0x4000	# check for message-security assist
+	jz	.Lekey_internal
+
+	lghi	%r0,0		# query capability vector
+	la	%r1,16($sp)
+	.long	0xb92f0042	# kmc %r4,%r2
+
+	llihh	%r1,0x8000
+	srlg	%r1,%r1,0(%r5)
+	ng	%r1,16($sp)
+	jz	.Lekey_internal
+
+	lmg	%r0,%r1,0($inp)	# just copy 128 bits...
+	stmg	%r0,%r1,0($key)
+	lhi	%r0,192
+	cr	$bits,%r0
+	jl	1f
+	lg	%r1,16($inp)
+	stg	%r1,16($key)
+	je	1f
+	lg	%r1,24($inp)
+	stg	%r1,24($key)
+1:	st	$bits,236($key)	# save bits
+	st	%r5,240($key)	# save km code
+	lghi	%r2,0
+	br	%r14
+___
+$code.=<<___;
+.align	16
+.Lekey_internal:
+	stmg	%r6,%r13,48($sp)	# all non-volatile regs
+
+	larl	$tbl,AES_Te+2048
+
+	llgf	$s0,0($inp)
+	llgf	$s1,4($inp)
+	llgf	$s2,8($inp)
+	llgf	$s3,12($inp)
+	st	$s0,0($key)
+	st	$s1,4($key)
+	st	$s2,8($key)
+	st	$s3,12($key)
+	lghi	$t0,128
+	cr	$bits,$t0
+	jne	.Lnot128
+
+	llill	$mask,0xff
+	lghi	$t3,0			# i=0
+	lghi	$rounds,10
+	st	$rounds,240($key)
+
+	llgfr	$t2,$s3			# temp=rk[3]
+	srlg	$i1,$s3,8
+	srlg	$i2,$s3,16
+	srlg	$i3,$s3,24
+	nr	$t2,$mask
+	nr	$i1,$mask
+	nr	$i2,$mask
+
+.align	16
+.L128_loop:
+	la	$t2,0($t2,$tbl)
+	la	$i1,0($i1,$tbl)
+	la	$i2,0($i2,$tbl)
+	la	$i3,0($i3,$tbl)
+	icm	$t2,2,0($t2)		# Te4[rk[3]>>0]<<8
+	icm	$t2,4,0($i1)		# Te4[rk[3]>>8]<<16
+	icm	$t2,8,0($i2)		# Te4[rk[3]>>16]<<24
+	icm	$t2,1,0($i3)		# Te4[rk[3]>>24]
+	x	$t2,256($t3,$tbl)	# rcon[i]
+	xr	$s0,$t2			# rk[4]=rk[0]^...
+	xr	$s1,$s0			# rk[5]=rk[1]^rk[4]
+	xr	$s2,$s1			# rk[6]=rk[2]^rk[5]
+	xr	$s3,$s2			# rk[7]=rk[3]^rk[6]
+
+	llgfr	$t2,$s3			# temp=rk[3]
+	srlg	$i1,$s3,8
+	srlg	$i2,$s3,16
+	nr	$t2,$mask
+	nr	$i1,$mask
+	srlg	$i3,$s3,24
+	nr	$i2,$mask
+
+	st	$s0,16($key)
+	st	$s1,20($key)
+	st	$s2,24($key)
+	st	$s3,28($key)
+	la	$key,16($key)		# key+=4
+	la	$t3,4($t3)		# i++
+	brct	$rounds,.L128_loop
+	lghi	%r2,0
+	lmg	%r6,%r13,48($sp)
+	br	$ra
+
+.align	16
+.Lnot128:
+	llgf	$t0,16($inp)
+	llgf	$t1,20($inp)
+	st	$t0,16($key)
+	st	$t1,20($key)
+	lghi	$t0,192
+	cr	$bits,$t0
+	jne	.Lnot192
+
+	llill	$mask,0xff
+	lghi	$t3,0			# i=0
+	lghi	$rounds,12
+	st	$rounds,240($key)
+	lghi	$rounds,8
+
+	srlg	$i1,$t1,8
+	srlg	$i2,$t1,16
+	srlg	$i3,$t1,24
+	nr	$t1,$mask
+	nr	$i1,$mask
+	nr	$i2,$mask
+
+.align	16
+.L192_loop:
+	la	$t1,0($t1,$tbl)
+	la	$i1,0($i1,$tbl)
+	la	$i2,0($i2,$tbl)
+	la	$i3,0($i3,$tbl)
+	icm	$t1,2,0($t1)		# Te4[rk[5]>>0]<<8
+	icm	$t1,4,0($i1)		# Te4[rk[5]>>8]<<16
+	icm	$t1,8,0($i2)		# Te4[rk[5]>>16]<<24
+	icm	$t1,1,0($i3)		# Te4[rk[5]>>24]
+	x	$t1,256($t3,$tbl)	# rcon[i]
+	xr	$s0,$t1			# rk[6]=rk[0]^...
+	xr	$s1,$s0			# rk[7]=rk[1]^rk[6]
+	xr	$s2,$s1			# rk[8]=rk[2]^rk[7]
+	xr	$s3,$s2			# rk[9]=rk[3]^rk[8]
+
+	st	$s0,24($key)
+	st	$s1,28($key)
+	st	$s2,32($key)
+	st	$s3,36($key)
+	brct	$rounds,.L192_continue
+	lghi	%r2,0
+	lmg	%r6,%r13,48($sp)
+	br	$ra
+
+.align	16
+.L192_continue:
+	lgr	$t1,$s3
+	x	$t1,16($key)		# rk[10]=rk[4]^rk[9]
+	st	$t1,40($key)
+	x	$t1,20($key)		# rk[11]=rk[5]^rk[10]
+	st	$t1,44($key)
+
+	srlg	$i1,$t1,8
+	srlg	$i2,$t1,16
+	srlg	$i3,$t1,24
+	nr	$t1,$mask
+	nr	$i1,$mask
+	nr	$i2,$mask
+
+	la	$key,24($key)		# key+=6
+	la	$t3,4($t3)		# i++
+	j	.L192_loop
+
+.align	16
+.Lnot192:
+	llgf	$t0,24($inp)
+	llgf	$t1,28($inp)
+	st	$t0,24($key)
+	st	$t1,28($key)
+	llill	$mask,0xff
+	lghi	$t3,0			# i=0
+	lghi	$rounds,14
+	st	$rounds,240($key)
+	lghi	$rounds,7
+
+	srlg	$i1,$t1,8
+	srlg	$i2,$t1,16
+	srlg	$i3,$t1,24
+	nr	$t1,$mask
+	nr	$i1,$mask
+	nr	$i2,$mask
+
+.align	16
+.L256_loop:
+	la	$t1,0($t1,$tbl)
+	la	$i1,0($i1,$tbl)
+	la	$i2,0($i2,$tbl)
+	la	$i3,0($i3,$tbl)
+	icm	$t1,2,0($t1)		# Te4[rk[7]>>0]<<8
+	icm	$t1,4,0($i1)		# Te4[rk[7]>>8]<<16
+	icm	$t1,8,0($i2)		# Te4[rk[7]>>16]<<24
+	icm	$t1,1,0($i3)		# Te4[rk[7]>>24]
+	x	$t1,256($t3,$tbl)	# rcon[i]
+	xr	$s0,$t1			# rk[8]=rk[0]^...
+	xr	$s1,$s0			# rk[9]=rk[1]^rk[8]
+	xr	$s2,$s1			# rk[10]=rk[2]^rk[9]
+	xr	$s3,$s2			# rk[11]=rk[3]^rk[10]
+	st	$s0,32($key)
+	st	$s1,36($key)
+	st	$s2,40($key)
+	st	$s3,44($key)
+	brct	$rounds,.L256_continue
+	lghi	%r2,0
+	lmg	%r6,%r13,48($sp)
+	br	$ra
+
+.align	16
+.L256_continue:
+	lgr	$t1,$s3			# temp=rk[11]
+	srlg	$i1,$s3,8
+	srlg	$i2,$s3,16
+	srlg	$i3,$s3,24
+	nr	$t1,$mask
+	nr	$i1,$mask
+	nr	$i2,$mask
+	la	$t1,0($t1,$tbl)
+	la	$i1,0($i1,$tbl)
+	la	$i2,0($i2,$tbl)
+	la	$i3,0($i3,$tbl)
+	llgc	$t1,0($t1)		# Te4[rk[11]>>0]
+	icm	$t1,2,0($i1)		# Te4[rk[11]>>8]<<8
+	icm	$t1,4,0($i2)		# Te4[rk[11]>>16]<<16
+	icm	$t1,8,0($i3)		# Te4[rk[11]>>24]<<24
+	x	$t1,16($key)		# rk[12]=rk[4]^...
+	st	$t1,48($key)
+	x	$t1,20($key)		# rk[13]=rk[5]^rk[12]
+	st	$t1,52($key)
+	x	$t1,24($key)		# rk[14]=rk[6]^rk[13]
+	st	$t1,56($key)
+	x	$t1,28($key)		# rk[15]=rk[7]^rk[14]
+	st	$t1,60($key)
+
+	srlg	$i1,$t1,8
+	srlg	$i2,$t1,16
+	srlg	$i3,$t1,24
+	nr	$t1,$mask
+	nr	$i1,$mask
+	nr	$i2,$mask
+
+	la	$key,32($key)		# key+=8
+	la	$t3,4($t3)		# i++
+	j	.L256_loop
+
+.Lminus1:
+	lghi	%r2,-1
+	br	$ra
+.size	AES_set_encrypt_key,.-AES_set_encrypt_key
+
+# void AES_set_decrypt_key(const unsigned char *in, int bits,
+# 		 AES_KEY *key) {
+.globl	AES_set_decrypt_key
+.type	AES_set_decrypt_key,\@function
+.align	16
+AES_set_decrypt_key:
+	stg	$key,32($sp)		# I rely on AES_set_encrypt_key to
+	stg	$ra,112($sp)		# save non-volatile registers!
+	bras	$ra,AES_set_encrypt_key
+	lg	$key,32($sp)
+	lg	$ra,112($sp)
+	ltgr	%r2,%r2
+	bnzr	$ra
+___
+$code.=<<___ if (!$softonly);
+	l	$t0,240($key)
+	lhi	$t1,16
+	cr	$t0,$t1
+	jl	.Lgo
+	oill	$t0,0x80	# set "decrypt" bit
+	st	$t0,240($key)
+	br	$ra
+
+.align	16
+.Ldkey_internal:
+	stg	$key,32($sp)
+	stg	$ra,40($sp)
+	bras	$ra,.Lekey_internal
+	lg	$key,32($sp)
+	lg	$ra,40($sp)
+___
+$code.=<<___;
+
+.Lgo:	llgf	$rounds,240($key)
+	la	$i1,0($key)
+	sllg	$i2,$rounds,4
+	la	$i2,0($i2,$key)
+	srl	$rounds,1
+	lghi	$t1,-16
+
+.align	16
+.Linv:	lmg	$s0,$s1,0($i1)
+	lmg	$s2,$s3,0($i2)
+	stmg	$s0,$s1,0($i2)
+	stmg	$s2,$s3,0($i1)
+	la	$i1,16($i1)
+	la	$i2,0($t1,$i2)
+	brct	$rounds,.Linv
+___
+$mask80=$i1;
+$mask1b=$i2;
+$maskfe=$i3;
+$code.=<<___;
+	llgf	$rounds,240($key)
+	aghi	$rounds,-1
+	sll	$rounds,2	# (rounds-1)*4
+	llilh	$mask80,0x8080
+	llilh	$mask1b,0x1b1b
+	llilh	$maskfe,0xfefe
+	oill	$mask80,0x8080
+	oill	$mask1b,0x1b1b
+	oill	$maskfe,0xfefe
+
+.align	16
+.Lmix:	l	$s0,16($key)	# tp1
+	lr	$s1,$s0
+	ngr	$s1,$mask80
+	srlg	$t1,$s1,7
+	slr	$s1,$t1
+	nr	$s1,$mask1b
+	sllg	$t1,$s0,1
+	nr	$t1,$maskfe
+	xr	$s1,$t1		# tp2
+
+	lr	$s2,$s1
+	ngr	$s2,$mask80
+	srlg	$t1,$s2,7
+	slr	$s2,$t1
+	nr	$s2,$mask1b
+	sllg	$t1,$s1,1
+	nr	$t1,$maskfe
+	xr	$s2,$t1		# tp4
+
+	lr	$s3,$s2
+	ngr	$s3,$mask80
+	srlg	$t1,$s3,7
+	slr	$s3,$t1
+	nr	$s3,$mask1b
+	sllg	$t1,$s2,1
+	nr	$t1,$maskfe
+	xr	$s3,$t1		# tp8
+
+	xr	$s1,$s0		# tp2^tp1
+	xr	$s2,$s0		# tp4^tp1
+	rll	$s0,$s0,24	# = ROTATE(tp1,8)
+	xr	$s2,$s3		# ^=tp8
+	xr	$s0,$s1		# ^=tp2^tp1
+	xr	$s1,$s3		# tp2^tp1^tp8
+	xr	$s0,$s2		# ^=tp4^tp1^tp8
+	rll	$s1,$s1,8
+	rll	$s2,$s2,16
+	xr	$s0,$s1		# ^= ROTATE(tp8^tp2^tp1,24)
+	rll	$s3,$s3,24
+	xr	$s0,$s2    	# ^= ROTATE(tp8^tp4^tp1,16)
+	xr	$s0,$s3		# ^= ROTATE(tp8,8)
+
+	st	$s0,16($key)
+	la	$key,4($key)
+	brct	$rounds,.Lmix
+
+	lmg	%r6,%r13,48($sp)# as was saved by AES_set_encrypt_key!
+	lghi	%r2,0
+	br	$ra
+.size	AES_set_decrypt_key,.-AES_set_decrypt_key
+___
+
+#void AES_cbc_encrypt(const unsigned char *in, unsigned char *out,
+#                     size_t length, const AES_KEY *key,
+#                     unsigned char *ivec, const int enc)
+{
+my $inp="%r2";
+my $out="%r4";	# length and out are swapped
+my $len="%r3";
+my $key="%r5";
+my $ivp="%r6";
+
+$code.=<<___;
+.globl	AES_cbc_encrypt
+.type	AES_cbc_encrypt,\@function
+.align	16
+AES_cbc_encrypt:
+	xgr	%r3,%r4		# flip %r3 and %r4, out and len
+	xgr	%r4,%r3
+	xgr	%r3,%r4
+___
+$code.=<<___ if (!$softonly);
+	lhi	%r0,16
+	cl	%r0,240($key)
+	jh	.Lcbc_software
+
+	lg	%r0,0($ivp)	# copy ivec
+	lg	%r1,8($ivp)
+	stmg	%r0,%r1,16($sp)
+	lmg	%r0,%r1,0($key)	# copy key, cover 256 bit
+	stmg	%r0,%r1,32($sp)
+	lmg	%r0,%r1,16($key)
+	stmg	%r0,%r1,48($sp)
+	l	%r0,240($key)	# load kmc code
+	lghi	$key,15		# res=len%16, len-=res;
+	ngr	$key,$len
+	slgr	$len,$key
+	la	%r1,16($sp)	# parameter block - ivec || key
+	jz	.Lkmc_truncated
+	.long	0xb92f0042	# kmc %r4,%r2
+	brc	1,.-4		# pay attention to "partial completion"
+	ltr	$key,$key
+	jnz	.Lkmc_truncated
+.Lkmc_done:
+	lmg	%r0,%r1,16($sp)	# copy ivec to caller
+	stg	%r0,0($ivp)
+	stg	%r1,8($ivp)
+	br	$ra
+.align	16
+.Lkmc_truncated:
+	ahi	$key,-1		# it's the way it's encoded in mvc
+	tmll	%r0,0x80
+	jnz	.Lkmc_truncated_dec
+	lghi	%r1,0
+	stg	%r1,128($sp)
+	stg	%r1,136($sp)
+	bras	%r1,1f
+	mvc	128(1,$sp),0($inp)
+1:	ex	$key,0(%r1)
+	la	%r1,16($sp)	# restore parameter block
+	la	$inp,128($sp)
+	lghi	$len,16
+	.long	0xb92f0042	# kmc %r4,%r2
+	j	.Lkmc_done
+.align	16
+.Lkmc_truncated_dec:
+	stg	$out,64($sp)
+	la	$out,128($sp)
+	lghi	$len,16
+	.long	0xb92f0042	# kmc %r4,%r2
+	lg	$out,64($sp)
+	bras	%r1,2f
+	mvc	0(1,$out),128($sp)
+2:	ex	$key,0(%r1)
+	j	.Lkmc_done
+.align	16
+.Lcbc_software:
+___
+$code.=<<___;
+	stmg	$key,$ra,40($sp)
+	lhi	%r0,0
+	cl	%r0,164($sp)
+	je	.Lcbc_decrypt
+
+	larl	$tbl,AES_Te
+
+	llgf	$s0,0($ivp)
+	llgf	$s1,4($ivp)
+	llgf	$s2,8($ivp)
+	llgf	$s3,12($ivp)
+
+	lghi	$t0,16
+	slgr	$len,$t0
+	brc	4,.Lcbc_enc_tail	# if borrow
+.Lcbc_enc_loop:
+	stmg	$inp,$out,16($sp)
+	x	$s0,0($inp)
+	x	$s1,4($inp)
+	x	$s2,8($inp)
+	x	$s3,12($inp)
+	lgr	%r4,$key
+
+	bras	$ra,_s390x_AES_encrypt
+
+	lmg	$inp,$key,16($sp)
+	st	$s0,0($out)
+	st	$s1,4($out)
+	st	$s2,8($out)
+	st	$s3,12($out)
+
+	la	$inp,16($inp)
+	la	$out,16($out)
+	lghi	$t0,16
+	ltgr	$len,$len
+	jz	.Lcbc_enc_done
+	slgr	$len,$t0
+	brc	4,.Lcbc_enc_tail	# if borrow
+	j	.Lcbc_enc_loop
+.align	16
+.Lcbc_enc_done:
+	lg	$ivp,48($sp)
+	st	$s0,0($ivp)
+	st	$s1,4($ivp)	
+	st	$s2,8($ivp)
+	st	$s3,12($ivp)
+
+	lmg	%r7,$ra,56($sp)
+	br	$ra
+
+.align	16
+.Lcbc_enc_tail:
+	aghi	$len,15
+	lghi	$t0,0
+	stg	$t0,128($sp)
+	stg	$t0,136($sp)
+	bras	$t1,3f
+	mvc	128(1,$sp),0($inp)
+3:	ex	$len,0($t1)
+	lghi	$len,0
+	la	$inp,128($sp)
+	j	.Lcbc_enc_loop
+
+.align	16
+.Lcbc_decrypt:
+	larl	$tbl,AES_Td
+
+	lg	$t0,0($ivp)
+	lg	$t1,8($ivp)
+	stmg	$t0,$t1,128($sp)
+
+.Lcbc_dec_loop:
+	stmg	$inp,$out,16($sp)
+	llgf	$s0,0($inp)
+	llgf	$s1,4($inp)
+	llgf	$s2,8($inp)
+	llgf	$s3,12($inp)
+	lgr	%r4,$key
+
+	bras	$ra,_s390x_AES_decrypt
+
+	lmg	$inp,$key,16($sp)
+	sllg	$s0,$s0,32
+	sllg	$s2,$s2,32
+	lr	$s0,$s1
+	lr	$s2,$s3
+
+	lg	$t0,0($inp)
+	lg	$t1,8($inp)
+	xg	$s0,128($sp)
+	xg	$s2,136($sp)
+	lghi	$s1,16
+	slgr	$len,$s1
+	brc	4,.Lcbc_dec_tail	# if borrow
+	brc	2,.Lcbc_dec_done	# if zero
+	stg	$s0,0($out)
+	stg	$s2,8($out)
+	stmg	$t0,$t1,128($sp)
+
+	la	$inp,16($inp)
+	la	$out,16($out)
+	j	.Lcbc_dec_loop
+
+.Lcbc_dec_done:
+	stg	$s0,0($out)
+	stg	$s2,8($out)
+.Lcbc_dec_exit:
+	lmg	$ivp,$ra,48($sp)
+	stmg	$t0,$t1,0($ivp)
+
+	br	$ra
+
+.align	16
+.Lcbc_dec_tail:
+	aghi	$len,15
+	stg	$s0,128($sp)
+	stg	$s2,136($sp)
+	bras	$s1,4f
+	mvc	0(1,$out),128($sp)
+4:	ex	$len,0($s1)
+	j	.Lcbc_dec_exit
+.size	AES_cbc_encrypt,.-AES_cbc_encrypt
+.comm  OPENSSL_s390xcap_P,8,8
+___
+}
+$code.=<<___;
+.string	"AES for s390x, CRYPTOGAMS by <appro\@openssl.org>"
+___
+
+$code =~ s/\`([^\`]*)\`/eval $1/gem;
+print $code;
diff --git a/openssl/crypto/aes/asm/aes-sparcv9.pl b/openssl/crypto/aes/asm/aes-sparcv9.pl
new file mode 100755
index 00000000..c57b3a2d
--- /dev/null
+++ b/openssl/crypto/aes/asm/aes-sparcv9.pl
@@ -0,0 +1,1181 @@
+#!/usr/bin/env perl
+#
+# ====================================================================
+# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
+# project. Rights for redistribution and usage in source and binary
+# forms are granted according to the OpenSSL license.
+# ====================================================================
+#
+# Version 1.1
+#
+# The major reason for undertaken effort was to mitigate the hazard of
+# cache-timing attack. This is [currently and initially!] addressed in
+# two ways. 1. S-boxes are compressed from 5KB to 2KB+256B size each.
+# 2. References to them are scheduled for L2 cache latency, meaning
+# that the tables don't have to reside in L1 cache. Once again, this
+# is an initial draft and one should expect more countermeasures to
+# be implemented...
+#
+# Version 1.1 prefetches T[ed]4 in order to mitigate attack on last
+# round.
+#
+# Even though performance was not the primary goal [on the contrary,
+# extra shifts "induced" by compressed S-box and longer loop epilogue
+# "induced" by scheduling for L2 have negative effect on performance],
+# the code turned out to run in ~23 cycles per processed byte en-/
+# decrypted with 128-bit key. This is pretty good result for code
+# with mentioned qualities and UltraSPARC core. Compared to Sun C
+# generated code my encrypt procedure runs just few percents faster,
+# while decrypt one - whole 50% faster [yes, Sun C failed to generate
+# optimal decrypt procedure]. Compared to GNU C generated code both
+# procedures are more than 60% faster:-)
+
+$bits=32;
+for (@ARGV)	{ $bits=64 if (/\-m64/ || /\-xarch\=v9/); }
+if ($bits==64)	{ $bias=2047; $frame=192; }
+else		{ $bias=0;    $frame=112; }
+$locals=16;
+
+$acc0="%l0";
+$acc1="%o0";
+$acc2="%o1";
+$acc3="%o2";
+
+$acc4="%l1";
+$acc5="%o3";
+$acc6="%o4";
+$acc7="%o5";
+
+$acc8="%l2";
+$acc9="%o7";
+$acc10="%g1";
+$acc11="%g2";
+
+$acc12="%l3";
+$acc13="%g3";
+$acc14="%g4";
+$acc15="%g5";
+
+$t0="%l4";
+$t1="%l5";
+$t2="%l6";
+$t3="%l7";
+
+$s0="%i0";
+$s1="%i1";
+$s2="%i2";
+$s3="%i3";
+$tbl="%i4";
+$key="%i5";
+$rounds="%i7";	# aliases with return address, which is off-loaded to stack
+
+sub _data_word()
+{ my $i;
+    while(defined($i=shift)) { $code.=sprintf"\t.long\t0x%08x,0x%08x\n",$i,$i; }
+}
+
+$code.=<<___ if ($bits==64);
+.register	%g2,#scratch
+.register	%g3,#scratch
+___
+$code.=<<___;
+.section	".text",#alloc,#execinstr
+
+.align	256
+AES_Te:
+___
+&_data_word(
+	0xc66363a5, 0xf87c7c84, 0xee777799, 0xf67b7b8d,
+	0xfff2f20d, 0xd66b6bbd, 0xde6f6fb1, 0x91c5c554,
+	0x60303050, 0x02010103, 0xce6767a9, 0x562b2b7d,
+	0xe7fefe19, 0xb5d7d762, 0x4dababe6, 0xec76769a,
+	0x8fcaca45, 0x1f82829d, 0x89c9c940, 0xfa7d7d87,
+	0xeffafa15, 0xb25959eb, 0x8e4747c9, 0xfbf0f00b,
+	0x41adadec, 0xb3d4d467, 0x5fa2a2fd, 0x45afafea,
+	0x239c9cbf, 0x53a4a4f7, 0xe4727296, 0x9bc0c05b,
+	0x75b7b7c2, 0xe1fdfd1c, 0x3d9393ae, 0x4c26266a,
+	0x6c36365a, 0x7e3f3f41, 0xf5f7f702, 0x83cccc4f,
+	0x6834345c, 0x51a5a5f4, 0xd1e5e534, 0xf9f1f108,
+	0xe2717193, 0xabd8d873, 0x62313153, 0x2a15153f,
+	0x0804040c, 0x95c7c752, 0x46232365, 0x9dc3c35e,
+	0x30181828, 0x379696a1, 0x0a05050f, 0x2f9a9ab5,
+	0x0e070709, 0x24121236, 0x1b80809b, 0xdfe2e23d,
+	0xcdebeb26, 0x4e272769, 0x7fb2b2cd, 0xea75759f,
+	0x1209091b, 0x1d83839e, 0x582c2c74, 0x341a1a2e,
+	0x361b1b2d, 0xdc6e6eb2, 0xb45a5aee, 0x5ba0a0fb,
+	0xa45252f6, 0x763b3b4d, 0xb7d6d661, 0x7db3b3ce,
+	0x5229297b, 0xdde3e33e, 0x5e2f2f71, 0x13848497,
+	0xa65353f5, 0xb9d1d168, 0x00000000, 0xc1eded2c,
+	0x40202060, 0xe3fcfc1f, 0x79b1b1c8, 0xb65b5bed,
+	0xd46a6abe, 0x8dcbcb46, 0x67bebed9, 0x7239394b,
+	0x944a4ade, 0x984c4cd4, 0xb05858e8, 0x85cfcf4a,
+	0xbbd0d06b, 0xc5efef2a, 0x4faaaae5, 0xedfbfb16,
+	0x864343c5, 0x9a4d4dd7, 0x66333355, 0x11858594,
+	0x8a4545cf, 0xe9f9f910, 0x04020206, 0xfe7f7f81,
+	0xa05050f0, 0x783c3c44, 0x259f9fba, 0x4ba8a8e3,
+	0xa25151f3, 0x5da3a3fe, 0x804040c0, 0x058f8f8a,
+	0x3f9292ad, 0x219d9dbc, 0x70383848, 0xf1f5f504,
+	0x63bcbcdf, 0x77b6b6c1, 0xafdada75, 0x42212163,
+	0x20101030, 0xe5ffff1a, 0xfdf3f30e, 0xbfd2d26d,
+	0x81cdcd4c, 0x180c0c14, 0x26131335, 0xc3ecec2f,
+	0xbe5f5fe1, 0x359797a2, 0x884444cc, 0x2e171739,
+	0x93c4c457, 0x55a7a7f2, 0xfc7e7e82, 0x7a3d3d47,
+	0xc86464ac, 0xba5d5de7, 0x3219192b, 0xe6737395,
+	0xc06060a0, 0x19818198, 0x9e4f4fd1, 0xa3dcdc7f,
+	0x44222266, 0x542a2a7e, 0x3b9090ab, 0x0b888883,
+	0x8c4646ca, 0xc7eeee29, 0x6bb8b8d3, 0x2814143c,
+	0xa7dede79, 0xbc5e5ee2, 0x160b0b1d, 0xaddbdb76,
+	0xdbe0e03b, 0x64323256, 0x743a3a4e, 0x140a0a1e,
+	0x924949db, 0x0c06060a, 0x4824246c, 0xb85c5ce4,
+	0x9fc2c25d, 0xbdd3d36e, 0x43acacef, 0xc46262a6,
+	0x399191a8, 0x319595a4, 0xd3e4e437, 0xf279798b,
+	0xd5e7e732, 0x8bc8c843, 0x6e373759, 0xda6d6db7,
+	0x018d8d8c, 0xb1d5d564, 0x9c4e4ed2, 0x49a9a9e0,
+	0xd86c6cb4, 0xac5656fa, 0xf3f4f407, 0xcfeaea25,
+	0xca6565af, 0xf47a7a8e, 0x47aeaee9, 0x10080818,
+	0x6fbabad5, 0xf0787888, 0x4a25256f, 0x5c2e2e72,
+	0x381c1c24, 0x57a6a6f1, 0x73b4b4c7, 0x97c6c651,
+	0xcbe8e823, 0xa1dddd7c, 0xe874749c, 0x3e1f1f21,
+	0x964b4bdd, 0x61bdbddc, 0x0d8b8b86, 0x0f8a8a85,
+	0xe0707090, 0x7c3e3e42, 0x71b5b5c4, 0xcc6666aa,
+	0x904848d8, 0x06030305, 0xf7f6f601, 0x1c0e0e12,
+	0xc26161a3, 0x6a35355f, 0xae5757f9, 0x69b9b9d0,
+	0x17868691, 0x99c1c158, 0x3a1d1d27, 0x279e9eb9,
+	0xd9e1e138, 0xebf8f813, 0x2b9898b3, 0x22111133,
+	0xd26969bb, 0xa9d9d970, 0x078e8e89, 0x339494a7,
+	0x2d9b9bb6, 0x3c1e1e22, 0x15878792, 0xc9e9e920,
+	0x87cece49, 0xaa5555ff, 0x50282878, 0xa5dfdf7a,
+	0x038c8c8f, 0x59a1a1f8, 0x09898980, 0x1a0d0d17,
+	0x65bfbfda, 0xd7e6e631, 0x844242c6, 0xd06868b8,
+	0x824141c3, 0x299999b0, 0x5a2d2d77, 0x1e0f0f11,
+	0x7bb0b0cb, 0xa85454fc, 0x6dbbbbd6, 0x2c16163a);
+$code.=<<___;
+	.byte	0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5
+	.byte	0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76
+	.byte	0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0
+	.byte	0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0
+	.byte	0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc
+	.byte	0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15
+	.byte	0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a
+	.byte	0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75
+	.byte	0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0
+	.byte	0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84
+	.byte	0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b
+	.byte	0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf
+	.byte	0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85
+	.byte	0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8
+	.byte	0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5
+	.byte	0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2
+	.byte	0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17
+	.byte	0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73
+	.byte	0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88
+	.byte	0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb
+	.byte	0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c
+	.byte	0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79
+	.byte	0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9
+	.byte	0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08
+	.byte	0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6
+	.byte	0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a
+	.byte	0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e
+	.byte	0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e
+	.byte	0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94
+	.byte	0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf
+	.byte	0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68
+	.byte	0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16
+.type	AES_Te,#object
+.size	AES_Te,(.-AES_Te)
+
+.align	64
+.skip	16
+_sparcv9_AES_encrypt:
+	save	%sp,-$frame-$locals,%sp
+	stx	%i7,[%sp+$bias+$frame+0]	! off-load return address
+	ld	[$key+240],$rounds
+	ld	[$key+0],$t0
+	ld	[$key+4],$t1			!
+	ld	[$key+8],$t2
+	srl	$rounds,1,$rounds
+	xor	$t0,$s0,$s0
+	ld	[$key+12],$t3
+	srl	$s0,21,$acc0
+	xor	$t1,$s1,$s1
+	ld	[$key+16],$t0
+	srl	$s1,13,$acc1			!
+	xor	$t2,$s2,$s2
+	ld	[$key+20],$t1
+	xor	$t3,$s3,$s3
+	ld	[$key+24],$t2
+	and	$acc0,2040,$acc0
+	ld	[$key+28],$t3
+	nop
+.Lenc_loop:
+	srl	$s2,5,$acc2			!
+	and	$acc1,2040,$acc1
+	ldx	[$tbl+$acc0],$acc0
+	sll	$s3,3,$acc3
+	and	$acc2,2040,$acc2
+	ldx	[$tbl+$acc1],$acc1
+	srl	$s1,21,$acc4
+	and	$acc3,2040,$acc3
+	ldx	[$tbl+$acc2],$acc2		!
+	srl	$s2,13,$acc5
+	and	$acc4,2040,$acc4
+	ldx	[$tbl+$acc3],$acc3
+	srl	$s3,5,$acc6
+	and	$acc5,2040,$acc5
+	ldx	[$tbl+$acc4],$acc4
+	fmovs	%f0,%f0
+	sll	$s0,3,$acc7			!
+	and	$acc6,2040,$acc6
+	ldx	[$tbl+$acc5],$acc5
+	srl	$s2,21,$acc8
+	and	$acc7,2040,$acc7
+	ldx	[$tbl+$acc6],$acc6
+	srl	$s3,13,$acc9
+	and	$acc8,2040,$acc8
+	ldx	[$tbl+$acc7],$acc7		!
+	srl	$s0,5,$acc10
+	and	$acc9,2040,$acc9
+	ldx	[$tbl+$acc8],$acc8
+	sll	$s1,3,$acc11
+	and	$acc10,2040,$acc10
+	ldx	[$tbl+$acc9],$acc9
+	fmovs	%f0,%f0
+	srl	$s3,21,$acc12			!
+	and	$acc11,2040,$acc11
+	ldx	[$tbl+$acc10],$acc10
+	srl	$s0,13,$acc13
+	and	$acc12,2040,$acc12
+	ldx	[$tbl+$acc11],$acc11
+	srl	$s1,5,$acc14
+	and	$acc13,2040,$acc13
+	ldx	[$tbl+$acc12],$acc12		!
+	sll	$s2,3,$acc15
+	and	$acc14,2040,$acc14
+	ldx	[$tbl+$acc13],$acc13
+	and	$acc15,2040,$acc15
+	add	$key,32,$key
+	ldx	[$tbl+$acc14],$acc14
+	fmovs	%f0,%f0
+	subcc	$rounds,1,$rounds		!
+	ldx	[$tbl+$acc15],$acc15
+	bz,a,pn	%icc,.Lenc_last
+	add	$tbl,2048,$rounds
+
+		srlx	$acc1,8,$acc1
+		xor	$acc0,$t0,$t0
+	ld	[$key+0],$s0
+	fmovs	%f0,%f0
+		srlx	$acc2,16,$acc2		!
+		xor	$acc1,$t0,$t0
+	ld	[$key+4],$s1
+		srlx	$acc3,24,$acc3
+		xor	$acc2,$t0,$t0
+	ld	[$key+8],$s2
+		srlx	$acc5,8,$acc5
+		xor	$acc3,$t0,$t0
+	ld	[$key+12],$s3			!
+		srlx	$acc6,16,$acc6
+		xor	$acc4,$t1,$t1
+	fmovs	%f0,%f0
+		srlx	$acc7,24,$acc7
+		xor	$acc5,$t1,$t1
+		srlx	$acc9,8,$acc9
+		xor	$acc6,$t1,$t1
+		srlx	$acc10,16,$acc10	!
+		xor	$acc7,$t1,$t1
+		srlx	$acc11,24,$acc11
+		xor	$acc8,$t2,$t2
+		srlx	$acc13,8,$acc13
+		xor	$acc9,$t2,$t2
+		srlx	$acc14,16,$acc14
+		xor	$acc10,$t2,$t2
+		srlx	$acc15,24,$acc15	!
+		xor	$acc11,$t2,$t2
+		xor	$acc12,$acc14,$acc14
+		xor	$acc13,$t3,$t3
+	srl	$t0,21,$acc0
+		xor	$acc14,$t3,$t3
+	srl	$t1,13,$acc1
+		xor	$acc15,$t3,$t3
+
+	and	$acc0,2040,$acc0		!
+	srl	$t2,5,$acc2
+	and	$acc1,2040,$acc1
+	ldx	[$tbl+$acc0],$acc0
+	sll	$t3,3,$acc3
+	and	$acc2,2040,$acc2
+	ldx	[$tbl+$acc1],$acc1
+	fmovs	%f0,%f0
+	srl	$t1,21,$acc4			!
+	and	$acc3,2040,$acc3
+	ldx	[$tbl+$acc2],$acc2
+	srl	$t2,13,$acc5
+	and	$acc4,2040,$acc4
+	ldx	[$tbl+$acc3],$acc3
+	srl	$t3,5,$acc6
+	and	$acc5,2040,$acc5
+	ldx	[$tbl+$acc4],$acc4		!
+	sll	$t0,3,$acc7
+	and	$acc6,2040,$acc6
+	ldx	[$tbl+$acc5],$acc5
+	srl	$t2,21,$acc8
+	and	$acc7,2040,$acc7
+	ldx	[$tbl+$acc6],$acc6
+	fmovs	%f0,%f0
+	srl	$t3,13,$acc9			!
+	and	$acc8,2040,$acc8
+	ldx	[$tbl+$acc7],$acc7
+	srl	$t0,5,$acc10
+	and	$acc9,2040,$acc9
+	ldx	[$tbl+$acc8],$acc8
+	sll	$t1,3,$acc11
+	and	$acc10,2040,$acc10
+	ldx	[$tbl+$acc9],$acc9		!
+	srl	$t3,21,$acc12
+	and	$acc11,2040,$acc11
+	ldx	[$tbl+$acc10],$acc10
+	srl	$t0,13,$acc13
+	and	$acc12,2040,$acc12
+	ldx	[$tbl+$acc11],$acc11
+	fmovs	%f0,%f0
+	srl	$t1,5,$acc14			!
+	and	$acc13,2040,$acc13
+	ldx	[$tbl+$acc12],$acc12
+	sll	$t2,3,$acc15
+	and	$acc14,2040,$acc14
+	ldx	[$tbl+$acc13],$acc13
+		srlx	$acc1,8,$acc1
+	and	$acc15,2040,$acc15
+	ldx	[$tbl+$acc14],$acc14		!
+
+		srlx	$acc2,16,$acc2
+		xor	$acc0,$s0,$s0
+	ldx	[$tbl+$acc15],$acc15
+		srlx	$acc3,24,$acc3
+		xor	$acc1,$s0,$s0
+	ld	[$key+16],$t0
+	fmovs	%f0,%f0
+		srlx	$acc5,8,$acc5		!
+		xor	$acc2,$s0,$s0
+	ld	[$key+20],$t1
+		srlx	$acc6,16,$acc6
+		xor	$acc3,$s0,$s0
+	ld	[$key+24],$t2
+		srlx	$acc7,24,$acc7
+		xor	$acc4,$s1,$s1
+	ld	[$key+28],$t3			!
+		srlx	$acc9,8,$acc9
+		xor	$acc5,$s1,$s1
+	ldx	[$tbl+2048+0],%g0		! prefetch te4
+		srlx	$acc10,16,$acc10
+		xor	$acc6,$s1,$s1
+	ldx	[$tbl+2048+32],%g0		! prefetch te4
+		srlx	$acc11,24,$acc11
+		xor	$acc7,$s1,$s1
+	ldx	[$tbl+2048+64],%g0		! prefetch te4
+		srlx	$acc13,8,$acc13
+		xor	$acc8,$s2,$s2
+	ldx	[$tbl+2048+96],%g0		! prefetch te4
+		srlx	$acc14,16,$acc14	!
+		xor	$acc9,$s2,$s2
+	ldx	[$tbl+2048+128],%g0		! prefetch te4
+		srlx	$acc15,24,$acc15
+		xor	$acc10,$s2,$s2
+	ldx	[$tbl+2048+160],%g0		! prefetch te4
+	srl	$s0,21,$acc0
+		xor	$acc11,$s2,$s2
+	ldx	[$tbl+2048+192],%g0		! prefetch te4
+		xor	$acc12,$acc14,$acc14
+		xor	$acc13,$s3,$s3
+	ldx	[$tbl+2048+224],%g0		! prefetch te4
+	srl	$s1,13,$acc1			!
+		xor	$acc14,$s3,$s3
+		xor	$acc15,$s3,$s3
+	ba	.Lenc_loop
+	and	$acc0,2040,$acc0
+
+.align	32
+.Lenc_last:
+		srlx	$acc1,8,$acc1		!
+		xor	$acc0,$t0,$t0
+	ld	[$key+0],$s0
+		srlx	$acc2,16,$acc2
+		xor	$acc1,$t0,$t0
+	ld	[$key+4],$s1
+		srlx	$acc3,24,$acc3
+		xor	$acc2,$t0,$t0
+	ld	[$key+8],$s2			!
+		srlx	$acc5,8,$acc5
+		xor	$acc3,$t0,$t0
+	ld	[$key+12],$s3
+		srlx	$acc6,16,$acc6
+		xor	$acc4,$t1,$t1
+		srlx	$acc7,24,$acc7
+		xor	$acc5,$t1,$t1
+		srlx	$acc9,8,$acc9		!
+		xor	$acc6,$t1,$t1
+		srlx	$acc10,16,$acc10
+		xor	$acc7,$t1,$t1
+		srlx	$acc11,24,$acc11
+		xor	$acc8,$t2,$t2
+		srlx	$acc13,8,$acc13
+		xor	$acc9,$t2,$t2
+		srlx	$acc14,16,$acc14	!
+		xor	$acc10,$t2,$t2
+		srlx	$acc15,24,$acc15
+		xor	$acc11,$t2,$t2
+		xor	$acc12,$acc14,$acc14
+		xor	$acc13,$t3,$t3
+	srl	$t0,24,$acc0
+		xor	$acc14,$t3,$t3
+	srl	$t1,16,$acc1			!
+		xor	$acc15,$t3,$t3
+
+	srl	$t2,8,$acc2
+	and	$acc1,255,$acc1
+	ldub	[$rounds+$acc0],$acc0
+	srl	$t1,24,$acc4
+	and	$acc2,255,$acc2
+	ldub	[$rounds+$acc1],$acc1
+	srl	$t2,16,$acc5			!
+	and	$t3,255,$acc3
+	ldub	[$rounds+$acc2],$acc2
+	ldub	[$rounds+$acc3],$acc3
+	srl	$t3,8,$acc6
+	and	$acc5,255,$acc5
+	ldub	[$rounds+$acc4],$acc4
+	fmovs	%f0,%f0
+	srl	$t2,24,$acc8			!
+	and	$acc6,255,$acc6
+	ldub	[$rounds+$acc5],$acc5
+	srl	$t3,16,$acc9
+	and	$t0,255,$acc7
+	ldub	[$rounds+$acc6],$acc6
+	ldub	[$rounds+$acc7],$acc7
+	fmovs	%f0,%f0
+	srl	$t0,8,$acc10			!
+	and	$acc9,255,$acc9
+	ldub	[$rounds+$acc8],$acc8
+	srl	$t3,24,$acc12
+	and	$acc10,255,$acc10
+	ldub	[$rounds+$acc9],$acc9
+	srl	$t0,16,$acc13
+	and	$t1,255,$acc11
+	ldub	[$rounds+$acc10],$acc10		!
+	srl	$t1,8,$acc14
+	and	$acc13,255,$acc13
+	ldub	[$rounds+$acc11],$acc11
+	ldub	[$rounds+$acc12],$acc12
+	and	$acc14,255,$acc14
+	ldub	[$rounds+$acc13],$acc13
+	and	$t2,255,$acc15
+	ldub	[$rounds+$acc14],$acc14		!
+
+		sll	$acc0,24,$acc0
+		xor	$acc3,$s0,$s0
+	ldub	[$rounds+$acc15],$acc15
+		sll	$acc1,16,$acc1
+		xor	$acc0,$s0,$s0
+	ldx	[%sp+$bias+$frame+0],%i7	! restore return address
+	fmovs	%f0,%f0
+		sll	$acc2,8,$acc2		!
+		xor	$acc1,$s0,$s0
+		sll	$acc4,24,$acc4
+		xor	$acc2,$s0,$s0
+		sll	$acc5,16,$acc5
+		xor	$acc7,$s1,$s1
+		sll	$acc6,8,$acc6
+		xor	$acc4,$s1,$s1
+		sll	$acc8,24,$acc8		!
+		xor	$acc5,$s1,$s1
+		sll	$acc9,16,$acc9
+		xor	$acc11,$s2,$s2
+		sll	$acc10,8,$acc10
+		xor	$acc6,$s1,$s1
+		sll	$acc12,24,$acc12
+		xor	$acc8,$s2,$s2
+		sll	$acc13,16,$acc13	!
+		xor	$acc9,$s2,$s2
+		sll	$acc14,8,$acc14
+		xor	$acc10,$s2,$s2
+		xor	$acc12,$acc14,$acc14
+		xor	$acc13,$s3,$s3
+		xor	$acc14,$s3,$s3
+		xor	$acc15,$s3,$s3
+
+	ret
+	restore
+.type	_sparcv9_AES_encrypt,#function
+.size	_sparcv9_AES_encrypt,(.-_sparcv9_AES_encrypt)
+
+.align	32
+.globl	AES_encrypt
+AES_encrypt:
+	or	%o0,%o1,%g1
+	andcc	%g1,3,%g0
+	bnz,pn	%xcc,.Lunaligned_enc
+	save	%sp,-$frame,%sp
+
+	ld	[%i0+0],%o0
+	ld	[%i0+4],%o1
+	ld	[%i0+8],%o2
+	ld	[%i0+12],%o3
+
+1:	call	.+8
+	add	%o7,AES_Te-1b,%o4
+	call	_sparcv9_AES_encrypt
+	mov	%i2,%o5
+
+	st	%o0,[%i1+0]
+	st	%o1,[%i1+4]
+	st	%o2,[%i1+8]
+	st	%o3,[%i1+12]
+
+	ret
+	restore
+
+.align	32
+.Lunaligned_enc:
+	ldub	[%i0+0],%l0
+	ldub	[%i0+1],%l1
+	ldub	[%i0+2],%l2
+
+	sll	%l0,24,%l0
+	ldub	[%i0+3],%l3
+	sll	%l1,16,%l1
+	ldub	[%i0+4],%l4
+	sll	%l2,8,%l2
+	or	%l1,%l0,%l0
+	ldub	[%i0+5],%l5
+	sll	%l4,24,%l4
+	or	%l3,%l2,%l2
+	ldub	[%i0+6],%l6
+	sll	%l5,16,%l5
+	or	%l0,%l2,%o0
+	ldub	[%i0+7],%l7
+
+	sll	%l6,8,%l6
+	or	%l5,%l4,%l4
+	ldub	[%i0+8],%l0
+	or	%l7,%l6,%l6
+	ldub	[%i0+9],%l1
+	or	%l4,%l6,%o1
+	ldub	[%i0+10],%l2
+
+	sll	%l0,24,%l0
+	ldub	[%i0+11],%l3
+	sll	%l1,16,%l1
+	ldub	[%i0+12],%l4
+	sll	%l2,8,%l2
+	or	%l1,%l0,%l0
+	ldub	[%i0+13],%l5
+	sll	%l4,24,%l4
+	or	%l3,%l2,%l2
+	ldub	[%i0+14],%l6
+	sll	%l5,16,%l5
+	or	%l0,%l2,%o2
+	ldub	[%i0+15],%l7
+
+	sll	%l6,8,%l6
+	or	%l5,%l4,%l4
+	or	%l7,%l6,%l6
+	or	%l4,%l6,%o3
+
+1:	call	.+8
+	add	%o7,AES_Te-1b,%o4
+	call	_sparcv9_AES_encrypt
+	mov	%i2,%o5
+
+	srl	%o0,24,%l0
+	srl	%o0,16,%l1
+	stb	%l0,[%i1+0]
+	srl	%o0,8,%l2
+	stb	%l1,[%i1+1]
+	stb	%l2,[%i1+2]
+	srl	%o1,24,%l4
+	stb	%o0,[%i1+3]
+
+	srl	%o1,16,%l5
+	stb	%l4,[%i1+4]
+	srl	%o1,8,%l6
+	stb	%l5,[%i1+5]
+	stb	%l6,[%i1+6]
+	srl	%o2,24,%l0
+	stb	%o1,[%i1+7]
+
+	srl	%o2,16,%l1
+	stb	%l0,[%i1+8]
+	srl	%o2,8,%l2
+	stb	%l1,[%i1+9]
+	stb	%l2,[%i1+10]
+	srl	%o3,24,%l4
+	stb	%o2,[%i1+11]
+
+	srl	%o3,16,%l5
+	stb	%l4,[%i1+12]
+	srl	%o3,8,%l6
+	stb	%l5,[%i1+13]
+	stb	%l6,[%i1+14]
+	stb	%o3,[%i1+15]
+
+	ret
+	restore
+.type	AES_encrypt,#function
+.size	AES_encrypt,(.-AES_encrypt)
+
+___
+
+$code.=<<___;
+.align	256
+AES_Td:
+___
+&_data_word(
+	0x51f4a750, 0x7e416553, 0x1a17a4c3, 0x3a275e96,
+	0x3bab6bcb, 0x1f9d45f1, 0xacfa58ab, 0x4be30393,
+	0x2030fa55, 0xad766df6, 0x88cc7691, 0xf5024c25,
+	0x4fe5d7fc, 0xc52acbd7, 0x26354480, 0xb562a38f,
+	0xdeb15a49, 0x25ba1b67, 0x45ea0e98, 0x5dfec0e1,
+	0xc32f7502, 0x814cf012, 0x8d4697a3, 0x6bd3f9c6,
+	0x038f5fe7, 0x15929c95, 0xbf6d7aeb, 0x955259da,
+	0xd4be832d, 0x587421d3, 0x49e06929, 0x8ec9c844,
+	0x75c2896a, 0xf48e7978, 0x99583e6b, 0x27b971dd,
+	0xbee14fb6, 0xf088ad17, 0xc920ac66, 0x7dce3ab4,
+	0x63df4a18, 0xe51a3182, 0x97513360, 0x62537f45,
+	0xb16477e0, 0xbb6bae84, 0xfe81a01c, 0xf9082b94,
+	0x70486858, 0x8f45fd19, 0x94de6c87, 0x527bf8b7,
+	0xab73d323, 0x724b02e2, 0xe31f8f57, 0x6655ab2a,
+	0xb2eb2807, 0x2fb5c203, 0x86c57b9a, 0xd33708a5,
+	0x302887f2, 0x23bfa5b2, 0x02036aba, 0xed16825c,
+	0x8acf1c2b, 0xa779b492, 0xf307f2f0, 0x4e69e2a1,
+	0x65daf4cd, 0x0605bed5, 0xd134621f, 0xc4a6fe8a,
+	0x342e539d, 0xa2f355a0, 0x058ae132, 0xa4f6eb75,
+	0x0b83ec39, 0x4060efaa, 0x5e719f06, 0xbd6e1051,
+	0x3e218af9, 0x96dd063d, 0xdd3e05ae, 0x4de6bd46,
+	0x91548db5, 0x71c45d05, 0x0406d46f, 0x605015ff,
+	0x1998fb24, 0xd6bde997, 0x894043cc, 0x67d99e77,
+	0xb0e842bd, 0x07898b88, 0xe7195b38, 0x79c8eedb,
+	0xa17c0a47, 0x7c420fe9, 0xf8841ec9, 0x00000000,
+	0x09808683, 0x322bed48, 0x1e1170ac, 0x6c5a724e,
+	0xfd0efffb, 0x0f853856, 0x3daed51e, 0x362d3927,
+	0x0a0fd964, 0x685ca621, 0x9b5b54d1, 0x24362e3a,
+	0x0c0a67b1, 0x9357e70f, 0xb4ee96d2, 0x1b9b919e,
+	0x80c0c54f, 0x61dc20a2, 0x5a774b69, 0x1c121a16,
+	0xe293ba0a, 0xc0a02ae5, 0x3c22e043, 0x121b171d,
+	0x0e090d0b, 0xf28bc7ad, 0x2db6a8b9, 0x141ea9c8,
+	0x57f11985, 0xaf75074c, 0xee99ddbb, 0xa37f60fd,
+	0xf701269f, 0x5c72f5bc, 0x44663bc5, 0x5bfb7e34,
+	0x8b432976, 0xcb23c6dc, 0xb6edfc68, 0xb8e4f163,
+	0xd731dcca, 0x42638510, 0x13972240, 0x84c61120,
+	0x854a247d, 0xd2bb3df8, 0xaef93211, 0xc729a16d,
+	0x1d9e2f4b, 0xdcb230f3, 0x0d8652ec, 0x77c1e3d0,
+	0x2bb3166c, 0xa970b999, 0x119448fa, 0x47e96422,
+	0xa8fc8cc4, 0xa0f03f1a, 0x567d2cd8, 0x223390ef,
+	0x87494ec7, 0xd938d1c1, 0x8ccaa2fe, 0x98d40b36,
+	0xa6f581cf, 0xa57ade28, 0xdab78e26, 0x3fadbfa4,
+	0x2c3a9de4, 0x5078920d, 0x6a5fcc9b, 0x547e4662,
+	0xf68d13c2, 0x90d8b8e8, 0x2e39f75e, 0x82c3aff5,
+	0x9f5d80be, 0x69d0937c, 0x6fd52da9, 0xcf2512b3,
+	0xc8ac993b, 0x10187da7, 0xe89c636e, 0xdb3bbb7b,
+	0xcd267809, 0x6e5918f4, 0xec9ab701, 0x834f9aa8,
+	0xe6956e65, 0xaaffe67e, 0x21bccf08, 0xef15e8e6,
+	0xbae79bd9, 0x4a6f36ce, 0xea9f09d4, 0x29b07cd6,
+	0x31a4b2af, 0x2a3f2331, 0xc6a59430, 0x35a266c0,
+	0x744ebc37, 0xfc82caa6, 0xe090d0b0, 0x33a7d815,
+	0xf104984a, 0x41ecdaf7, 0x7fcd500e, 0x1791f62f,
+	0x764dd68d, 0x43efb04d, 0xccaa4d54, 0xe49604df,
+	0x9ed1b5e3, 0x4c6a881b, 0xc12c1fb8, 0x4665517f,
+	0x9d5eea04, 0x018c355d, 0xfa877473, 0xfb0b412e,
+	0xb3671d5a, 0x92dbd252, 0xe9105633, 0x6dd64713,
+	0x9ad7618c, 0x37a10c7a, 0x59f8148e, 0xeb133c89,
+	0xcea927ee, 0xb761c935, 0xe11ce5ed, 0x7a47b13c,
+	0x9cd2df59, 0x55f2733f, 0x1814ce79, 0x73c737bf,
+	0x53f7cdea, 0x5ffdaa5b, 0xdf3d6f14, 0x7844db86,
+	0xcaaff381, 0xb968c43e, 0x3824342c, 0xc2a3405f,
+	0x161dc372, 0xbce2250c, 0x283c498b, 0xff0d9541,
+	0x39a80171, 0x080cb3de, 0xd8b4e49c, 0x6456c190,
+	0x7bcb8461, 0xd532b670, 0x486c5c74, 0xd0b85742);
+$code.=<<___;
+	.byte	0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38
+	.byte	0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb
+	.byte	0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87
+	.byte	0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb
+	.byte	0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d
+	.byte	0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e
+	.byte	0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2
+	.byte	0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25
+	.byte	0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16
+	.byte	0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92
+	.byte	0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda
+	.byte	0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84
+	.byte	0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a
+	.byte	0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06
+	.byte	0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02
+	.byte	0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b
+	.byte	0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea
+	.byte	0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73
+	.byte	0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85
+	.byte	0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e
+	.byte	0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89
+	.byte	0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b
+	.byte	0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20
+	.byte	0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4
+	.byte	0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31
+	.byte	0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f
+	.byte	0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d
+	.byte	0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef
+	.byte	0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0
+	.byte	0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61
+	.byte	0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26
+	.byte	0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d
+.type	AES_Td,#object
+.size	AES_Td,(.-AES_Td)
+
+.align	64
+.skip	16
+_sparcv9_AES_decrypt:
+	save	%sp,-$frame-$locals,%sp
+	stx	%i7,[%sp+$bias+$frame+0]	! off-load return address
+	ld	[$key+240],$rounds
+	ld	[$key+0],$t0
+	ld	[$key+4],$t1			!
+	ld	[$key+8],$t2
+	ld	[$key+12],$t3
+	srl	$rounds,1,$rounds
+	xor	$t0,$s0,$s0
+	ld	[$key+16],$t0
+	xor	$t1,$s1,$s1
+	ld	[$key+20],$t1
+	srl	$s0,21,$acc0			!
+	xor	$t2,$s2,$s2
+	ld	[$key+24],$t2
+	xor	$t3,$s3,$s3
+	and	$acc0,2040,$acc0
+	ld	[$key+28],$t3
+	srl	$s3,13,$acc1
+	nop
+.Ldec_loop:
+	srl	$s2,5,$acc2			!
+	and	$acc1,2040,$acc1
+	ldx	[$tbl+$acc0],$acc0
+	sll	$s1,3,$acc3
+	and	$acc2,2040,$acc2
+	ldx	[$tbl+$acc1],$acc1
+	srl	$s1,21,$acc4
+	and	$acc3,2040,$acc3
+	ldx	[$tbl+$acc2],$acc2		!
+	srl	$s0,13,$acc5
+	and	$acc4,2040,$acc4
+	ldx	[$tbl+$acc3],$acc3
+	srl	$s3,5,$acc6
+	and	$acc5,2040,$acc5
+	ldx	[$tbl+$acc4],$acc4
+	fmovs	%f0,%f0
+	sll	$s2,3,$acc7			!
+	and	$acc6,2040,$acc6
+	ldx	[$tbl+$acc5],$acc5
+	srl	$s2,21,$acc8
+	and	$acc7,2040,$acc7
+	ldx	[$tbl+$acc6],$acc6
+	srl	$s1,13,$acc9
+	and	$acc8,2040,$acc8
+	ldx	[$tbl+$acc7],$acc7		!
+	srl	$s0,5,$acc10
+	and	$acc9,2040,$acc9
+	ldx	[$tbl+$acc8],$acc8
+	sll	$s3,3,$acc11
+	and	$acc10,2040,$acc10
+	ldx	[$tbl+$acc9],$acc9
+	fmovs	%f0,%f0
+	srl	$s3,21,$acc12			!
+	and	$acc11,2040,$acc11
+	ldx	[$tbl+$acc10],$acc10
+	srl	$s2,13,$acc13
+	and	$acc12,2040,$acc12
+	ldx	[$tbl+$acc11],$acc11
+	srl	$s1,5,$acc14
+	and	$acc13,2040,$acc13
+	ldx	[$tbl+$acc12],$acc12		!
+	sll	$s0,3,$acc15
+	and	$acc14,2040,$acc14
+	ldx	[$tbl+$acc13],$acc13
+	and	$acc15,2040,$acc15
+	add	$key,32,$key
+	ldx	[$tbl+$acc14],$acc14
+	fmovs	%f0,%f0
+	subcc	$rounds,1,$rounds		!
+	ldx	[$tbl+$acc15],$acc15
+	bz,a,pn	%icc,.Ldec_last
+	add	$tbl,2048,$rounds
+
+		srlx	$acc1,8,$acc1
+		xor	$acc0,$t0,$t0
+	ld	[$key+0],$s0
+	fmovs	%f0,%f0
+		srlx	$acc2,16,$acc2		!
+		xor	$acc1,$t0,$t0
+	ld	[$key+4],$s1
+		srlx	$acc3,24,$acc3
+		xor	$acc2,$t0,$t0
+	ld	[$key+8],$s2
+		srlx	$acc5,8,$acc5
+		xor	$acc3,$t0,$t0
+	ld	[$key+12],$s3			!
+		srlx	$acc6,16,$acc6
+		xor	$acc4,$t1,$t1
+	fmovs	%f0,%f0
+		srlx	$acc7,24,$acc7
+		xor	$acc5,$t1,$t1
+		srlx	$acc9,8,$acc9
+		xor	$acc6,$t1,$t1
+		srlx	$acc10,16,$acc10	!
+		xor	$acc7,$t1,$t1
+		srlx	$acc11,24,$acc11
+		xor	$acc8,$t2,$t2
+		srlx	$acc13,8,$acc13
+		xor	$acc9,$t2,$t2
+		srlx	$acc14,16,$acc14
+		xor	$acc10,$t2,$t2
+		srlx	$acc15,24,$acc15	!
+		xor	$acc11,$t2,$t2
+		xor	$acc12,$acc14,$acc14
+		xor	$acc13,$t3,$t3
+	srl	$t0,21,$acc0
+		xor	$acc14,$t3,$t3
+		xor	$acc15,$t3,$t3
+	srl	$t3,13,$acc1
+
+	and	$acc0,2040,$acc0		!
+	srl	$t2,5,$acc2
+	and	$acc1,2040,$acc1
+	ldx	[$tbl+$acc0],$acc0
+	sll	$t1,3,$acc3
+	and	$acc2,2040,$acc2
+	ldx	[$tbl+$acc1],$acc1
+	fmovs	%f0,%f0
+	srl	$t1,21,$acc4			!
+	and	$acc3,2040,$acc3
+	ldx	[$tbl+$acc2],$acc2
+	srl	$t0,13,$acc5
+	and	$acc4,2040,$acc4
+	ldx	[$tbl+$acc3],$acc3
+	srl	$t3,5,$acc6
+	and	$acc5,2040,$acc5
+	ldx	[$tbl+$acc4],$acc4		!
+	sll	$t2,3,$acc7
+	and	$acc6,2040,$acc6
+	ldx	[$tbl+$acc5],$acc5
+	srl	$t2,21,$acc8
+	and	$acc7,2040,$acc7
+	ldx	[$tbl+$acc6],$acc6
+	fmovs	%f0,%f0
+	srl	$t1,13,$acc9			!
+	and	$acc8,2040,$acc8
+	ldx	[$tbl+$acc7],$acc7
+	srl	$t0,5,$acc10
+	and	$acc9,2040,$acc9
+	ldx	[$tbl+$acc8],$acc8
+	sll	$t3,3,$acc11
+	and	$acc10,2040,$acc10
+	ldx	[$tbl+$acc9],$acc9		!
+	srl	$t3,21,$acc12
+	and	$acc11,2040,$acc11
+	ldx	[$tbl+$acc10],$acc10
+	srl	$t2,13,$acc13
+	and	$acc12,2040,$acc12
+	ldx	[$tbl+$acc11],$acc11
+	fmovs	%f0,%f0
+	srl	$t1,5,$acc14			!
+	and	$acc13,2040,$acc13
+	ldx	[$tbl+$acc12],$acc12
+	sll	$t0,3,$acc15
+	and	$acc14,2040,$acc14
+	ldx	[$tbl+$acc13],$acc13
+		srlx	$acc1,8,$acc1
+	and	$acc15,2040,$acc15
+	ldx	[$tbl+$acc14],$acc14		!
+
+		srlx	$acc2,16,$acc2
+		xor	$acc0,$s0,$s0
+	ldx	[$tbl+$acc15],$acc15
+		srlx	$acc3,24,$acc3
+		xor	$acc1,$s0,$s0
+	ld	[$key+16],$t0
+	fmovs	%f0,%f0
+		srlx	$acc5,8,$acc5		!
+		xor	$acc2,$s0,$s0
+	ld	[$key+20],$t1
+		srlx	$acc6,16,$acc6
+		xor	$acc3,$s0,$s0
+	ld	[$key+24],$t2
+		srlx	$acc7,24,$acc7
+		xor	$acc4,$s1,$s1
+	ld	[$key+28],$t3			!
+		srlx	$acc9,8,$acc9
+		xor	$acc5,$s1,$s1
+	ldx	[$tbl+2048+0],%g0		! prefetch td4
+		srlx	$acc10,16,$acc10
+		xor	$acc6,$s1,$s1
+	ldx	[$tbl+2048+32],%g0		! prefetch td4
+		srlx	$acc11,24,$acc11
+		xor	$acc7,$s1,$s1
+	ldx	[$tbl+2048+64],%g0		! prefetch td4
+		srlx	$acc13,8,$acc13
+		xor	$acc8,$s2,$s2
+	ldx	[$tbl+2048+96],%g0		! prefetch td4
+		srlx	$acc14,16,$acc14	!
+		xor	$acc9,$s2,$s2
+	ldx	[$tbl+2048+128],%g0		! prefetch td4
+		srlx	$acc15,24,$acc15
+		xor	$acc10,$s2,$s2
+	ldx	[$tbl+2048+160],%g0		! prefetch td4
+	srl	$s0,21,$acc0
+		xor	$acc11,$s2,$s2
+	ldx	[$tbl+2048+192],%g0		! prefetch td4
+		xor	$acc12,$acc14,$acc14
+		xor	$acc13,$s3,$s3
+	ldx	[$tbl+2048+224],%g0		! prefetch td4
+	and	$acc0,2040,$acc0		!
+		xor	$acc14,$s3,$s3
+		xor	$acc15,$s3,$s3
+	ba	.Ldec_loop
+	srl	$s3,13,$acc1
+
+.align	32
+.Ldec_last:
+		srlx	$acc1,8,$acc1		!
+		xor	$acc0,$t0,$t0
+	ld	[$key+0],$s0
+		srlx	$acc2,16,$acc2
+		xor	$acc1,$t0,$t0
+	ld	[$key+4],$s1
+		srlx	$acc3,24,$acc3
+		xor	$acc2,$t0,$t0
+	ld	[$key+8],$s2			!
+		srlx	$acc5,8,$acc5
+		xor	$acc3,$t0,$t0
+	ld	[$key+12],$s3
+		srlx	$acc6,16,$acc6
+		xor	$acc4,$t1,$t1
+		srlx	$acc7,24,$acc7
+		xor	$acc5,$t1,$t1
+		srlx	$acc9,8,$acc9		!
+		xor	$acc6,$t1,$t1
+		srlx	$acc10,16,$acc10
+		xor	$acc7,$t1,$t1
+		srlx	$acc11,24,$acc11
+		xor	$acc8,$t2,$t2
+		srlx	$acc13,8,$acc13
+		xor	$acc9,$t2,$t2
+		srlx	$acc14,16,$acc14	!
+		xor	$acc10,$t2,$t2
+		srlx	$acc15,24,$acc15
+		xor	$acc11,$t2,$t2
+		xor	$acc12,$acc14,$acc14
+		xor	$acc13,$t3,$t3
+	srl	$t0,24,$acc0
+		xor	$acc14,$t3,$t3
+		xor	$acc15,$t3,$t3		!
+	srl	$t3,16,$acc1
+
+	srl	$t2,8,$acc2
+	and	$acc1,255,$acc1
+	ldub	[$rounds+$acc0],$acc0
+	srl	$t1,24,$acc4
+	and	$acc2,255,$acc2
+	ldub	[$rounds+$acc1],$acc1
+	srl	$t0,16,$acc5			!
+	and	$t1,255,$acc3
+	ldub	[$rounds+$acc2],$acc2
+	ldub	[$rounds+$acc3],$acc3
+	srl	$t3,8,$acc6
+	and	$acc5,255,$acc5
+	ldub	[$rounds+$acc4],$acc4
+	fmovs	%f0,%f0
+	srl	$t2,24,$acc8			!
+	and	$acc6,255,$acc6
+	ldub	[$rounds+$acc5],$acc5
+	srl	$t1,16,$acc9
+	and	$t2,255,$acc7
+	ldub	[$rounds+$acc6],$acc6
+	ldub	[$rounds+$acc7],$acc7
+	fmovs	%f0,%f0
+	srl	$t0,8,$acc10			!
+	and	$acc9,255,$acc9
+	ldub	[$rounds+$acc8],$acc8
+	srl	$t3,24,$acc12
+	and	$acc10,255,$acc10
+	ldub	[$rounds+$acc9],$acc9
+	srl	$t2,16,$acc13
+	and	$t3,255,$acc11
+	ldub	[$rounds+$acc10],$acc10		!
+	srl	$t1,8,$acc14
+	and	$acc13,255,$acc13
+	ldub	[$rounds+$acc11],$acc11
+	ldub	[$rounds+$acc12],$acc12
+	and	$acc14,255,$acc14
+	ldub	[$rounds+$acc13],$acc13
+	and	$t0,255,$acc15
+	ldub	[$rounds+$acc14],$acc14		!
+
+		sll	$acc0,24,$acc0
+		xor	$acc3,$s0,$s0
+	ldub	[$rounds+$acc15],$acc15
+		sll	$acc1,16,$acc1
+		xor	$acc0,$s0,$s0
+	ldx	[%sp+$bias+$frame+0],%i7	! restore return address
+	fmovs	%f0,%f0
+		sll	$acc2,8,$acc2		!
+		xor	$acc1,$s0,$s0
+		sll	$acc4,24,$acc4
+		xor	$acc2,$s0,$s0
+		sll	$acc5,16,$acc5
+		xor	$acc7,$s1,$s1
+		sll	$acc6,8,$acc6
+		xor	$acc4,$s1,$s1
+		sll	$acc8,24,$acc8		!
+		xor	$acc5,$s1,$s1
+		sll	$acc9,16,$acc9
+		xor	$acc11,$s2,$s2
+		sll	$acc10,8,$acc10
+		xor	$acc6,$s1,$s1
+		sll	$acc12,24,$acc12
+		xor	$acc8,$s2,$s2
+		sll	$acc13,16,$acc13	!
+		xor	$acc9,$s2,$s2
+		sll	$acc14,8,$acc14
+		xor	$acc10,$s2,$s2
+		xor	$acc12,$acc14,$acc14
+		xor	$acc13,$s3,$s3
+		xor	$acc14,$s3,$s3
+		xor	$acc15,$s3,$s3
+
+	ret
+	restore
+.type	_sparcv9_AES_decrypt,#function
+.size	_sparcv9_AES_decrypt,(.-_sparcv9_AES_decrypt)
+
+.align	32
+.globl	AES_decrypt
+AES_decrypt:
+	or	%o0,%o1,%g1
+	andcc	%g1,3,%g0
+	bnz,pn	%xcc,.Lunaligned_dec
+	save	%sp,-$frame,%sp
+
+	ld	[%i0+0],%o0
+	ld	[%i0+4],%o1
+	ld	[%i0+8],%o2
+	ld	[%i0+12],%o3
+
+1:	call	.+8
+	add	%o7,AES_Td-1b,%o4
+	call	_sparcv9_AES_decrypt
+	mov	%i2,%o5
+
+	st	%o0,[%i1+0]
+	st	%o1,[%i1+4]
+	st	%o2,[%i1+8]
+	st	%o3,[%i1+12]
+
+	ret
+	restore
+
+.align	32
+.Lunaligned_dec:
+	ldub	[%i0+0],%l0
+	ldub	[%i0+1],%l1
+	ldub	[%i0+2],%l2
+
+	sll	%l0,24,%l0
+	ldub	[%i0+3],%l3
+	sll	%l1,16,%l1
+	ldub	[%i0+4],%l4
+	sll	%l2,8,%l2
+	or	%l1,%l0,%l0
+	ldub	[%i0+5],%l5
+	sll	%l4,24,%l4
+	or	%l3,%l2,%l2
+	ldub	[%i0+6],%l6
+	sll	%l5,16,%l5
+	or	%l0,%l2,%o0
+	ldub	[%i0+7],%l7
+
+	sll	%l6,8,%l6
+	or	%l5,%l4,%l4
+	ldub	[%i0+8],%l0
+	or	%l7,%l6,%l6
+	ldub	[%i0+9],%l1
+	or	%l4,%l6,%o1
+	ldub	[%i0+10],%l2
+
+	sll	%l0,24,%l0
+	ldub	[%i0+11],%l3
+	sll	%l1,16,%l1
+	ldub	[%i0+12],%l4
+	sll	%l2,8,%l2
+	or	%l1,%l0,%l0
+	ldub	[%i0+13],%l5
+	sll	%l4,24,%l4
+	or	%l3,%l2,%l2
+	ldub	[%i0+14],%l6
+	sll	%l5,16,%l5
+	or	%l0,%l2,%o2
+	ldub	[%i0+15],%l7
+
+	sll	%l6,8,%l6
+	or	%l5,%l4,%l4
+	or	%l7,%l6,%l6
+	or	%l4,%l6,%o3
+
+1:	call	.+8
+	add	%o7,AES_Td-1b,%o4
+	call	_sparcv9_AES_decrypt
+	mov	%i2,%o5
+
+	srl	%o0,24,%l0
+	srl	%o0,16,%l1
+	stb	%l0,[%i1+0]
+	srl	%o0,8,%l2
+	stb	%l1,[%i1+1]
+	stb	%l2,[%i1+2]
+	srl	%o1,24,%l4
+	stb	%o0,[%i1+3]
+
+	srl	%o1,16,%l5
+	stb	%l4,[%i1+4]
+	srl	%o1,8,%l6
+	stb	%l5,[%i1+5]
+	stb	%l6,[%i1+6]
+	srl	%o2,24,%l0
+	stb	%o1,[%i1+7]
+
+	srl	%o2,16,%l1
+	stb	%l0,[%i1+8]
+	srl	%o2,8,%l2
+	stb	%l1,[%i1+9]
+	stb	%l2,[%i1+10]
+	srl	%o3,24,%l4
+	stb	%o2,[%i1+11]
+
+	srl	%o3,16,%l5
+	stb	%l4,[%i1+12]
+	srl	%o3,8,%l6
+	stb	%l5,[%i1+13]
+	stb	%l6,[%i1+14]
+	stb	%o3,[%i1+15]
+
+	ret
+	restore
+.type	AES_decrypt,#function
+.size	AES_decrypt,(.-AES_decrypt)
+___
+
+# fmovs instructions substituting for FP nops were originally added
+# to meet specific instruction alignment requirements to maximize ILP.
+# As UltraSPARC T1, a.k.a. Niagara, has shared FPU, FP nops can have
+# undesired effect, so just omit them and sacrifice some portion of
+# percent in performance...
+$code =~ s/fmovs.*$//gem;
+
+print $code;
diff --git a/openssl/crypto/aes/asm/aes-x86_64.pl b/openssl/crypto/aes/asm/aes-x86_64.pl
new file mode 100755
index 00000000..a545e892
--- /dev/null
+++ b/openssl/crypto/aes/asm/aes-x86_64.pl
@@ -0,0 +1,2809 @@
+#!/usr/bin/env perl
+#
+# ====================================================================
+# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+#
+# Version 2.1.
+#
+# aes-*-cbc benchmarks are improved by >70% [compared to gcc 3.3.2 on
+# Opteron 240 CPU] plus all the bells-n-whistles from 32-bit version
+# [you'll notice a lot of resemblance], such as compressed S-boxes
+# in little-endian byte order, prefetch of these tables in CBC mode,
+# as well as avoiding L1 cache aliasing between stack frame and key
+# schedule and already mentioned tables, compressed Td4...
+#
+# Performance in number of cycles per processed byte for 128-bit key:
+#
+#		ECB encrypt	ECB decrypt	CBC large chunk
+# AMD64		33		41		13.0
+# EM64T		38		59		18.6(*)
+# Core 2	30		43		14.5(*)
+#
+# (*) with hyper-threading off
+
+$flavour = shift;
+$output  = shift;
+if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
+
+$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
+
+$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
+( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or
+( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or
+die "can't locate x86_64-xlate.pl";
+
+open STDOUT,"| $^X $xlate $flavour $output";
+
+$verticalspin=1;	# unlike 32-bit version $verticalspin performs
+			# ~15% better on both AMD and Intel cores
+$speed_limit=512;	# see aes-586.pl for details
+
+$code=".text\n";
+
+$s0="%eax";
+$s1="%ebx";
+$s2="%ecx";
+$s3="%edx";
+$acc0="%esi";	$mask80="%rsi";
+$acc1="%edi";	$maskfe="%rdi";
+$acc2="%ebp";	$mask1b="%rbp";
+$inp="%r8";
+$out="%r9";
+$t0="%r10d";
+$t1="%r11d";
+$t2="%r12d";
+$rnds="%r13d";
+$sbox="%r14";
+$key="%r15";
+
+sub hi() { my $r=shift;	$r =~ s/%[er]([a-d])x/%\1h/;	$r; }
+sub lo() { my $r=shift;	$r =~ s/%[er]([a-d])x/%\1l/;
+			$r =~ s/%[er]([sd]i)/%\1l/;
+			$r =~ s/%(r[0-9]+)[d]?/%\1b/;	$r; }
+sub LO() { my $r=shift; $r =~ s/%r([a-z]+)/%e\1/;
+			$r =~ s/%r([0-9]+)/%r\1d/;	$r; }
+sub _data_word()
+{ my $i;
+    while(defined($i=shift)) { $code.=sprintf".long\t0x%08x,0x%08x\n",$i,$i; }
+}
+sub data_word()
+{ my $i;
+  my $last=pop(@_);
+    $code.=".long\t";
+    while(defined($i=shift)) { $code.=sprintf"0x%08x,",$i; }
+    $code.=sprintf"0x%08x\n",$last;
+}
+
+sub data_byte()
+{ my $i;
+  my $last=pop(@_);
+    $code.=".byte\t";
+    while(defined($i=shift)) { $code.=sprintf"0x%02x,",$i&0xff; }
+    $code.=sprintf"0x%02x\n",$last&0xff;
+}
+
+sub encvert()
+{ my $t3="%r8d";	# zaps $inp!
+
+$code.=<<___;
+	# favor 3-way issue Opteron pipeline...
+	movzb	`&lo("$s0")`,$acc0
+	movzb	`&lo("$s1")`,$acc1
+	movzb	`&lo("$s2")`,$acc2
+	mov	0($sbox,$acc0,8),$t0
+	mov	0($sbox,$acc1,8),$t1
+	mov	0($sbox,$acc2,8),$t2
+
+	movzb	`&hi("$s1")`,$acc0
+	movzb	`&hi("$s2")`,$acc1
+	movzb	`&lo("$s3")`,$acc2
+	xor	3($sbox,$acc0,8),$t0
+	xor	3($sbox,$acc1,8),$t1
+	mov	0($sbox,$acc2,8),$t3
+
+	movzb	`&hi("$s3")`,$acc0
+	shr	\$16,$s2
+	movzb	`&hi("$s0")`,$acc2
+	xor	3($sbox,$acc0,8),$t2
+	shr	\$16,$s3
+	xor	3($sbox,$acc2,8),$t3
+
+	shr	\$16,$s1
+	lea	16($key),$key
+	shr	\$16,$s0
+
+	movzb	`&lo("$s2")`,$acc0
+	movzb	`&lo("$s3")`,$acc1
+	movzb	`&lo("$s0")`,$acc2
+	xor	2($sbox,$acc0,8),$t0
+	xor	2($sbox,$acc1,8),$t1
+	xor	2($sbox,$acc2,8),$t2
+
+	movzb	`&hi("$s3")`,$acc0
+	movzb	`&hi("$s0")`,$acc1
+	movzb	`&lo("$s1")`,$acc2
+	xor	1($sbox,$acc0,8),$t0
+	xor	1($sbox,$acc1,8),$t1
+	xor	2($sbox,$acc2,8),$t3
+
+	mov	12($key),$s3
+	movzb	`&hi("$s1")`,$acc1
+	movzb	`&hi("$s2")`,$acc2
+	mov	0($key),$s0
+	xor	1($sbox,$acc1,8),$t2
+	xor	1($sbox,$acc2,8),$t3
+
+	mov	4($key),$s1
+	mov	8($key),$s2
+	xor	$t0,$s0
+	xor	$t1,$s1
+	xor	$t2,$s2
+	xor	$t3,$s3
+___
+}
+
+sub enclastvert()
+{ my $t3="%r8d";	# zaps $inp!
+
+$code.=<<___;
+	movzb	`&lo("$s0")`,$acc0
+	movzb	`&lo("$s1")`,$acc1
+	movzb	`&lo("$s2")`,$acc2
+	movzb	2($sbox,$acc0,8),$t0
+	movzb	2($sbox,$acc1,8),$t1
+	movzb	2($sbox,$acc2,8),$t2
+
+	movzb	`&lo("$s3")`,$acc0
+	movzb	`&hi("$s1")`,$acc1
+	movzb	`&hi("$s2")`,$acc2
+	movzb	2($sbox,$acc0,8),$t3
+	mov	0($sbox,$acc1,8),$acc1	#$t0
+	mov	0($sbox,$acc2,8),$acc2	#$t1
+
+	and	\$0x0000ff00,$acc1
+	and	\$0x0000ff00,$acc2
+
+	xor	$acc1,$t0
+	xor	$acc2,$t1
+	shr	\$16,$s2
+
+	movzb	`&hi("$s3")`,$acc0
+	movzb	`&hi("$s0")`,$acc1
+	shr	\$16,$s3
+	mov	0($sbox,$acc0,8),$acc0	#$t2
+	mov	0($sbox,$acc1,8),$acc1	#$t3
+
+	and	\$0x0000ff00,$acc0
+	and	\$0x0000ff00,$acc1
+	shr	\$16,$s1
+	xor	$acc0,$t2
+	xor	$acc1,$t3
+	shr	\$16,$s0
+
+	movzb	`&lo("$s2")`,$acc0
+	movzb	`&lo("$s3")`,$acc1
+	movzb	`&lo("$s0")`,$acc2
+	mov	0($sbox,$acc0,8),$acc0	#$t0
+	mov	0($sbox,$acc1,8),$acc1	#$t1
+	mov	0($sbox,$acc2,8),$acc2	#$t2
+
+	and	\$0x00ff0000,$acc0
+	and	\$0x00ff0000,$acc1
+	and	\$0x00ff0000,$acc2
+
+	xor	$acc0,$t0
+	xor	$acc1,$t1
+	xor	$acc2,$t2
+
+	movzb	`&lo("$s1")`,$acc0
+	movzb	`&hi("$s3")`,$acc1
+	movzb	`&hi("$s0")`,$acc2
+	mov	0($sbox,$acc0,8),$acc0	#$t3
+	mov	2($sbox,$acc1,8),$acc1	#$t0
+	mov	2($sbox,$acc2,8),$acc2	#$t1
+
+	and	\$0x00ff0000,$acc0
+	and	\$0xff000000,$acc1
+	and	\$0xff000000,$acc2
+
+	xor	$acc0,$t3
+	xor	$acc1,$t0
+	xor	$acc2,$t1
+
+	movzb	`&hi("$s1")`,$acc0
+	movzb	`&hi("$s2")`,$acc1
+	mov	16+12($key),$s3
+	mov	2($sbox,$acc0,8),$acc0	#$t2
+	mov	2($sbox,$acc1,8),$acc1	#$t3
+	mov	16+0($key),$s0
+
+	and	\$0xff000000,$acc0
+	and	\$0xff000000,$acc1
+
+	xor	$acc0,$t2
+	xor	$acc1,$t3
+
+	mov	16+4($key),$s1
+	mov	16+8($key),$s2
+	xor	$t0,$s0
+	xor	$t1,$s1
+	xor	$t2,$s2
+	xor	$t3,$s3
+___
+}
+
+sub encstep()
+{ my ($i,@s) = @_;
+  my $tmp0=$acc0;
+  my $tmp1=$acc1;
+  my $tmp2=$acc2;
+  my $out=($t0,$t1,$t2,$s[0])[$i];
+
+	if ($i==3) {
+		$tmp0=$s[1];
+		$tmp1=$s[2];
+		$tmp2=$s[3];
+	}
+	$code.="	movzb	".&lo($s[0]).",$out\n";
+	$code.="	mov	$s[2],$tmp1\n"		if ($i!=3);
+	$code.="	lea	16($key),$key\n"	if ($i==0);
+
+	$code.="	movzb	".&hi($s[1]).",$tmp0\n";
+	$code.="	mov	0($sbox,$out,8),$out\n";
+
+	$code.="	shr	\$16,$tmp1\n";
+	$code.="	mov	$s[3],$tmp2\n"		if ($i!=3);
+	$code.="	xor	3($sbox,$tmp0,8),$out\n";
+
+	$code.="	movzb	".&lo($tmp1).",$tmp1\n";
+	$code.="	shr	\$24,$tmp2\n";
+	$code.="	xor	4*$i($key),$out\n";
+
+	$code.="	xor	2($sbox,$tmp1,8),$out\n";
+	$code.="	xor	1($sbox,$tmp2,8),$out\n";
+
+	$code.="	mov	$t0,$s[1]\n"		if ($i==3);
+	$code.="	mov	$t1,$s[2]\n"		if ($i==3);
+	$code.="	mov	$t2,$s[3]\n"		if ($i==3);
+	$code.="\n";
+}
+
+sub enclast()
+{ my ($i,@s)=@_;
+  my $tmp0=$acc0;
+  my $tmp1=$acc1;
+  my $tmp2=$acc2;
+  my $out=($t0,$t1,$t2,$s[0])[$i];
+
+	if ($i==3) {
+		$tmp0=$s[1];
+		$tmp1=$s[2];
+		$tmp2=$s[3];
+	}
+	$code.="	movzb	".&lo($s[0]).",$out\n";
+	$code.="	mov	$s[2],$tmp1\n"		if ($i!=3);
+
+	$code.="	mov	2($sbox,$out,8),$out\n";
+	$code.="	shr	\$16,$tmp1\n";
+	$code.="	mov	$s[3],$tmp2\n"		if ($i!=3);
+
+	$code.="	and	\$0x000000ff,$out\n";
+	$code.="	movzb	".&hi($s[1]).",$tmp0\n";
+	$code.="	movzb	".&lo($tmp1).",$tmp1\n";
+	$code.="	shr	\$24,$tmp2\n";
+
+	$code.="	mov	0($sbox,$tmp0,8),$tmp0\n";
+	$code.="	mov	0($sbox,$tmp1,8),$tmp1\n";
+	$code.="	mov	2($sbox,$tmp2,8),$tmp2\n";
+
+	$code.="	and	\$0x0000ff00,$tmp0\n";
+	$code.="	and	\$0x00ff0000,$tmp1\n";
+	$code.="	and	\$0xff000000,$tmp2\n";
+
+	$code.="	xor	$tmp0,$out\n";
+	$code.="	mov	$t0,$s[1]\n"		if ($i==3);
+	$code.="	xor	$tmp1,$out\n";
+	$code.="	mov	$t1,$s[2]\n"		if ($i==3);
+	$code.="	xor	$tmp2,$out\n";
+	$code.="	mov	$t2,$s[3]\n"		if ($i==3);
+	$code.="\n";
+}
+
+$code.=<<___;
+.type	_x86_64_AES_encrypt,\@abi-omnipotent
+.align	16
+_x86_64_AES_encrypt:
+	xor	0($key),$s0			# xor with key
+	xor	4($key),$s1
+	xor	8($key),$s2
+	xor	12($key),$s3
+
+	mov	240($key),$rnds			# load key->rounds
+	sub	\$1,$rnds
+	jmp	.Lenc_loop
+.align	16
+.Lenc_loop:
+___
+	if ($verticalspin) { &encvert(); }
+	else {	&encstep(0,$s0,$s1,$s2,$s3);
+		&encstep(1,$s1,$s2,$s3,$s0);
+		&encstep(2,$s2,$s3,$s0,$s1);
+		&encstep(3,$s3,$s0,$s1,$s2);
+	}
+$code.=<<___;
+	sub	\$1,$rnds
+	jnz	.Lenc_loop
+___
+	if ($verticalspin) { &enclastvert(); }
+	else {	&enclast(0,$s0,$s1,$s2,$s3);
+		&enclast(1,$s1,$s2,$s3,$s0);
+		&enclast(2,$s2,$s3,$s0,$s1);
+		&enclast(3,$s3,$s0,$s1,$s2);
+		$code.=<<___;
+		xor	16+0($key),$s0		# xor with key
+		xor	16+4($key),$s1
+		xor	16+8($key),$s2
+		xor	16+12($key),$s3
+___
+	}
+$code.=<<___;
+	.byte	0xf3,0xc3			# rep ret
+.size	_x86_64_AES_encrypt,.-_x86_64_AES_encrypt
+___
+
+# it's possible to implement this by shifting tN by 8, filling least
+# significant byte with byte load and finally bswap-ing at the end,
+# but such partial register load kills Core 2...
+sub enccompactvert()
+{ my ($t3,$t4,$t5)=("%r8d","%r9d","%r13d");
+
+$code.=<<___;
+	movzb	`&lo("$s0")`,$t0
+	movzb	`&lo("$s1")`,$t1
+	movzb	`&lo("$s2")`,$t2
+	movzb	($sbox,$t0,1),$t0
+	movzb	($sbox,$t1,1),$t1
+	movzb	($sbox,$t2,1),$t2
+
+	movzb	`&lo("$s3")`,$t3
+	movzb	`&hi("$s1")`,$acc0
+	movzb	`&hi("$s2")`,$acc1
+	movzb	($sbox,$t3,1),$t3
+	movzb	($sbox,$acc0,1),$t4	#$t0
+	movzb	($sbox,$acc1,1),$t5	#$t1
+
+	movzb	`&hi("$s3")`,$acc2
+	movzb	`&hi("$s0")`,$acc0
+	shr	\$16,$s2
+	movzb	($sbox,$acc2,1),$acc2	#$t2
+	movzb	($sbox,$acc0,1),$acc0	#$t3
+	shr	\$16,$s3
+
+	movzb	`&lo("$s2")`,$acc1
+	shl	\$8,$t4
+	shl	\$8,$t5
+	movzb	($sbox,$acc1,1),$acc1	#$t0
+	xor	$t4,$t0
+	xor	$t5,$t1
+
+	movzb	`&lo("$s3")`,$t4
+	shr	\$16,$s0
+	shr	\$16,$s1
+	movzb	`&lo("$s0")`,$t5
+	shl	\$8,$acc2
+	shl	\$8,$acc0
+	movzb	($sbox,$t4,1),$t4	#$t1
+	movzb	($sbox,$t5,1),$t5	#$t2
+	xor	$acc2,$t2
+	xor	$acc0,$t3
+
+	movzb	`&lo("$s1")`,$acc2
+	movzb	`&hi("$s3")`,$acc0
+	shl	\$16,$acc1
+	movzb	($sbox,$acc2,1),$acc2	#$t3
+	movzb	($sbox,$acc0,1),$acc0	#$t0
+	xor	$acc1,$t0
+
+	movzb	`&hi("$s0")`,$acc1
+	shr	\$8,$s2
+	shr	\$8,$s1
+	movzb	($sbox,$acc1,1),$acc1	#$t1
+	movzb	($sbox,$s2,1),$s3	#$t3
+	movzb	($sbox,$s1,1),$s2	#$t2
+	shl	\$16,$t4
+	shl	\$16,$t5
+	shl	\$16,$acc2
+	xor	$t4,$t1
+	xor	$t5,$t2
+	xor	$acc2,$t3
+
+	shl	\$24,$acc0
+	shl	\$24,$acc1
+	shl	\$24,$s3
+	xor	$acc0,$t0
+	shl	\$24,$s2
+	xor	$acc1,$t1
+	mov	$t0,$s0
+	mov	$t1,$s1
+	xor	$t2,$s2
+	xor	$t3,$s3
+___
+}
+
+sub enctransform_ref()
+{ my $sn = shift;
+  my ($acc,$r2,$tmp)=("%r8d","%r9d","%r13d");
+
+$code.=<<___;
+	mov	$sn,$acc
+	and	\$0x80808080,$acc
+	mov	$acc,$tmp
+	shr	\$7,$tmp
+	lea	($sn,$sn),$r2
+	sub	$tmp,$acc
+	and	\$0xfefefefe,$r2
+	and	\$0x1b1b1b1b,$acc
+	mov	$sn,$tmp
+	xor	$acc,$r2
+
+	xor	$r2,$sn
+	rol	\$24,$sn
+	xor	$r2,$sn
+	ror	\$16,$tmp
+	xor	$tmp,$sn
+	ror	\$8,$tmp
+	xor	$tmp,$sn
+___
+}
+
+# unlike decrypt case it does not pay off to parallelize enctransform
+sub enctransform()
+{ my ($t3,$r20,$r21)=($acc2,"%r8d","%r9d");
+
+$code.=<<___;
+	mov	$s0,$acc0
+	mov	$s1,$acc1
+	and	\$0x80808080,$acc0
+	and	\$0x80808080,$acc1
+	mov	$acc0,$t0
+	mov	$acc1,$t1
+	shr	\$7,$t0
+	lea	($s0,$s0),$r20
+	shr	\$7,$t1
+	lea	($s1,$s1),$r21
+	sub	$t0,$acc0
+	sub	$t1,$acc1
+	and	\$0xfefefefe,$r20
+	and	\$0xfefefefe,$r21
+	and	\$0x1b1b1b1b,$acc0
+	and	\$0x1b1b1b1b,$acc1
+	mov	$s0,$t0
+	mov	$s1,$t1
+	xor	$acc0,$r20
+	xor	$acc1,$r21
+
+	xor	$r20,$s0
+	xor	$r21,$s1
+	 mov	$s2,$acc0
+	 mov	$s3,$acc1
+	rol	\$24,$s0
+	rol	\$24,$s1
+	 and	\$0x80808080,$acc0
+	 and	\$0x80808080,$acc1
+	xor	$r20,$s0
+	xor	$r21,$s1
+	 mov	$acc0,$t2
+	 mov	$acc1,$t3
+	ror	\$16,$t0
+	ror	\$16,$t1
+	 shr	\$7,$t2
+	 lea	($s2,$s2),$r20
+	xor	$t0,$s0
+	xor	$t1,$s1
+	 shr	\$7,$t3
+	 lea	($s3,$s3),$r21
+	ror	\$8,$t0
+	ror	\$8,$t1
+	 sub	$t2,$acc0
+	 sub	$t3,$acc1
+	xor	$t0,$s0
+	xor	$t1,$s1
+
+	and	\$0xfefefefe,$r20
+	and	\$0xfefefefe,$r21
+	and	\$0x1b1b1b1b,$acc0
+	and	\$0x1b1b1b1b,$acc1
+	mov	$s2,$t2
+	mov	$s3,$t3
+	xor	$acc0,$r20
+	xor	$acc1,$r21
+
+	xor	$r20,$s2
+	xor	$r21,$s3
+	rol	\$24,$s2
+	rol	\$24,$s3
+	xor	$r20,$s2
+	xor	$r21,$s3
+	mov	0($sbox),$acc0			# prefetch Te4
+	ror	\$16,$t2
+	ror	\$16,$t3
+	mov	64($sbox),$acc1
+	xor	$t2,$s2
+	xor	$t3,$s3
+	mov	128($sbox),$r20
+	ror	\$8,$t2
+	ror	\$8,$t3
+	mov	192($sbox),$r21
+	xor	$t2,$s2
+	xor	$t3,$s3
+___
+}
+
+$code.=<<___;
+.type	_x86_64_AES_encrypt_compact,\@abi-omnipotent
+.align	16
+_x86_64_AES_encrypt_compact:
+	lea	128($sbox),$inp			# size optimization
+	mov	0-128($inp),$acc1		# prefetch Te4
+	mov	32-128($inp),$acc2
+	mov	64-128($inp),$t0
+	mov	96-128($inp),$t1
+	mov	128-128($inp),$acc1
+	mov	160-128($inp),$acc2
+	mov	192-128($inp),$t0
+	mov	224-128($inp),$t1
+	jmp	.Lenc_loop_compact
+.align	16
+.Lenc_loop_compact:
+		xor	0($key),$s0		# xor with key
+		xor	4($key),$s1
+		xor	8($key),$s2
+		xor	12($key),$s3
+		lea	16($key),$key
+___
+		&enccompactvert();
+$code.=<<___;
+		cmp	16(%rsp),$key
+		je	.Lenc_compact_done
+___
+		&enctransform();
+$code.=<<___;
+	jmp	.Lenc_loop_compact
+.align	16
+.Lenc_compact_done:
+	xor	0($key),$s0
+	xor	4($key),$s1
+	xor	8($key),$s2
+	xor	12($key),$s3
+	.byte	0xf3,0xc3			# rep ret
+.size	_x86_64_AES_encrypt_compact,.-_x86_64_AES_encrypt_compact
+___
+
+# void AES_encrypt (const void *inp,void *out,const AES_KEY *key);
+$code.=<<___;
+.globl	AES_encrypt
+.type	AES_encrypt,\@function,3
+.align	16
+AES_encrypt:
+	push	%rbx
+	push	%rbp
+	push	%r12
+	push	%r13
+	push	%r14
+	push	%r15
+
+	# allocate frame "above" key schedule
+	mov	%rsp,%r10
+	lea	-63(%rdx),%rcx	# %rdx is key argument
+	and	\$-64,%rsp
+	sub	%rsp,%rcx
+	neg	%rcx
+	and	\$0x3c0,%rcx
+	sub	%rcx,%rsp
+	sub	\$32,%rsp
+
+	mov	%rsi,16(%rsp)	# save out
+	mov	%r10,24(%rsp)	# save real stack pointer
+.Lenc_prologue:
+
+	mov	%rdx,$key
+	mov	240($key),$rnds	# load rounds
+
+	mov	0(%rdi),$s0	# load input vector
+	mov	4(%rdi),$s1
+	mov	8(%rdi),$s2
+	mov	12(%rdi),$s3
+
+	shl	\$4,$rnds
+	lea	($key,$rnds),%rbp
+	mov	$key,(%rsp)	# key schedule
+	mov	%rbp,8(%rsp)	# end of key schedule
+
+	# pick Te4 copy which can't "overlap" with stack frame or key schedule
+	lea	.LAES_Te+2048(%rip),$sbox
+	lea	768(%rsp),%rbp
+	sub	$sbox,%rbp
+	and	\$0x300,%rbp
+	lea	($sbox,%rbp),$sbox
+
+	call	_x86_64_AES_encrypt_compact
+
+	mov	16(%rsp),$out	# restore out
+	mov	24(%rsp),%rsi	# restore saved stack pointer
+	mov	$s0,0($out)	# write output vector
+	mov	$s1,4($out)
+	mov	$s2,8($out)
+	mov	$s3,12($out)
+
+	mov	(%rsi),%r15
+	mov	8(%rsi),%r14
+	mov	16(%rsi),%r13
+	mov	24(%rsi),%r12
+	mov	32(%rsi),%rbp
+	mov	40(%rsi),%rbx
+	lea	48(%rsi),%rsp
+.Lenc_epilogue:
+	ret
+.size	AES_encrypt,.-AES_encrypt
+___
+
+#------------------------------------------------------------------#
+
+sub decvert()
+{ my $t3="%r8d";	# zaps $inp!
+
+$code.=<<___;
+	# favor 3-way issue Opteron pipeline...
+	movzb	`&lo("$s0")`,$acc0
+	movzb	`&lo("$s1")`,$acc1
+	movzb	`&lo("$s2")`,$acc2
+	mov	0($sbox,$acc0,8),$t0
+	mov	0($sbox,$acc1,8),$t1
+	mov	0($sbox,$acc2,8),$t2
+
+	movzb	`&hi("$s3")`,$acc0
+	movzb	`&hi("$s0")`,$acc1
+	movzb	`&lo("$s3")`,$acc2
+	xor	3($sbox,$acc0,8),$t0
+	xor	3($sbox,$acc1,8),$t1
+	mov	0($sbox,$acc2,8),$t3
+
+	movzb	`&hi("$s1")`,$acc0
+	shr	\$16,$s0
+	movzb	`&hi("$s2")`,$acc2
+	xor	3($sbox,$acc0,8),$t2
+	shr	\$16,$s3
+	xor	3($sbox,$acc2,8),$t3
+
+	shr	\$16,$s1
+	lea	16($key),$key
+	shr	\$16,$s2
+
+	movzb	`&lo("$s2")`,$acc0
+	movzb	`&lo("$s3")`,$acc1
+	movzb	`&lo("$s0")`,$acc2
+	xor	2($sbox,$acc0,8),$t0
+	xor	2($sbox,$acc1,8),$t1
+	xor	2($sbox,$acc2,8),$t2
+
+	movzb	`&hi("$s1")`,$acc0
+	movzb	`&hi("$s2")`,$acc1
+	movzb	`&lo("$s1")`,$acc2
+	xor	1($sbox,$acc0,8),$t0
+	xor	1($sbox,$acc1,8),$t1
+	xor	2($sbox,$acc2,8),$t3
+
+	movzb	`&hi("$s3")`,$acc0
+	mov	12($key),$s3
+	movzb	`&hi("$s0")`,$acc2
+	xor	1($sbox,$acc0,8),$t2
+	mov	0($key),$s0
+	xor	1($sbox,$acc2,8),$t3
+
+	xor	$t0,$s0
+	mov	4($key),$s1
+	mov	8($key),$s2
+	xor	$t2,$s2
+	xor	$t1,$s1
+	xor	$t3,$s3
+___
+}
+
+sub declastvert()
+{ my $t3="%r8d";	# zaps $inp!
+
+$code.=<<___;
+	lea	2048($sbox),$sbox	# size optimization
+	movzb	`&lo("$s0")`,$acc0
+	movzb	`&lo("$s1")`,$acc1
+	movzb	`&lo("$s2")`,$acc2
+	movzb	($sbox,$acc0,1),$t0
+	movzb	($sbox,$acc1,1),$t1
+	movzb	($sbox,$acc2,1),$t2
+
+	movzb	`&lo("$s3")`,$acc0
+	movzb	`&hi("$s3")`,$acc1
+	movzb	`&hi("$s0")`,$acc2
+	movzb	($sbox,$acc0,1),$t3
+	movzb	($sbox,$acc1,1),$acc1	#$t0
+	movzb	($sbox,$acc2,1),$acc2	#$t1
+
+	shl	\$8,$acc1
+	shl	\$8,$acc2
+
+	xor	$acc1,$t0
+	xor	$acc2,$t1
+	shr	\$16,$s3
+
+	movzb	`&hi("$s1")`,$acc0
+	movzb	`&hi("$s2")`,$acc1
+	shr	\$16,$s0
+	movzb	($sbox,$acc0,1),$acc0	#$t2
+	movzb	($sbox,$acc1,1),$acc1	#$t3
+
+	shl	\$8,$acc0
+	shl	\$8,$acc1
+	shr	\$16,$s1
+	xor	$acc0,$t2
+	xor	$acc1,$t3
+	shr	\$16,$s2
+
+	movzb	`&lo("$s2")`,$acc0
+	movzb	`&lo("$s3")`,$acc1
+	movzb	`&lo("$s0")`,$acc2
+	movzb	($sbox,$acc0,1),$acc0	#$t0
+	movzb	($sbox,$acc1,1),$acc1	#$t1
+	movzb	($sbox,$acc2,1),$acc2	#$t2
+
+	shl	\$16,$acc0
+	shl	\$16,$acc1
+	shl	\$16,$acc2
+
+	xor	$acc0,$t0
+	xor	$acc1,$t1
+	xor	$acc2,$t2
+
+	movzb	`&lo("$s1")`,$acc0
+	movzb	`&hi("$s1")`,$acc1
+	movzb	`&hi("$s2")`,$acc2
+	movzb	($sbox,$acc0,1),$acc0	#$t3
+	movzb	($sbox,$acc1,1),$acc1	#$t0
+	movzb	($sbox,$acc2,1),$acc2	#$t1
+
+	shl	\$16,$acc0
+	shl	\$24,$acc1
+	shl	\$24,$acc2
+
+	xor	$acc0,$t3
+	xor	$acc1,$t0
+	xor	$acc2,$t1
+
+	movzb	`&hi("$s3")`,$acc0
+	movzb	`&hi("$s0")`,$acc1
+	mov	16+12($key),$s3
+	movzb	($sbox,$acc0,1),$acc0	#$t2
+	movzb	($sbox,$acc1,1),$acc1	#$t3
+	mov	16+0($key),$s0
+
+	shl	\$24,$acc0
+	shl	\$24,$acc1
+
+	xor	$acc0,$t2
+	xor	$acc1,$t3
+
+	mov	16+4($key),$s1
+	mov	16+8($key),$s2
+	lea	-2048($sbox),$sbox
+	xor	$t0,$s0
+	xor	$t1,$s1
+	xor	$t2,$s2
+	xor	$t3,$s3
+___
+}
+
+sub decstep()
+{ my ($i,@s) = @_;
+  my $tmp0=$acc0;
+  my $tmp1=$acc1;
+  my $tmp2=$acc2;
+  my $out=($t0,$t1,$t2,$s[0])[$i];
+
+	$code.="	mov	$s[0],$out\n"		if ($i!=3);
+			$tmp1=$s[2]			if ($i==3);
+	$code.="	mov	$s[2],$tmp1\n"		if ($i!=3);
+	$code.="	and	\$0xFF,$out\n";
+
+	$code.="	mov	0($sbox,$out,8),$out\n";
+	$code.="	shr	\$16,$tmp1\n";
+			$tmp2=$s[3]			if ($i==3);
+	$code.="	mov	$s[3],$tmp2\n"		if ($i!=3);
+
+			$tmp0=$s[1]			if ($i==3);
+	$code.="	movzb	".&hi($s[1]).",$tmp0\n";
+	$code.="	and	\$0xFF,$tmp1\n";
+	$code.="	shr	\$24,$tmp2\n";
+
+	$code.="	xor	3($sbox,$tmp0,8),$out\n";
+	$code.="	xor	2($sbox,$tmp1,8),$out\n";
+	$code.="	xor	1($sbox,$tmp2,8),$out\n";
+
+	$code.="	mov	$t2,$s[1]\n"		if ($i==3);
+	$code.="	mov	$t1,$s[2]\n"		if ($i==3);
+	$code.="	mov	$t0,$s[3]\n"		if ($i==3);
+	$code.="\n";
+}
+
+sub declast()
+{ my ($i,@s)=@_;
+  my $tmp0=$acc0;
+  my $tmp1=$acc1;
+  my $tmp2=$acc2;
+  my $out=($t0,$t1,$t2,$s[0])[$i];
+
+	$code.="	mov	$s[0],$out\n"		if ($i!=3);
+			$tmp1=$s[2]			if ($i==3);
+	$code.="	mov	$s[2],$tmp1\n"		if ($i!=3);
+	$code.="	and	\$0xFF,$out\n";
+
+	$code.="	movzb	2048($sbox,$out,1),$out\n";
+	$code.="	shr	\$16,$tmp1\n";
+			$tmp2=$s[3]			if ($i==3);
+	$code.="	mov	$s[3],$tmp2\n"		if ($i!=3);
+
+			$tmp0=$s[1]			if ($i==3);
+	$code.="	movzb	".&hi($s[1]).",$tmp0\n";
+	$code.="	and	\$0xFF,$tmp1\n";
+	$code.="	shr	\$24,$tmp2\n";
+
+	$code.="	movzb	2048($sbox,$tmp0,1),$tmp0\n";
+	$code.="	movzb	2048($sbox,$tmp1,1),$tmp1\n";
+	$code.="	movzb	2048($sbox,$tmp2,1),$tmp2\n";
+
+	$code.="	shl	\$8,$tmp0\n";
+	$code.="	shl	\$16,$tmp1\n";
+	$code.="	shl	\$24,$tmp2\n";
+
+	$code.="	xor	$tmp0,$out\n";
+	$code.="	mov	$t2,$s[1]\n"		if ($i==3);
+	$code.="	xor	$tmp1,$out\n";
+	$code.="	mov	$t1,$s[2]\n"		if ($i==3);
+	$code.="	xor	$tmp2,$out\n";
+	$code.="	mov	$t0,$s[3]\n"		if ($i==3);
+	$code.="\n";
+}
+
+$code.=<<___;
+.type	_x86_64_AES_decrypt,\@abi-omnipotent
+.align	16
+_x86_64_AES_decrypt:
+	xor	0($key),$s0			# xor with key
+	xor	4($key),$s1
+	xor	8($key),$s2
+	xor	12($key),$s3
+
+	mov	240($key),$rnds			# load key->rounds
+	sub	\$1,$rnds
+	jmp	.Ldec_loop
+.align	16
+.Ldec_loop:
+___
+	if ($verticalspin) { &decvert(); }
+	else {	&decstep(0,$s0,$s3,$s2,$s1);
+		&decstep(1,$s1,$s0,$s3,$s2);
+		&decstep(2,$s2,$s1,$s0,$s3);
+		&decstep(3,$s3,$s2,$s1,$s0);
+		$code.=<<___;
+		lea	16($key),$key
+		xor	0($key),$s0			# xor with key
+		xor	4($key),$s1
+		xor	8($key),$s2
+		xor	12($key),$s3
+___
+	}
+$code.=<<___;
+	sub	\$1,$rnds
+	jnz	.Ldec_loop
+___
+	if ($verticalspin) { &declastvert(); }
+	else {	&declast(0,$s0,$s3,$s2,$s1);
+		&declast(1,$s1,$s0,$s3,$s2);
+		&declast(2,$s2,$s1,$s0,$s3);
+		&declast(3,$s3,$s2,$s1,$s0);
+		$code.=<<___;
+		xor	16+0($key),$s0			# xor with key
+		xor	16+4($key),$s1
+		xor	16+8($key),$s2
+		xor	16+12($key),$s3
+___
+	}
+$code.=<<___;
+	.byte	0xf3,0xc3			# rep ret
+.size	_x86_64_AES_decrypt,.-_x86_64_AES_decrypt
+___
+
+sub deccompactvert()
+{ my ($t3,$t4,$t5)=("%r8d","%r9d","%r13d");
+
+$code.=<<___;
+	movzb	`&lo("$s0")`,$t0
+	movzb	`&lo("$s1")`,$t1
+	movzb	`&lo("$s2")`,$t2
+	movzb	($sbox,$t0,1),$t0
+	movzb	($sbox,$t1,1),$t1
+	movzb	($sbox,$t2,1),$t2
+
+	movzb	`&lo("$s3")`,$t3
+	movzb	`&hi("$s3")`,$acc0
+	movzb	`&hi("$s0")`,$acc1
+	movzb	($sbox,$t3,1),$t3
+	movzb	($sbox,$acc0,1),$t4	#$t0
+	movzb	($sbox,$acc1,1),$t5	#$t1
+
+	movzb	`&hi("$s1")`,$acc2
+	movzb	`&hi("$s2")`,$acc0
+	shr	\$16,$s2
+	movzb	($sbox,$acc2,1),$acc2	#$t2
+	movzb	($sbox,$acc0,1),$acc0	#$t3
+	shr	\$16,$s3
+
+	movzb	`&lo("$s2")`,$acc1
+	shl	\$8,$t4
+	shl	\$8,$t5
+	movzb	($sbox,$acc1,1),$acc1	#$t0
+	xor	$t4,$t0
+	xor	$t5,$t1
+
+	movzb	`&lo("$s3")`,$t4
+	shr	\$16,$s0
+	shr	\$16,$s1
+	movzb	`&lo("$s0")`,$t5
+	shl	\$8,$acc2
+	shl	\$8,$acc0
+	movzb	($sbox,$t4,1),$t4	#$t1
+	movzb	($sbox,$t5,1),$t5	#$t2
+	xor	$acc2,$t2
+	xor	$acc0,$t3
+
+	movzb	`&lo("$s1")`,$acc2
+	movzb	`&hi("$s1")`,$acc0
+	shl	\$16,$acc1
+	movzb	($sbox,$acc2,1),$acc2	#$t3
+	movzb	($sbox,$acc0,1),$acc0	#$t0
+	xor	$acc1,$t0
+
+	movzb	`&hi("$s2")`,$acc1
+	shl	\$16,$t4
+	shl	\$16,$t5
+	movzb	($sbox,$acc1,1),$s1	#$t1
+	xor	$t4,$t1
+	xor	$t5,$t2
+
+	movzb	`&hi("$s3")`,$acc1
+	shr	\$8,$s0
+	shl	\$16,$acc2
+	movzb	($sbox,$acc1,1),$s2	#$t2
+	movzb	($sbox,$s0,1),$s3	#$t3
+	xor	$acc2,$t3
+
+	shl	\$24,$acc0
+	shl	\$24,$s1
+	shl	\$24,$s2
+	xor	$acc0,$t0
+	shl	\$24,$s3
+	xor	$t1,$s1
+	mov	$t0,$s0
+	xor	$t2,$s2
+	xor	$t3,$s3
+___
+}
+
+# parallelized version! input is pair of 64-bit values: %rax=s1.s0
+# and %rcx=s3.s2, output is four 32-bit values in %eax=s0, %ebx=s1,
+# %ecx=s2 and %edx=s3.
+sub dectransform()
+{ my ($tp10,$tp20,$tp40,$tp80,$acc0)=("%rax","%r8", "%r9", "%r10","%rbx");
+  my ($tp18,$tp28,$tp48,$tp88,$acc8)=("%rcx","%r11","%r12","%r13","%rdx");
+  my $prefetch = shift;
+
+$code.=<<___;
+	mov	$tp10,$acc0
+	mov	$tp18,$acc8
+	and	$mask80,$acc0
+	and	$mask80,$acc8
+	mov	$acc0,$tp40
+	mov	$acc8,$tp48
+	shr	\$7,$tp40
+	lea	($tp10,$tp10),$tp20
+	shr	\$7,$tp48
+	lea	($tp18,$tp18),$tp28
+	sub	$tp40,$acc0
+	sub	$tp48,$acc8
+	and	$maskfe,$tp20
+	and	$maskfe,$tp28
+	and	$mask1b,$acc0
+	and	$mask1b,$acc8
+	xor	$tp20,$acc0
+	xor	$tp28,$acc8
+	mov	$acc0,$tp20
+	mov	$acc8,$tp28
+
+	and	$mask80,$acc0
+	and	$mask80,$acc8
+	mov	$acc0,$tp80
+	mov	$acc8,$tp88
+	shr	\$7,$tp80
+	lea	($tp20,$tp20),$tp40
+	shr	\$7,$tp88
+	lea	($tp28,$tp28),$tp48
+	sub	$tp80,$acc0
+	sub	$tp88,$acc8
+	and	$maskfe,$tp40
+	and	$maskfe,$tp48
+	and	$mask1b,$acc0
+	and	$mask1b,$acc8
+	xor	$tp40,$acc0
+	xor	$tp48,$acc8
+	mov	$acc0,$tp40
+	mov	$acc8,$tp48
+
+	and	$mask80,$acc0
+	and	$mask80,$acc8
+	mov	$acc0,$tp80
+	mov	$acc8,$tp88
+	shr	\$7,$tp80
+	 xor	$tp10,$tp20		# tp2^=tp1
+	shr	\$7,$tp88
+	 xor	$tp18,$tp28		# tp2^=tp1
+	sub	$tp80,$acc0
+	sub	$tp88,$acc8
+	lea	($tp40,$tp40),$tp80
+	lea	($tp48,$tp48),$tp88
+	 xor	$tp10,$tp40		# tp4^=tp1
+	 xor	$tp18,$tp48		# tp4^=tp1
+	and	$maskfe,$tp80
+	and	$maskfe,$tp88
+	and	$mask1b,$acc0
+	and	$mask1b,$acc8
+	xor	$acc0,$tp80
+	xor	$acc8,$tp88
+
+	xor	$tp80,$tp10		# tp1^=tp8
+	xor	$tp88,$tp18		# tp1^=tp8
+	xor	$tp80,$tp20		# tp2^tp1^=tp8
+	xor	$tp88,$tp28		# tp2^tp1^=tp8
+	mov	$tp10,$acc0
+	mov	$tp18,$acc8
+	xor	$tp80,$tp40		# tp4^tp1^=tp8
+	xor	$tp88,$tp48		# tp4^tp1^=tp8
+	shr	\$32,$acc0
+	shr	\$32,$acc8
+	xor	$tp20,$tp80		# tp8^=tp8^tp2^tp1=tp2^tp1
+	xor	$tp28,$tp88		# tp8^=tp8^tp2^tp1=tp2^tp1
+	rol	\$8,`&LO("$tp10")`	# ROTATE(tp1^tp8,8)
+	rol	\$8,`&LO("$tp18")`	# ROTATE(tp1^tp8,8)
+	xor	$tp40,$tp80		# tp2^tp1^=tp8^tp4^tp1=tp8^tp4^tp2
+	xor	$tp48,$tp88		# tp2^tp1^=tp8^tp4^tp1=tp8^tp4^tp2
+
+	rol	\$8,`&LO("$acc0")`	# ROTATE(tp1^tp8,8)
+	rol	\$8,`&LO("$acc8")`	# ROTATE(tp1^tp8,8)
+	xor	`&LO("$tp80")`,`&LO("$tp10")`
+	xor	`&LO("$tp88")`,`&LO("$tp18")`
+	shr	\$32,$tp80
+	shr	\$32,$tp88
+	xor	`&LO("$tp80")`,`&LO("$acc0")`
+	xor	`&LO("$tp88")`,`&LO("$acc8")`
+
+	mov	$tp20,$tp80
+	mov	$tp28,$tp88
+	shr	\$32,$tp80
+	shr	\$32,$tp88
+	rol	\$24,`&LO("$tp20")`	# ROTATE(tp2^tp1^tp8,24)
+	rol	\$24,`&LO("$tp28")`	# ROTATE(tp2^tp1^tp8,24)
+	rol	\$24,`&LO("$tp80")`	# ROTATE(tp2^tp1^tp8,24)
+	rol	\$24,`&LO("$tp88")`	# ROTATE(tp2^tp1^tp8,24)
+	xor	`&LO("$tp20")`,`&LO("$tp10")`
+	xor	`&LO("$tp28")`,`&LO("$tp18")`
+	mov	$tp40,$tp20
+	mov	$tp48,$tp28
+	xor	`&LO("$tp80")`,`&LO("$acc0")`
+	xor	`&LO("$tp88")`,`&LO("$acc8")`
+
+	`"mov	0($sbox),$mask80"	if ($prefetch)`
+	shr	\$32,$tp20
+	shr	\$32,$tp28
+	`"mov	64($sbox),$maskfe"	if ($prefetch)`
+	rol	\$16,`&LO("$tp40")`	# ROTATE(tp4^tp1^tp8,16)
+	rol	\$16,`&LO("$tp48")`	# ROTATE(tp4^tp1^tp8,16)
+	`"mov	128($sbox),$mask1b"	if ($prefetch)`
+	rol	\$16,`&LO("$tp20")`	# ROTATE(tp4^tp1^tp8,16)
+	rol	\$16,`&LO("$tp28")`	# ROTATE(tp4^tp1^tp8,16)
+	`"mov	192($sbox),$tp80"	if ($prefetch)`
+	xor	`&LO("$tp40")`,`&LO("$tp10")`
+	xor	`&LO("$tp48")`,`&LO("$tp18")`
+	`"mov	256($sbox),$tp88"	if ($prefetch)`
+	xor	`&LO("$tp20")`,`&LO("$acc0")`
+	xor	`&LO("$tp28")`,`&LO("$acc8")`
+___
+}
+
+$code.=<<___;
+.type	_x86_64_AES_decrypt_compact,\@abi-omnipotent
+.align	16
+_x86_64_AES_decrypt_compact:
+	lea	128($sbox),$inp			# size optimization
+	mov	0-128($inp),$acc1		# prefetch Td4
+	mov	32-128($inp),$acc2
+	mov	64-128($inp),$t0
+	mov	96-128($inp),$t1
+	mov	128-128($inp),$acc1
+	mov	160-128($inp),$acc2
+	mov	192-128($inp),$t0
+	mov	224-128($inp),$t1
+	jmp	.Ldec_loop_compact
+
+.align	16
+.Ldec_loop_compact:
+		xor	0($key),$s0		# xor with key
+		xor	4($key),$s1
+		xor	8($key),$s2
+		xor	12($key),$s3
+		lea	16($key),$key
+___
+		&deccompactvert();
+$code.=<<___;
+		cmp	16(%rsp),$key
+		je	.Ldec_compact_done
+
+		mov	256+0($sbox),$mask80
+		shl	\$32,%rbx
+		shl	\$32,%rdx
+		mov	256+8($sbox),$maskfe
+		or	%rbx,%rax
+		or	%rdx,%rcx
+		mov	256+16($sbox),$mask1b
+___
+		&dectransform(1);
+$code.=<<___;
+	jmp	.Ldec_loop_compact
+.align	16
+.Ldec_compact_done:
+	xor	0($key),$s0
+	xor	4($key),$s1
+	xor	8($key),$s2
+	xor	12($key),$s3
+	.byte	0xf3,0xc3			# rep ret
+.size	_x86_64_AES_decrypt_compact,.-_x86_64_AES_decrypt_compact
+___
+
+# void AES_decrypt (const void *inp,void *out,const AES_KEY *key);
+$code.=<<___;
+.globl	AES_decrypt
+.type	AES_decrypt,\@function,3
+.align	16
+AES_decrypt:
+	push	%rbx
+	push	%rbp
+	push	%r12
+	push	%r13
+	push	%r14
+	push	%r15
+
+	# allocate frame "above" key schedule
+	mov	%rsp,%r10
+	lea	-63(%rdx),%rcx	# %rdx is key argument
+	and	\$-64,%rsp
+	sub	%rsp,%rcx
+	neg	%rcx
+	and	\$0x3c0,%rcx
+	sub	%rcx,%rsp
+	sub	\$32,%rsp
+
+	mov	%rsi,16(%rsp)	# save out
+	mov	%r10,24(%rsp)	# save real stack pointer
+.Ldec_prologue:
+
+	mov	%rdx,$key
+	mov	240($key),$rnds	# load rounds
+
+	mov	0(%rdi),$s0	# load input vector
+	mov	4(%rdi),$s1
+	mov	8(%rdi),$s2
+	mov	12(%rdi),$s3
+
+	shl	\$4,$rnds
+	lea	($key,$rnds),%rbp
+	mov	$key,(%rsp)	# key schedule
+	mov	%rbp,8(%rsp)	# end of key schedule
+
+	# pick Td4 copy which can't "overlap" with stack frame or key schedule
+	lea	.LAES_Td+2048(%rip),$sbox
+	lea	768(%rsp),%rbp
+	sub	$sbox,%rbp
+	and	\$0x300,%rbp
+	lea	($sbox,%rbp),$sbox
+	shr	\$3,%rbp	# recall "magic" constants!
+	add	%rbp,$sbox
+
+	call	_x86_64_AES_decrypt_compact
+
+	mov	16(%rsp),$out	# restore out
+	mov	24(%rsp),%rsi	# restore saved stack pointer
+	mov	$s0,0($out)	# write output vector
+	mov	$s1,4($out)
+	mov	$s2,8($out)
+	mov	$s3,12($out)
+
+	mov	(%rsi),%r15
+	mov	8(%rsi),%r14
+	mov	16(%rsi),%r13
+	mov	24(%rsi),%r12
+	mov	32(%rsi),%rbp
+	mov	40(%rsi),%rbx
+	lea	48(%rsi),%rsp
+.Ldec_epilogue:
+	ret
+.size	AES_decrypt,.-AES_decrypt
+___
+#------------------------------------------------------------------#
+
+sub enckey()
+{
+$code.=<<___;
+	movz	%dl,%esi		# rk[i]>>0
+	movzb	-128(%rbp,%rsi),%ebx
+	movz	%dh,%esi		# rk[i]>>8
+	shl	\$24,%ebx
+	xor	%ebx,%eax
+
+	movzb	-128(%rbp,%rsi),%ebx
+	shr	\$16,%edx
+	movz	%dl,%esi		# rk[i]>>16
+	xor	%ebx,%eax
+
+	movzb	-128(%rbp,%rsi),%ebx
+	movz	%dh,%esi		# rk[i]>>24
+	shl	\$8,%ebx
+	xor	%ebx,%eax
+
+	movzb	-128(%rbp,%rsi),%ebx
+	shl	\$16,%ebx
+	xor	%ebx,%eax
+
+	xor	1024-128(%rbp,%rcx,4),%eax		# rcon
+___
+}
+
+# int AES_set_encrypt_key(const unsigned char *userKey, const int bits,
+#                        AES_KEY *key)
+$code.=<<___;
+.globl	AES_set_encrypt_key
+.type	AES_set_encrypt_key,\@function,3
+.align	16
+AES_set_encrypt_key:
+	push	%rbx
+	push	%rbp
+	push	%r12			# redundant, but allows to share 
+	push	%r13			# exception handler...
+	push	%r14
+	push	%r15
+	sub	\$8,%rsp
+.Lenc_key_prologue:
+
+	call	_x86_64_AES_set_encrypt_key
+
+	mov	8(%rsp),%r15
+	mov	16(%rsp),%r14
+	mov	24(%rsp),%r13
+	mov	32(%rsp),%r12
+	mov	40(%rsp),%rbp
+	mov	48(%rsp),%rbx
+	add	\$56,%rsp
+.Lenc_key_epilogue:
+	ret
+.size	AES_set_encrypt_key,.-AES_set_encrypt_key
+
+.type	_x86_64_AES_set_encrypt_key,\@abi-omnipotent
+.align	16
+_x86_64_AES_set_encrypt_key:
+	mov	%esi,%ecx			# %ecx=bits
+	mov	%rdi,%rsi			# %rsi=userKey
+	mov	%rdx,%rdi			# %rdi=key
+
+	test	\$-1,%rsi
+	jz	.Lbadpointer
+	test	\$-1,%rdi
+	jz	.Lbadpointer
+
+	lea	.LAES_Te(%rip),%rbp
+	lea	2048+128(%rbp),%rbp
+
+	# prefetch Te4
+	mov	0-128(%rbp),%eax
+	mov	32-128(%rbp),%ebx
+	mov	64-128(%rbp),%r8d
+	mov	96-128(%rbp),%edx
+	mov	128-128(%rbp),%eax
+	mov	160-128(%rbp),%ebx
+	mov	192-128(%rbp),%r8d
+	mov	224-128(%rbp),%edx
+
+	cmp	\$128,%ecx
+	je	.L10rounds
+	cmp	\$192,%ecx
+	je	.L12rounds
+	cmp	\$256,%ecx
+	je	.L14rounds
+	mov	\$-2,%rax			# invalid number of bits
+	jmp	.Lexit
+
+.L10rounds:
+	mov	0(%rsi),%rax			# copy first 4 dwords
+	mov	8(%rsi),%rdx
+	mov	%rax,0(%rdi)
+	mov	%rdx,8(%rdi)
+
+	shr	\$32,%rdx
+	xor	%ecx,%ecx
+	jmp	.L10shortcut
+.align	4
+.L10loop:
+		mov	0(%rdi),%eax			# rk[0]
+		mov	12(%rdi),%edx			# rk[3]
+.L10shortcut:
+___
+		&enckey	();
+$code.=<<___;
+		mov	%eax,16(%rdi)			# rk[4]
+		xor	4(%rdi),%eax
+		mov	%eax,20(%rdi)			# rk[5]
+		xor	8(%rdi),%eax
+		mov	%eax,24(%rdi)			# rk[6]
+		xor	12(%rdi),%eax
+		mov	%eax,28(%rdi)			# rk[7]
+		add	\$1,%ecx
+		lea	16(%rdi),%rdi
+		cmp	\$10,%ecx
+	jl	.L10loop
+
+	movl	\$10,80(%rdi)			# setup number of rounds
+	xor	%rax,%rax
+	jmp	.Lexit
+
+.L12rounds:
+	mov	0(%rsi),%rax			# copy first 6 dwords
+	mov	8(%rsi),%rbx
+	mov	16(%rsi),%rdx
+	mov	%rax,0(%rdi)
+	mov	%rbx,8(%rdi)
+	mov	%rdx,16(%rdi)
+
+	shr	\$32,%rdx
+	xor	%ecx,%ecx
+	jmp	.L12shortcut
+.align	4
+.L12loop:
+		mov	0(%rdi),%eax			# rk[0]
+		mov	20(%rdi),%edx			# rk[5]
+.L12shortcut:
+___
+		&enckey	();
+$code.=<<___;
+		mov	%eax,24(%rdi)			# rk[6]
+		xor	4(%rdi),%eax
+		mov	%eax,28(%rdi)			# rk[7]
+		xor	8(%rdi),%eax
+		mov	%eax,32(%rdi)			# rk[8]
+		xor	12(%rdi),%eax
+		mov	%eax,36(%rdi)			# rk[9]
+
+		cmp	\$7,%ecx
+		je	.L12break
+		add	\$1,%ecx
+
+		xor	16(%rdi),%eax
+		mov	%eax,40(%rdi)			# rk[10]
+		xor	20(%rdi),%eax
+		mov	%eax,44(%rdi)			# rk[11]
+
+		lea	24(%rdi),%rdi
+	jmp	.L12loop
+.L12break:
+	movl	\$12,72(%rdi)		# setup number of rounds
+	xor	%rax,%rax
+	jmp	.Lexit
+
+.L14rounds:		
+	mov	0(%rsi),%rax			# copy first 8 dwords
+	mov	8(%rsi),%rbx
+	mov	16(%rsi),%rcx
+	mov	24(%rsi),%rdx
+	mov	%rax,0(%rdi)
+	mov	%rbx,8(%rdi)
+	mov	%rcx,16(%rdi)
+	mov	%rdx,24(%rdi)
+
+	shr	\$32,%rdx
+	xor	%ecx,%ecx
+	jmp	.L14shortcut
+.align	4
+.L14loop:
+		mov	0(%rdi),%eax			# rk[0]
+		mov	28(%rdi),%edx			# rk[4]
+.L14shortcut:
+___
+		&enckey	();
+$code.=<<___;
+		mov	%eax,32(%rdi)			# rk[8]
+		xor	4(%rdi),%eax
+		mov	%eax,36(%rdi)			# rk[9]
+		xor	8(%rdi),%eax
+		mov	%eax,40(%rdi)			# rk[10]
+		xor	12(%rdi),%eax
+		mov	%eax,44(%rdi)			# rk[11]
+
+		cmp	\$6,%ecx
+		je	.L14break
+		add	\$1,%ecx
+
+		mov	%eax,%edx
+		mov	16(%rdi),%eax			# rk[4]
+		movz	%dl,%esi			# rk[11]>>0
+		movzb	-128(%rbp,%rsi),%ebx
+		movz	%dh,%esi			# rk[11]>>8
+		xor	%ebx,%eax
+
+		movzb	-128(%rbp,%rsi),%ebx
+		shr	\$16,%edx
+		shl	\$8,%ebx
+		movz	%dl,%esi			# rk[11]>>16
+		xor	%ebx,%eax
+
+		movzb	-128(%rbp,%rsi),%ebx
+		movz	%dh,%esi			# rk[11]>>24
+		shl	\$16,%ebx
+		xor	%ebx,%eax
+
+		movzb	-128(%rbp,%rsi),%ebx
+		shl	\$24,%ebx
+		xor	%ebx,%eax
+
+		mov	%eax,48(%rdi)			# rk[12]
+		xor	20(%rdi),%eax
+		mov	%eax,52(%rdi)			# rk[13]
+		xor	24(%rdi),%eax
+		mov	%eax,56(%rdi)			# rk[14]
+		xor	28(%rdi),%eax
+		mov	%eax,60(%rdi)			# rk[15]
+
+		lea	32(%rdi),%rdi
+	jmp	.L14loop
+.L14break:
+	movl	\$14,48(%rdi)		# setup number of rounds
+	xor	%rax,%rax
+	jmp	.Lexit
+
+.Lbadpointer:
+	mov	\$-1,%rax
+.Lexit:
+	.byte	0xf3,0xc3			# rep ret
+.size	_x86_64_AES_set_encrypt_key,.-_x86_64_AES_set_encrypt_key
+___
+
+sub deckey_ref()
+{ my ($i,$ptr,$te,$td) = @_;
+  my ($tp1,$tp2,$tp4,$tp8,$acc)=("%eax","%ebx","%edi","%edx","%r8d");
+$code.=<<___;
+	mov	$i($ptr),$tp1
+	mov	$tp1,$acc
+	and	\$0x80808080,$acc
+	mov	$acc,$tp4
+	shr	\$7,$tp4
+	lea	0($tp1,$tp1),$tp2
+	sub	$tp4,$acc
+	and	\$0xfefefefe,$tp2
+	and	\$0x1b1b1b1b,$acc
+	xor	$tp2,$acc
+	mov	$acc,$tp2
+
+	and	\$0x80808080,$acc
+	mov	$acc,$tp8
+	shr	\$7,$tp8
+	lea	0($tp2,$tp2),$tp4
+	sub	$tp8,$acc
+	and	\$0xfefefefe,$tp4
+	and	\$0x1b1b1b1b,$acc
+	 xor	$tp1,$tp2		# tp2^tp1
+	xor	$tp4,$acc
+	mov	$acc,$tp4
+
+	and	\$0x80808080,$acc
+	mov	$acc,$tp8
+	shr	\$7,$tp8
+	sub	$tp8,$acc
+	lea	0($tp4,$tp4),$tp8
+	 xor	$tp1,$tp4		# tp4^tp1
+	and	\$0xfefefefe,$tp8
+	and	\$0x1b1b1b1b,$acc
+	xor	$acc,$tp8
+
+	xor	$tp8,$tp1		# tp1^tp8
+	rol	\$8,$tp1		# ROTATE(tp1^tp8,8)
+	xor	$tp8,$tp2		# tp2^tp1^tp8
+	xor	$tp8,$tp4		# tp4^tp1^tp8
+	xor	$tp2,$tp8
+	xor	$tp4,$tp8		# tp8^(tp8^tp4^tp1)^(tp8^tp2^tp1)=tp8^tp4^tp2
+
+	xor	$tp8,$tp1
+	rol	\$24,$tp2		# ROTATE(tp2^tp1^tp8,24)
+	xor	$tp2,$tp1
+	rol	\$16,$tp4		# ROTATE(tp4^tp1^tp8,16)
+	xor	$tp4,$tp1
+
+	mov	$tp1,$i($ptr)
+___
+}
+
+# int AES_set_decrypt_key(const unsigned char *userKey, const int bits,
+#                        AES_KEY *key)
+$code.=<<___;
+.globl	AES_set_decrypt_key
+.type	AES_set_decrypt_key,\@function,3
+.align	16
+AES_set_decrypt_key:
+	push	%rbx
+	push	%rbp
+	push	%r12
+	push	%r13
+	push	%r14
+	push	%r15
+	push	%rdx			# save key schedule
+.Ldec_key_prologue:
+
+	call	_x86_64_AES_set_encrypt_key
+	mov	(%rsp),%r8		# restore key schedule
+	cmp	\$0,%eax
+	jne	.Labort
+
+	mov	240(%r8),%r14d		# pull number of rounds
+	xor	%rdi,%rdi
+	lea	(%rdi,%r14d,4),%rcx
+	mov	%r8,%rsi
+	lea	(%r8,%rcx,4),%rdi	# pointer to last chunk
+.align	4
+.Linvert:
+		mov	0(%rsi),%rax
+		mov	8(%rsi),%rbx
+		mov	0(%rdi),%rcx
+		mov	8(%rdi),%rdx
+		mov	%rax,0(%rdi)
+		mov	%rbx,8(%rdi)
+		mov	%rcx,0(%rsi)
+		mov	%rdx,8(%rsi)
+		lea	16(%rsi),%rsi
+		lea	-16(%rdi),%rdi
+		cmp	%rsi,%rdi
+	jne	.Linvert
+
+	lea	.LAES_Te+2048+1024(%rip),%rax	# rcon
+
+	mov	40(%rax),$mask80
+	mov	48(%rax),$maskfe
+	mov	56(%rax),$mask1b
+
+	mov	%r8,$key
+	sub	\$1,%r14d
+.align	4
+.Lpermute:
+		lea	16($key),$key
+		mov	0($key),%rax
+		mov	8($key),%rcx
+___
+		&dectransform ();
+$code.=<<___;
+		mov	%eax,0($key)
+		mov	%ebx,4($key)
+		mov	%ecx,8($key)
+		mov	%edx,12($key)
+		sub	\$1,%r14d
+	jnz	.Lpermute
+
+	xor	%rax,%rax
+.Labort:
+	mov	8(%rsp),%r15
+	mov	16(%rsp),%r14
+	mov	24(%rsp),%r13
+	mov	32(%rsp),%r12
+	mov	40(%rsp),%rbp
+	mov	48(%rsp),%rbx
+	add	\$56,%rsp
+.Ldec_key_epilogue:
+	ret
+.size	AES_set_decrypt_key,.-AES_set_decrypt_key
+___
+
+# void AES_cbc_encrypt (const void char *inp, unsigned char *out,
+#			size_t length, const AES_KEY *key,
+#			unsigned char *ivp,const int enc);
+{
+# stack frame layout
+# -8(%rsp)		return address
+my $keyp="0(%rsp)";		# one to pass as $key
+my $keyend="8(%rsp)";		# &(keyp->rd_key[4*keyp->rounds])
+my $_rsp="16(%rsp)";		# saved %rsp
+my $_inp="24(%rsp)";		# copy of 1st parameter, inp
+my $_out="32(%rsp)";		# copy of 2nd parameter, out
+my $_len="40(%rsp)";		# copy of 3rd parameter, length
+my $_key="48(%rsp)";		# copy of 4th parameter, key
+my $_ivp="56(%rsp)";		# copy of 5th parameter, ivp
+my $ivec="64(%rsp)";		# ivec[16]
+my $aes_key="80(%rsp)";		# copy of aes_key
+my $mark="80+240(%rsp)";	# copy of aes_key->rounds
+
+$code.=<<___;
+.globl	AES_cbc_encrypt
+.type	AES_cbc_encrypt,\@function,6
+.align	16
+.extern	OPENSSL_ia32cap_P
+AES_cbc_encrypt:
+	cmp	\$0,%rdx	# check length
+	je	.Lcbc_epilogue
+	pushfq
+	push	%rbx
+	push	%rbp
+	push	%r12
+	push	%r13
+	push	%r14
+	push	%r15
+.Lcbc_prologue:
+
+	cld
+	mov	%r9d,%r9d	# clear upper half of enc
+
+	lea	.LAES_Te(%rip),$sbox
+	cmp	\$0,%r9
+	jne	.Lcbc_picked_te
+	lea	.LAES_Td(%rip),$sbox
+.Lcbc_picked_te:
+
+	mov	OPENSSL_ia32cap_P(%rip),%r10d
+	cmp	\$$speed_limit,%rdx
+	jb	.Lcbc_slow_prologue
+	test	\$15,%rdx
+	jnz	.Lcbc_slow_prologue
+	bt	\$28,%r10d
+	jc	.Lcbc_slow_prologue
+
+	# allocate aligned stack frame...
+	lea	-88-248(%rsp),$key
+	and	\$-64,$key
+
+	# ... and make sure it doesn't alias with AES_T[ed] modulo 4096
+	mov	$sbox,%r10
+	lea	2304($sbox),%r11
+	mov	$key,%r12
+	and	\$0xFFF,%r10	# s = $sbox&0xfff
+	and	\$0xFFF,%r11	# e = ($sbox+2048)&0xfff
+	and	\$0xFFF,%r12	# p = %rsp&0xfff
+
+	cmp	%r11,%r12	# if (p=>e) %rsp =- (p-e);
+	jb	.Lcbc_te_break_out
+	sub	%r11,%r12
+	sub	%r12,$key
+	jmp	.Lcbc_te_ok
+.Lcbc_te_break_out:		# else %rsp -= (p-s)&0xfff + framesz
+	sub	%r10,%r12
+	and	\$0xFFF,%r12
+	add	\$320,%r12
+	sub	%r12,$key
+.align	4
+.Lcbc_te_ok:
+
+	xchg	%rsp,$key
+	#add	\$8,%rsp	# reserve for return address!
+	mov	$key,$_rsp	# save %rsp
+.Lcbc_fast_body:
+	mov	%rdi,$_inp	# save copy of inp
+	mov	%rsi,$_out	# save copy of out
+	mov	%rdx,$_len	# save copy of len
+	mov	%rcx,$_key	# save copy of key
+	mov	%r8,$_ivp	# save copy of ivp
+	movl	\$0,$mark	# copy of aes_key->rounds = 0;
+	mov	%r8,%rbp	# rearrange input arguments
+	mov	%r9,%rbx
+	mov	%rsi,$out
+	mov	%rdi,$inp
+	mov	%rcx,$key
+
+	mov	240($key),%eax		# key->rounds
+	# do we copy key schedule to stack?
+	mov	$key,%r10
+	sub	$sbox,%r10
+	and	\$0xfff,%r10
+	cmp	\$2304,%r10
+	jb	.Lcbc_do_ecopy
+	cmp	\$4096-248,%r10
+	jb	.Lcbc_skip_ecopy
+.align	4
+.Lcbc_do_ecopy:
+		mov	$key,%rsi
+		lea	$aes_key,%rdi
+		lea	$aes_key,$key
+		mov	\$240/8,%ecx
+		.long	0x90A548F3	# rep movsq
+		mov	%eax,(%rdi)	# copy aes_key->rounds
+.Lcbc_skip_ecopy:
+	mov	$key,$keyp	# save key pointer
+
+	mov	\$18,%ecx
+.align	4
+.Lcbc_prefetch_te:
+		mov	0($sbox),%r10
+		mov	32($sbox),%r11
+		mov	64($sbox),%r12
+		mov	96($sbox),%r13
+		lea	128($sbox),$sbox
+		sub	\$1,%ecx
+	jnz	.Lcbc_prefetch_te
+	lea	-2304($sbox),$sbox
+
+	cmp	\$0,%rbx
+	je	.LFAST_DECRYPT
+
+#----------------------------- ENCRYPT -----------------------------#
+	mov	0(%rbp),$s0		# load iv
+	mov	4(%rbp),$s1
+	mov	8(%rbp),$s2
+	mov	12(%rbp),$s3
+
+.align	4
+.Lcbc_fast_enc_loop:
+		xor	0($inp),$s0
+		xor	4($inp),$s1
+		xor	8($inp),$s2
+		xor	12($inp),$s3
+		mov	$keyp,$key	# restore key
+		mov	$inp,$_inp	# if ($verticalspin) save inp
+
+		call	_x86_64_AES_encrypt
+
+		mov	$_inp,$inp	# if ($verticalspin) restore inp
+		mov	$_len,%r10
+		mov	$s0,0($out)
+		mov	$s1,4($out)
+		mov	$s2,8($out)
+		mov	$s3,12($out)
+
+		lea	16($inp),$inp
+		lea	16($out),$out
+		sub	\$16,%r10
+		test	\$-16,%r10
+		mov	%r10,$_len
+	jnz	.Lcbc_fast_enc_loop
+	mov	$_ivp,%rbp	# restore ivp
+	mov	$s0,0(%rbp)	# save ivec
+	mov	$s1,4(%rbp)
+	mov	$s2,8(%rbp)
+	mov	$s3,12(%rbp)
+
+	jmp	.Lcbc_fast_cleanup
+
+#----------------------------- DECRYPT -----------------------------#
+.align	16
+.LFAST_DECRYPT:
+	cmp	$inp,$out
+	je	.Lcbc_fast_dec_in_place
+
+	mov	%rbp,$ivec
+.align	4
+.Lcbc_fast_dec_loop:
+		mov	0($inp),$s0	# read input
+		mov	4($inp),$s1
+		mov	8($inp),$s2
+		mov	12($inp),$s3
+		mov	$keyp,$key	# restore key
+		mov	$inp,$_inp	# if ($verticalspin) save inp
+
+		call	_x86_64_AES_decrypt
+
+		mov	$ivec,%rbp	# load ivp
+		mov	$_inp,$inp	# if ($verticalspin) restore inp
+		mov	$_len,%r10	# load len
+		xor	0(%rbp),$s0	# xor iv
+		xor	4(%rbp),$s1
+		xor	8(%rbp),$s2
+		xor	12(%rbp),$s3
+		mov	$inp,%rbp	# current input, next iv
+
+		sub	\$16,%r10
+		mov	%r10,$_len	# update len
+		mov	%rbp,$ivec	# update ivp
+
+		mov	$s0,0($out)	# write output
+		mov	$s1,4($out)
+		mov	$s2,8($out)
+		mov	$s3,12($out)
+
+		lea	16($inp),$inp
+		lea	16($out),$out
+	jnz	.Lcbc_fast_dec_loop
+	mov	$_ivp,%r12		# load user ivp
+	mov	0(%rbp),%r10		# load iv
+	mov	8(%rbp),%r11
+	mov	%r10,0(%r12)		# copy back to user
+	mov	%r11,8(%r12)
+	jmp	.Lcbc_fast_cleanup
+
+.align	16
+.Lcbc_fast_dec_in_place:
+	mov	0(%rbp),%r10		# copy iv to stack
+	mov	8(%rbp),%r11
+	mov	%r10,0+$ivec
+	mov	%r11,8+$ivec
+.align	4
+.Lcbc_fast_dec_in_place_loop:
+		mov	0($inp),$s0	# load input
+		mov	4($inp),$s1
+		mov	8($inp),$s2
+		mov	12($inp),$s3
+		mov	$keyp,$key	# restore key
+		mov	$inp,$_inp	# if ($verticalspin) save inp
+
+		call	_x86_64_AES_decrypt
+
+		mov	$_inp,$inp	# if ($verticalspin) restore inp
+		mov	$_len,%r10
+		xor	0+$ivec,$s0
+		xor	4+$ivec,$s1
+		xor	8+$ivec,$s2
+		xor	12+$ivec,$s3
+
+		mov	0($inp),%r11	# load input
+		mov	8($inp),%r12
+		sub	\$16,%r10
+		jz	.Lcbc_fast_dec_in_place_done
+
+		mov	%r11,0+$ivec	# copy input to iv
+		mov	%r12,8+$ivec
+
+		mov	$s0,0($out)	# save output [zaps input]
+		mov	$s1,4($out)
+		mov	$s2,8($out)
+		mov	$s3,12($out)
+
+		lea	16($inp),$inp
+		lea	16($out),$out
+		mov	%r10,$_len
+	jmp	.Lcbc_fast_dec_in_place_loop
+.Lcbc_fast_dec_in_place_done:
+	mov	$_ivp,%rdi
+	mov	%r11,0(%rdi)	# copy iv back to user
+	mov	%r12,8(%rdi)
+
+	mov	$s0,0($out)	# save output [zaps input]
+	mov	$s1,4($out)
+	mov	$s2,8($out)
+	mov	$s3,12($out)
+
+.align	4
+.Lcbc_fast_cleanup:
+	cmpl	\$0,$mark	# was the key schedule copied?
+	lea	$aes_key,%rdi
+	je	.Lcbc_exit
+		mov	\$240/8,%ecx
+		xor	%rax,%rax
+		.long	0x90AB48F3	# rep stosq
+
+	jmp	.Lcbc_exit
+
+#--------------------------- SLOW ROUTINE ---------------------------#
+.align	16
+.Lcbc_slow_prologue:
+	# allocate aligned stack frame...
+	lea	-88(%rsp),%rbp
+	and	\$-64,%rbp
+	# ... just "above" key schedule
+	lea	-88-63(%rcx),%r10
+	sub	%rbp,%r10
+	neg	%r10
+	and	\$0x3c0,%r10
+	sub	%r10,%rbp
+
+	xchg	%rsp,%rbp
+	#add	\$8,%rsp	# reserve for return address!
+	mov	%rbp,$_rsp	# save %rsp
+.Lcbc_slow_body:
+	#mov	%rdi,$_inp	# save copy of inp
+	#mov	%rsi,$_out	# save copy of out
+	#mov	%rdx,$_len	# save copy of len
+	#mov	%rcx,$_key	# save copy of key
+	mov	%r8,$_ivp	# save copy of ivp
+	mov	%r8,%rbp	# rearrange input arguments
+	mov	%r9,%rbx
+	mov	%rsi,$out
+	mov	%rdi,$inp
+	mov	%rcx,$key
+	mov	%rdx,%r10
+
+	mov	240($key),%eax
+	mov	$key,$keyp	# save key pointer
+	shl	\$4,%eax
+	lea	($key,%rax),%rax
+	mov	%rax,$keyend
+
+	# pick Te4 copy which can't "overlap" with stack frame or key scdedule
+	lea	2048($sbox),$sbox
+	lea	768-8(%rsp),%rax
+	sub	$sbox,%rax
+	and	\$0x300,%rax
+	lea	($sbox,%rax),$sbox
+
+	cmp	\$0,%rbx
+	je	.LSLOW_DECRYPT
+
+#--------------------------- SLOW ENCRYPT ---------------------------#
+	test	\$-16,%r10		# check upon length
+	mov	0(%rbp),$s0		# load iv
+	mov	4(%rbp),$s1
+	mov	8(%rbp),$s2
+	mov	12(%rbp),$s3
+	jz	.Lcbc_slow_enc_tail	# short input...
+
+.align	4
+.Lcbc_slow_enc_loop:
+		xor	0($inp),$s0
+		xor	4($inp),$s1
+		xor	8($inp),$s2
+		xor	12($inp),$s3
+		mov	$keyp,$key	# restore key
+		mov	$inp,$_inp	# save inp
+		mov	$out,$_out	# save out
+		mov	%r10,$_len	# save len
+
+		call	_x86_64_AES_encrypt_compact
+
+		mov	$_inp,$inp	# restore inp
+		mov	$_out,$out	# restore out
+		mov	$_len,%r10	# restore len
+		mov	$s0,0($out)
+		mov	$s1,4($out)
+		mov	$s2,8($out)
+		mov	$s3,12($out)
+
+		lea	16($inp),$inp
+		lea	16($out),$out
+		sub	\$16,%r10
+		test	\$-16,%r10
+	jnz	.Lcbc_slow_enc_loop
+	test	\$15,%r10
+	jnz	.Lcbc_slow_enc_tail
+	mov	$_ivp,%rbp	# restore ivp
+	mov	$s0,0(%rbp)	# save ivec
+	mov	$s1,4(%rbp)
+	mov	$s2,8(%rbp)
+	mov	$s3,12(%rbp)
+
+	jmp	.Lcbc_exit
+
+.align	4
+.Lcbc_slow_enc_tail:
+	mov	%rax,%r11
+	mov	%rcx,%r12
+	mov	%r10,%rcx
+	mov	$inp,%rsi
+	mov	$out,%rdi
+	.long	0x9066A4F3		# rep movsb
+	mov	\$16,%rcx		# zero tail
+	sub	%r10,%rcx
+	xor	%rax,%rax
+	.long	0x9066AAF3		# rep stosb
+	mov	$out,$inp		# this is not a mistake!
+	mov	\$16,%r10		# len=16
+	mov	%r11,%rax
+	mov	%r12,%rcx
+	jmp	.Lcbc_slow_enc_loop	# one more spin...
+#--------------------------- SLOW DECRYPT ---------------------------#
+.align	16
+.LSLOW_DECRYPT:
+	shr	\$3,%rax
+	add	%rax,$sbox		# recall "magic" constants!
+
+	mov	0(%rbp),%r11		# copy iv to stack
+	mov	8(%rbp),%r12
+	mov	%r11,0+$ivec
+	mov	%r12,8+$ivec
+
+.align	4
+.Lcbc_slow_dec_loop:
+		mov	0($inp),$s0	# load input
+		mov	4($inp),$s1
+		mov	8($inp),$s2
+		mov	12($inp),$s3
+		mov	$keyp,$key	# restore key
+		mov	$inp,$_inp	# save inp
+		mov	$out,$_out	# save out
+		mov	%r10,$_len	# save len
+
+		call	_x86_64_AES_decrypt_compact
+
+		mov	$_inp,$inp	# restore inp
+		mov	$_out,$out	# restore out
+		mov	$_len,%r10
+		xor	0+$ivec,$s0
+		xor	4+$ivec,$s1
+		xor	8+$ivec,$s2
+		xor	12+$ivec,$s3
+
+		mov	0($inp),%r11	# load input
+		mov	8($inp),%r12
+		sub	\$16,%r10
+		jc	.Lcbc_slow_dec_partial
+		jz	.Lcbc_slow_dec_done
+
+		mov	%r11,0+$ivec	# copy input to iv
+		mov	%r12,8+$ivec
+
+		mov	$s0,0($out)	# save output [can zap input]
+		mov	$s1,4($out)
+		mov	$s2,8($out)
+		mov	$s3,12($out)
+
+		lea	16($inp),$inp
+		lea	16($out),$out
+	jmp	.Lcbc_slow_dec_loop
+.Lcbc_slow_dec_done:
+	mov	$_ivp,%rdi
+	mov	%r11,0(%rdi)		# copy iv back to user
+	mov	%r12,8(%rdi)
+
+	mov	$s0,0($out)		# save output [can zap input]
+	mov	$s1,4($out)
+	mov	$s2,8($out)
+	mov	$s3,12($out)
+
+	jmp	.Lcbc_exit
+
+.align	4
+.Lcbc_slow_dec_partial:
+	mov	$_ivp,%rdi
+	mov	%r11,0(%rdi)		# copy iv back to user
+	mov	%r12,8(%rdi)
+
+	mov	$s0,0+$ivec		# save output to stack
+	mov	$s1,4+$ivec
+	mov	$s2,8+$ivec
+	mov	$s3,12+$ivec
+
+	mov	$out,%rdi
+	lea	$ivec,%rsi
+	lea	16(%r10),%rcx
+	.long	0x9066A4F3	# rep movsb
+	jmp	.Lcbc_exit
+
+.align	16
+.Lcbc_exit:
+	mov	$_rsp,%rsi
+	mov	(%rsi),%r15
+	mov	8(%rsi),%r14
+	mov	16(%rsi),%r13
+	mov	24(%rsi),%r12
+	mov	32(%rsi),%rbp
+	mov	40(%rsi),%rbx
+	lea	48(%rsi),%rsp
+.Lcbc_popfq:
+	popfq
+.Lcbc_epilogue:
+	ret
+.size	AES_cbc_encrypt,.-AES_cbc_encrypt
+___
+}
+
+$code.=<<___;
+.align	64
+.LAES_Te:
+___
+	&_data_word(0xa56363c6, 0x847c7cf8, 0x997777ee, 0x8d7b7bf6);
+	&_data_word(0x0df2f2ff, 0xbd6b6bd6, 0xb16f6fde, 0x54c5c591);
+	&_data_word(0x50303060, 0x03010102, 0xa96767ce, 0x7d2b2b56);
+	&_data_word(0x19fefee7, 0x62d7d7b5, 0xe6abab4d, 0x9a7676ec);
+	&_data_word(0x45caca8f, 0x9d82821f, 0x40c9c989, 0x877d7dfa);
+	&_data_word(0x15fafaef, 0xeb5959b2, 0xc947478e, 0x0bf0f0fb);
+	&_data_word(0xecadad41, 0x67d4d4b3, 0xfda2a25f, 0xeaafaf45);
+	&_data_word(0xbf9c9c23, 0xf7a4a453, 0x967272e4, 0x5bc0c09b);
+	&_data_word(0xc2b7b775, 0x1cfdfde1, 0xae93933d, 0x6a26264c);
+	&_data_word(0x5a36366c, 0x413f3f7e, 0x02f7f7f5, 0x4fcccc83);
+	&_data_word(0x5c343468, 0xf4a5a551, 0x34e5e5d1, 0x08f1f1f9);
+	&_data_word(0x937171e2, 0x73d8d8ab, 0x53313162, 0x3f15152a);
+	&_data_word(0x0c040408, 0x52c7c795, 0x65232346, 0x5ec3c39d);
+	&_data_word(0x28181830, 0xa1969637, 0x0f05050a, 0xb59a9a2f);
+	&_data_word(0x0907070e, 0x36121224, 0x9b80801b, 0x3de2e2df);
+	&_data_word(0x26ebebcd, 0x6927274e, 0xcdb2b27f, 0x9f7575ea);
+	&_data_word(0x1b090912, 0x9e83831d, 0x742c2c58, 0x2e1a1a34);
+	&_data_word(0x2d1b1b36, 0xb26e6edc, 0xee5a5ab4, 0xfba0a05b);
+	&_data_word(0xf65252a4, 0x4d3b3b76, 0x61d6d6b7, 0xceb3b37d);
+	&_data_word(0x7b292952, 0x3ee3e3dd, 0x712f2f5e, 0x97848413);
+	&_data_word(0xf55353a6, 0x68d1d1b9, 0x00000000, 0x2cededc1);
+	&_data_word(0x60202040, 0x1ffcfce3, 0xc8b1b179, 0xed5b5bb6);
+	&_data_word(0xbe6a6ad4, 0x46cbcb8d, 0xd9bebe67, 0x4b393972);
+	&_data_word(0xde4a4a94, 0xd44c4c98, 0xe85858b0, 0x4acfcf85);
+	&_data_word(0x6bd0d0bb, 0x2aefefc5, 0xe5aaaa4f, 0x16fbfbed);
+	&_data_word(0xc5434386, 0xd74d4d9a, 0x55333366, 0x94858511);
+	&_data_word(0xcf45458a, 0x10f9f9e9, 0x06020204, 0x817f7ffe);
+	&_data_word(0xf05050a0, 0x443c3c78, 0xba9f9f25, 0xe3a8a84b);
+	&_data_word(0xf35151a2, 0xfea3a35d, 0xc0404080, 0x8a8f8f05);
+	&_data_word(0xad92923f, 0xbc9d9d21, 0x48383870, 0x04f5f5f1);
+	&_data_word(0xdfbcbc63, 0xc1b6b677, 0x75dadaaf, 0x63212142);
+	&_data_word(0x30101020, 0x1affffe5, 0x0ef3f3fd, 0x6dd2d2bf);
+	&_data_word(0x4ccdcd81, 0x140c0c18, 0x35131326, 0x2fececc3);
+	&_data_word(0xe15f5fbe, 0xa2979735, 0xcc444488, 0x3917172e);
+	&_data_word(0x57c4c493, 0xf2a7a755, 0x827e7efc, 0x473d3d7a);
+	&_data_word(0xac6464c8, 0xe75d5dba, 0x2b191932, 0x957373e6);
+	&_data_word(0xa06060c0, 0x98818119, 0xd14f4f9e, 0x7fdcdca3);
+	&_data_word(0x66222244, 0x7e2a2a54, 0xab90903b, 0x8388880b);
+	&_data_word(0xca46468c, 0x29eeeec7, 0xd3b8b86b, 0x3c141428);
+	&_data_word(0x79dedea7, 0xe25e5ebc, 0x1d0b0b16, 0x76dbdbad);
+	&_data_word(0x3be0e0db, 0x56323264, 0x4e3a3a74, 0x1e0a0a14);
+	&_data_word(0xdb494992, 0x0a06060c, 0x6c242448, 0xe45c5cb8);
+	&_data_word(0x5dc2c29f, 0x6ed3d3bd, 0xefacac43, 0xa66262c4);
+	&_data_word(0xa8919139, 0xa4959531, 0x37e4e4d3, 0x8b7979f2);
+	&_data_word(0x32e7e7d5, 0x43c8c88b, 0x5937376e, 0xb76d6dda);
+	&_data_word(0x8c8d8d01, 0x64d5d5b1, 0xd24e4e9c, 0xe0a9a949);
+	&_data_word(0xb46c6cd8, 0xfa5656ac, 0x07f4f4f3, 0x25eaeacf);
+	&_data_word(0xaf6565ca, 0x8e7a7af4, 0xe9aeae47, 0x18080810);
+	&_data_word(0xd5baba6f, 0x887878f0, 0x6f25254a, 0x722e2e5c);
+	&_data_word(0x241c1c38, 0xf1a6a657, 0xc7b4b473, 0x51c6c697);
+	&_data_word(0x23e8e8cb, 0x7cdddda1, 0x9c7474e8, 0x211f1f3e);
+	&_data_word(0xdd4b4b96, 0xdcbdbd61, 0x868b8b0d, 0x858a8a0f);
+	&_data_word(0x907070e0, 0x423e3e7c, 0xc4b5b571, 0xaa6666cc);
+	&_data_word(0xd8484890, 0x05030306, 0x01f6f6f7, 0x120e0e1c);
+	&_data_word(0xa36161c2, 0x5f35356a, 0xf95757ae, 0xd0b9b969);
+	&_data_word(0x91868617, 0x58c1c199, 0x271d1d3a, 0xb99e9e27);
+	&_data_word(0x38e1e1d9, 0x13f8f8eb, 0xb398982b, 0x33111122);
+	&_data_word(0xbb6969d2, 0x70d9d9a9, 0x898e8e07, 0xa7949433);
+	&_data_word(0xb69b9b2d, 0x221e1e3c, 0x92878715, 0x20e9e9c9);
+	&_data_word(0x49cece87, 0xff5555aa, 0x78282850, 0x7adfdfa5);
+	&_data_word(0x8f8c8c03, 0xf8a1a159, 0x80898909, 0x170d0d1a);
+	&_data_word(0xdabfbf65, 0x31e6e6d7, 0xc6424284, 0xb86868d0);
+	&_data_word(0xc3414182, 0xb0999929, 0x772d2d5a, 0x110f0f1e);
+	&_data_word(0xcbb0b07b, 0xfc5454a8, 0xd6bbbb6d, 0x3a16162c);
+
+#Te4	# four copies of Te4 to choose from to avoid L1 aliasing
+	&data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5);
+	&data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76);
+	&data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0);
+	&data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0);
+	&data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc);
+	&data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15);
+	&data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a);
+	&data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75);
+	&data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0);
+	&data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84);
+	&data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b);
+	&data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf);
+	&data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85);
+	&data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8);
+	&data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5);
+	&data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2);
+	&data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17);
+	&data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73);
+	&data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88);
+	&data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb);
+	&data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c);
+	&data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79);
+	&data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9);
+	&data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08);
+	&data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6);
+	&data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a);
+	&data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e);
+	&data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e);
+	&data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94);
+	&data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf);
+	&data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68);
+	&data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16);
+
+	&data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5);
+	&data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76);
+	&data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0);
+	&data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0);
+	&data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc);
+	&data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15);
+	&data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a);
+	&data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75);
+	&data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0);
+	&data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84);
+	&data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b);
+	&data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf);
+	&data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85);
+	&data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8);
+	&data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5);
+	&data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2);
+	&data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17);
+	&data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73);
+	&data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88);
+	&data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb);
+	&data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c);
+	&data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79);
+	&data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9);
+	&data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08);
+	&data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6);
+	&data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a);
+	&data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e);
+	&data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e);
+	&data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94);
+	&data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf);
+	&data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68);
+	&data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16);
+
+	&data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5);
+	&data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76);
+	&data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0);
+	&data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0);
+	&data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc);
+	&data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15);
+	&data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a);
+	&data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75);
+	&data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0);
+	&data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84);
+	&data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b);
+	&data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf);
+	&data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85);
+	&data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8);
+	&data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5);
+	&data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2);
+	&data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17);
+	&data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73);
+	&data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88);
+	&data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb);
+	&data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c);
+	&data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79);
+	&data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9);
+	&data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08);
+	&data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6);
+	&data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a);
+	&data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e);
+	&data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e);
+	&data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94);
+	&data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf);
+	&data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68);
+	&data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16);
+
+	&data_byte(0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5);
+	&data_byte(0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76);
+	&data_byte(0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0);
+	&data_byte(0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0);
+	&data_byte(0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc);
+	&data_byte(0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15);
+	&data_byte(0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a);
+	&data_byte(0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75);
+	&data_byte(0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0);
+	&data_byte(0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84);
+	&data_byte(0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b);
+	&data_byte(0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf);
+	&data_byte(0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85);
+	&data_byte(0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8);
+	&data_byte(0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5);
+	&data_byte(0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2);
+	&data_byte(0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17);
+	&data_byte(0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73);
+	&data_byte(0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88);
+	&data_byte(0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb);
+	&data_byte(0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c);
+	&data_byte(0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79);
+	&data_byte(0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9);
+	&data_byte(0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08);
+	&data_byte(0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6);
+	&data_byte(0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a);
+	&data_byte(0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e);
+	&data_byte(0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e);
+	&data_byte(0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94);
+	&data_byte(0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf);
+	&data_byte(0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68);
+	&data_byte(0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16);
+#rcon:
+$code.=<<___;
+	.long	0x00000001, 0x00000002, 0x00000004, 0x00000008
+	.long	0x00000010, 0x00000020, 0x00000040, 0x00000080
+	.long	0x0000001b, 0x00000036, 0x80808080, 0x80808080
+	.long	0xfefefefe, 0xfefefefe, 0x1b1b1b1b, 0x1b1b1b1b
+___
+$code.=<<___;
+.align	64
+.LAES_Td:
+___
+	&_data_word(0x50a7f451, 0x5365417e, 0xc3a4171a, 0x965e273a);
+	&_data_word(0xcb6bab3b, 0xf1459d1f, 0xab58faac, 0x9303e34b);
+	&_data_word(0x55fa3020, 0xf66d76ad, 0x9176cc88, 0x254c02f5);
+	&_data_word(0xfcd7e54f, 0xd7cb2ac5, 0x80443526, 0x8fa362b5);
+	&_data_word(0x495ab1de, 0x671bba25, 0x980eea45, 0xe1c0fe5d);
+	&_data_word(0x02752fc3, 0x12f04c81, 0xa397468d, 0xc6f9d36b);
+	&_data_word(0xe75f8f03, 0x959c9215, 0xeb7a6dbf, 0xda595295);
+	&_data_word(0x2d83bed4, 0xd3217458, 0x2969e049, 0x44c8c98e);
+	&_data_word(0x6a89c275, 0x78798ef4, 0x6b3e5899, 0xdd71b927);
+	&_data_word(0xb64fe1be, 0x17ad88f0, 0x66ac20c9, 0xb43ace7d);
+	&_data_word(0x184adf63, 0x82311ae5, 0x60335197, 0x457f5362);
+	&_data_word(0xe07764b1, 0x84ae6bbb, 0x1ca081fe, 0x942b08f9);
+	&_data_word(0x58684870, 0x19fd458f, 0x876cde94, 0xb7f87b52);
+	&_data_word(0x23d373ab, 0xe2024b72, 0x578f1fe3, 0x2aab5566);
+	&_data_word(0x0728ebb2, 0x03c2b52f, 0x9a7bc586, 0xa50837d3);
+	&_data_word(0xf2872830, 0xb2a5bf23, 0xba6a0302, 0x5c8216ed);
+	&_data_word(0x2b1ccf8a, 0x92b479a7, 0xf0f207f3, 0xa1e2694e);
+	&_data_word(0xcdf4da65, 0xd5be0506, 0x1f6234d1, 0x8afea6c4);
+	&_data_word(0x9d532e34, 0xa055f3a2, 0x32e18a05, 0x75ebf6a4);
+	&_data_word(0x39ec830b, 0xaaef6040, 0x069f715e, 0x51106ebd);
+	&_data_word(0xf98a213e, 0x3d06dd96, 0xae053edd, 0x46bde64d);
+	&_data_word(0xb58d5491, 0x055dc471, 0x6fd40604, 0xff155060);
+	&_data_word(0x24fb9819, 0x97e9bdd6, 0xcc434089, 0x779ed967);
+	&_data_word(0xbd42e8b0, 0x888b8907, 0x385b19e7, 0xdbeec879);
+	&_data_word(0x470a7ca1, 0xe90f427c, 0xc91e84f8, 0x00000000);
+	&_data_word(0x83868009, 0x48ed2b32, 0xac70111e, 0x4e725a6c);
+	&_data_word(0xfbff0efd, 0x5638850f, 0x1ed5ae3d, 0x27392d36);
+	&_data_word(0x64d90f0a, 0x21a65c68, 0xd1545b9b, 0x3a2e3624);
+	&_data_word(0xb1670a0c, 0x0fe75793, 0xd296eeb4, 0x9e919b1b);
+	&_data_word(0x4fc5c080, 0xa220dc61, 0x694b775a, 0x161a121c);
+	&_data_word(0x0aba93e2, 0xe52aa0c0, 0x43e0223c, 0x1d171b12);
+	&_data_word(0x0b0d090e, 0xadc78bf2, 0xb9a8b62d, 0xc8a91e14);
+	&_data_word(0x8519f157, 0x4c0775af, 0xbbdd99ee, 0xfd607fa3);
+	&_data_word(0x9f2601f7, 0xbcf5725c, 0xc53b6644, 0x347efb5b);
+	&_data_word(0x7629438b, 0xdcc623cb, 0x68fcedb6, 0x63f1e4b8);
+	&_data_word(0xcadc31d7, 0x10856342, 0x40229713, 0x2011c684);
+	&_data_word(0x7d244a85, 0xf83dbbd2, 0x1132f9ae, 0x6da129c7);
+	&_data_word(0x4b2f9e1d, 0xf330b2dc, 0xec52860d, 0xd0e3c177);
+	&_data_word(0x6c16b32b, 0x99b970a9, 0xfa489411, 0x2264e947);
+	&_data_word(0xc48cfca8, 0x1a3ff0a0, 0xd82c7d56, 0xef903322);
+	&_data_word(0xc74e4987, 0xc1d138d9, 0xfea2ca8c, 0x360bd498);
+	&_data_word(0xcf81f5a6, 0x28de7aa5, 0x268eb7da, 0xa4bfad3f);
+	&_data_word(0xe49d3a2c, 0x0d927850, 0x9bcc5f6a, 0x62467e54);
+	&_data_word(0xc2138df6, 0xe8b8d890, 0x5ef7392e, 0xf5afc382);
+	&_data_word(0xbe805d9f, 0x7c93d069, 0xa92dd56f, 0xb31225cf);
+	&_data_word(0x3b99acc8, 0xa77d1810, 0x6e639ce8, 0x7bbb3bdb);
+	&_data_word(0x097826cd, 0xf418596e, 0x01b79aec, 0xa89a4f83);
+	&_data_word(0x656e95e6, 0x7ee6ffaa, 0x08cfbc21, 0xe6e815ef);
+	&_data_word(0xd99be7ba, 0xce366f4a, 0xd4099fea, 0xd67cb029);
+	&_data_word(0xafb2a431, 0x31233f2a, 0x3094a5c6, 0xc066a235);
+	&_data_word(0x37bc4e74, 0xa6ca82fc, 0xb0d090e0, 0x15d8a733);
+	&_data_word(0x4a9804f1, 0xf7daec41, 0x0e50cd7f, 0x2ff69117);
+	&_data_word(0x8dd64d76, 0x4db0ef43, 0x544daacc, 0xdf0496e4);
+	&_data_word(0xe3b5d19e, 0x1b886a4c, 0xb81f2cc1, 0x7f516546);
+	&_data_word(0x04ea5e9d, 0x5d358c01, 0x737487fa, 0x2e410bfb);
+	&_data_word(0x5a1d67b3, 0x52d2db92, 0x335610e9, 0x1347d66d);
+	&_data_word(0x8c61d79a, 0x7a0ca137, 0x8e14f859, 0x893c13eb);
+	&_data_word(0xee27a9ce, 0x35c961b7, 0xede51ce1, 0x3cb1477a);
+	&_data_word(0x59dfd29c, 0x3f73f255, 0x79ce1418, 0xbf37c773);
+	&_data_word(0xeacdf753, 0x5baafd5f, 0x146f3ddf, 0x86db4478);
+	&_data_word(0x81f3afca, 0x3ec468b9, 0x2c342438, 0x5f40a3c2);
+	&_data_word(0x72c31d16, 0x0c25e2bc, 0x8b493c28, 0x41950dff);
+	&_data_word(0x7101a839, 0xdeb30c08, 0x9ce4b4d8, 0x90c15664);
+	&_data_word(0x6184cb7b, 0x70b632d5, 0x745c6c48, 0x4257b8d0);
+
+#Td4:	# four copies of Td4 to choose from to avoid L1 aliasing
+	&data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38);
+	&data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb);
+	&data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87);
+	&data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb);
+	&data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d);
+	&data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e);
+	&data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2);
+	&data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25);
+	&data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16);
+	&data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92);
+	&data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda);
+	&data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84);
+	&data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a);
+	&data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06);
+	&data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02);
+	&data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b);
+	&data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea);
+	&data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73);
+	&data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85);
+	&data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e);
+	&data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89);
+	&data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b);
+	&data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20);
+	&data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4);
+	&data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31);
+	&data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f);
+	&data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d);
+	&data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef);
+	&data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0);
+	&data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61);
+	&data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26);
+	&data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d);
+$code.=<<___;
+	.long	0x80808080, 0x80808080, 0xfefefefe, 0xfefefefe
+	.long	0x1b1b1b1b, 0x1b1b1b1b, 0, 0
+___
+	&data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38);
+	&data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb);
+	&data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87);
+	&data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb);
+	&data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d);
+	&data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e);
+	&data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2);
+	&data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25);
+	&data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16);
+	&data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92);
+	&data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda);
+	&data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84);
+	&data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a);
+	&data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06);
+	&data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02);
+	&data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b);
+	&data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea);
+	&data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73);
+	&data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85);
+	&data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e);
+	&data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89);
+	&data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b);
+	&data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20);
+	&data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4);
+	&data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31);
+	&data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f);
+	&data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d);
+	&data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef);
+	&data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0);
+	&data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61);
+	&data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26);
+	&data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d);
+$code.=<<___;
+	.long	0x80808080, 0x80808080, 0xfefefefe, 0xfefefefe
+	.long	0x1b1b1b1b, 0x1b1b1b1b, 0, 0
+___
+	&data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38);
+	&data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb);
+	&data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87);
+	&data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb);
+	&data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d);
+	&data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e);
+	&data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2);
+	&data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25);
+	&data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16);
+	&data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92);
+	&data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda);
+	&data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84);
+	&data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a);
+	&data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06);
+	&data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02);
+	&data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b);
+	&data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea);
+	&data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73);
+	&data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85);
+	&data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e);
+	&data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89);
+	&data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b);
+	&data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20);
+	&data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4);
+	&data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31);
+	&data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f);
+	&data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d);
+	&data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef);
+	&data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0);
+	&data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61);
+	&data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26);
+	&data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d);
+$code.=<<___;
+	.long	0x80808080, 0x80808080, 0xfefefefe, 0xfefefefe
+	.long	0x1b1b1b1b, 0x1b1b1b1b, 0, 0
+___
+	&data_byte(0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38);
+	&data_byte(0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb);
+	&data_byte(0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87);
+	&data_byte(0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb);
+	&data_byte(0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d);
+	&data_byte(0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e);
+	&data_byte(0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2);
+	&data_byte(0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25);
+	&data_byte(0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16);
+	&data_byte(0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92);
+	&data_byte(0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda);
+	&data_byte(0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84);
+	&data_byte(0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a);
+	&data_byte(0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06);
+	&data_byte(0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02);
+	&data_byte(0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b);
+	&data_byte(0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea);
+	&data_byte(0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73);
+	&data_byte(0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85);
+	&data_byte(0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e);
+	&data_byte(0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89);
+	&data_byte(0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b);
+	&data_byte(0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20);
+	&data_byte(0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4);
+	&data_byte(0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31);
+	&data_byte(0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f);
+	&data_byte(0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d);
+	&data_byte(0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef);
+	&data_byte(0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0);
+	&data_byte(0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61);
+	&data_byte(0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26);
+	&data_byte(0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d);
+$code.=<<___;
+	.long	0x80808080, 0x80808080, 0xfefefefe, 0xfefefefe
+	.long	0x1b1b1b1b, 0x1b1b1b1b, 0, 0
+.asciz  "AES for x86_64, CRYPTOGAMS by <appro\@openssl.org>"
+.align	64
+___
+
+# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame,
+#		CONTEXT *context,DISPATCHER_CONTEXT *disp)
+if ($win64) {
+$rec="%rcx";
+$frame="%rdx";
+$context="%r8";
+$disp="%r9";
+
+$code.=<<___;
+.extern	__imp_RtlVirtualUnwind
+.type	block_se_handler,\@abi-omnipotent
+.align	16
+block_se_handler:
+	push	%rsi
+	push	%rdi
+	push	%rbx
+	push	%rbp
+	push	%r12
+	push	%r13
+	push	%r14
+	push	%r15
+	pushfq
+	sub	\$64,%rsp
+
+	mov	120($context),%rax	# pull context->Rax
+	mov	248($context),%rbx	# pull context->Rip
+
+	mov	8($disp),%rsi		# disp->ImageBase
+	mov	56($disp),%r11		# disp->HandlerData
+
+	mov	0(%r11),%r10d		# HandlerData[0]
+	lea	(%rsi,%r10),%r10	# prologue label
+	cmp	%r10,%rbx		# context->Rip<prologue label
+	jb	.Lin_block_prologue
+
+	mov	152($context),%rax	# pull context->Rsp
+
+	mov	4(%r11),%r10d		# HandlerData[1]
+	lea	(%rsi,%r10),%r10	# epilogue label
+	cmp	%r10,%rbx		# context->Rip>=epilogue label
+	jae	.Lin_block_prologue
+
+	mov	24(%rax),%rax		# pull saved real stack pointer
+	lea	48(%rax),%rax		# adjust...
+
+	mov	-8(%rax),%rbx
+	mov	-16(%rax),%rbp
+	mov	-24(%rax),%r12
+	mov	-32(%rax),%r13
+	mov	-40(%rax),%r14
+	mov	-48(%rax),%r15
+	mov	%rbx,144($context)	# restore context->Rbx
+	mov	%rbp,160($context)	# restore context->Rbp
+	mov	%r12,216($context)	# restore context->R12
+	mov	%r13,224($context)	# restore context->R13
+	mov	%r14,232($context)	# restore context->R14
+	mov	%r15,240($context)	# restore context->R15
+
+.Lin_block_prologue:
+	mov	8(%rax),%rdi
+	mov	16(%rax),%rsi
+	mov	%rax,152($context)	# restore context->Rsp
+	mov	%rsi,168($context)	# restore context->Rsi
+	mov	%rdi,176($context)	# restore context->Rdi
+
+	jmp	.Lcommon_seh_exit
+.size	block_se_handler,.-block_se_handler
+
+.type	key_se_handler,\@abi-omnipotent
+.align	16
+key_se_handler:
+	push	%rsi
+	push	%rdi
+	push	%rbx
+	push	%rbp
+	push	%r12
+	push	%r13
+	push	%r14
+	push	%r15
+	pushfq
+	sub	\$64,%rsp
+
+	mov	120($context),%rax	# pull context->Rax
+	mov	248($context),%rbx	# pull context->Rip
+
+	mov	8($disp),%rsi		# disp->ImageBase
+	mov	56($disp),%r11		# disp->HandlerData
+
+	mov	0(%r11),%r10d		# HandlerData[0]
+	lea	(%rsi,%r10),%r10	# prologue label
+	cmp	%r10,%rbx		# context->Rip<prologue label
+	jb	.Lin_key_prologue
+
+	mov	152($context),%rax	# pull context->Rsp
+
+	mov	4(%r11),%r10d		# HandlerData[1]
+	lea	(%rsi,%r10),%r10	# epilogue label
+	cmp	%r10,%rbx		# context->Rip>=epilogue label
+	jae	.Lin_key_prologue
+
+	lea	56(%rax),%rax
+
+	mov	-8(%rax),%rbx
+	mov	-16(%rax),%rbp
+	mov	-24(%rax),%r12
+	mov	-32(%rax),%r13
+	mov	-40(%rax),%r14
+	mov	-48(%rax),%r15
+	mov	%rbx,144($context)	# restore context->Rbx
+	mov	%rbp,160($context)	# restore context->Rbp
+	mov	%r12,216($context)	# restore context->R12
+	mov	%r13,224($context)	# restore context->R13
+	mov	%r14,232($context)	# restore context->R14
+	mov	%r15,240($context)	# restore context->R15
+
+.Lin_key_prologue:
+	mov	8(%rax),%rdi
+	mov	16(%rax),%rsi
+	mov	%rax,152($context)	# restore context->Rsp
+	mov	%rsi,168($context)	# restore context->Rsi
+	mov	%rdi,176($context)	# restore context->Rdi
+
+	jmp	.Lcommon_seh_exit
+.size	key_se_handler,.-key_se_handler
+
+.type	cbc_se_handler,\@abi-omnipotent
+.align	16
+cbc_se_handler:
+	push	%rsi
+	push	%rdi
+	push	%rbx
+	push	%rbp
+	push	%r12
+	push	%r13
+	push	%r14
+	push	%r15
+	pushfq
+	sub	\$64,%rsp
+
+	mov	120($context),%rax	# pull context->Rax
+	mov	248($context),%rbx	# pull context->Rip
+
+	lea	.Lcbc_prologue(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip<.Lcbc_prologue
+	jb	.Lin_cbc_prologue
+
+	lea	.Lcbc_fast_body(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip<.Lcbc_fast_body
+	jb	.Lin_cbc_frame_setup
+
+	lea	.Lcbc_slow_prologue(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip<.Lcbc_slow_prologue
+	jb	.Lin_cbc_body
+
+	lea	.Lcbc_slow_body(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip<.Lcbc_slow_body
+	jb	.Lin_cbc_frame_setup
+
+.Lin_cbc_body:
+	mov	152($context),%rax	# pull context->Rsp
+
+	lea	.Lcbc_epilogue(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip>=.Lcbc_epilogue
+	jae	.Lin_cbc_prologue
+
+	lea	8(%rax),%rax
+
+	lea	.Lcbc_popfq(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip>=.Lcbc_popfq
+	jae	.Lin_cbc_prologue
+
+	mov	`16-8`(%rax),%rax	# biased $_rsp
+	lea	56(%rax),%rax
+
+.Lin_cbc_frame_setup:
+	mov	-16(%rax),%rbx
+	mov	-24(%rax),%rbp
+	mov	-32(%rax),%r12
+	mov	-40(%rax),%r13
+	mov	-48(%rax),%r14
+	mov	-56(%rax),%r15
+	mov	%rbx,144($context)	# restore context->Rbx
+	mov	%rbp,160($context)	# restore context->Rbp
+	mov	%r12,216($context)	# restore context->R12
+	mov	%r13,224($context)	# restore context->R13
+	mov	%r14,232($context)	# restore context->R14
+	mov	%r15,240($context)	# restore context->R15
+
+.Lin_cbc_prologue:
+	mov	8(%rax),%rdi
+	mov	16(%rax),%rsi
+	mov	%rax,152($context)	# restore context->Rsp
+	mov	%rsi,168($context)	# restore context->Rsi
+	mov	%rdi,176($context)	# restore context->Rdi
+
+.Lcommon_seh_exit:
+
+	mov	40($disp),%rdi		# disp->ContextRecord
+	mov	$context,%rsi		# context
+	mov	\$`1232/8`,%ecx		# sizeof(CONTEXT)
+	.long	0xa548f3fc		# cld; rep movsq
+
+	mov	$disp,%rsi
+	xor	%rcx,%rcx		# arg1, UNW_FLAG_NHANDLER
+	mov	8(%rsi),%rdx		# arg2, disp->ImageBase
+	mov	0(%rsi),%r8		# arg3, disp->ControlPc
+	mov	16(%rsi),%r9		# arg4, disp->FunctionEntry
+	mov	40(%rsi),%r10		# disp->ContextRecord
+	lea	56(%rsi),%r11		# &disp->HandlerData
+	lea	24(%rsi),%r12		# &disp->EstablisherFrame
+	mov	%r10,32(%rsp)		# arg5
+	mov	%r11,40(%rsp)		# arg6
+	mov	%r12,48(%rsp)		# arg7
+	mov	%rcx,56(%rsp)		# arg8, (NULL)
+	call	*__imp_RtlVirtualUnwind(%rip)
+
+	mov	\$1,%eax		# ExceptionContinueSearch
+	add	\$64,%rsp
+	popfq
+	pop	%r15
+	pop	%r14
+	pop	%r13
+	pop	%r12
+	pop	%rbp
+	pop	%rbx
+	pop	%rdi
+	pop	%rsi
+	ret
+.size	cbc_se_handler,.-cbc_se_handler
+
+.section	.pdata
+.align	4
+	.rva	.LSEH_begin_AES_encrypt
+	.rva	.LSEH_end_AES_encrypt
+	.rva	.LSEH_info_AES_encrypt
+
+	.rva	.LSEH_begin_AES_decrypt
+	.rva	.LSEH_end_AES_decrypt
+	.rva	.LSEH_info_AES_decrypt
+
+	.rva	.LSEH_begin_AES_set_encrypt_key
+	.rva	.LSEH_end_AES_set_encrypt_key
+	.rva	.LSEH_info_AES_set_encrypt_key
+
+	.rva	.LSEH_begin_AES_set_decrypt_key
+	.rva	.LSEH_end_AES_set_decrypt_key
+	.rva	.LSEH_info_AES_set_decrypt_key
+
+	.rva	.LSEH_begin_AES_cbc_encrypt
+	.rva	.LSEH_end_AES_cbc_encrypt
+	.rva	.LSEH_info_AES_cbc_encrypt
+
+.section	.xdata
+.align	8
+.LSEH_info_AES_encrypt:
+	.byte	9,0,0,0
+	.rva	block_se_handler
+	.rva	.Lenc_prologue,.Lenc_epilogue	# HandlerData[]
+.LSEH_info_AES_decrypt:
+	.byte	9,0,0,0
+	.rva	block_se_handler
+	.rva	.Ldec_prologue,.Ldec_epilogue	# HandlerData[]
+.LSEH_info_AES_set_encrypt_key:
+	.byte	9,0,0,0
+	.rva	key_se_handler
+	.rva	.Lenc_key_prologue,.Lenc_key_epilogue	# HandlerData[]
+.LSEH_info_AES_set_decrypt_key:
+	.byte	9,0,0,0
+	.rva	key_se_handler
+	.rva	.Ldec_key_prologue,.Ldec_key_epilogue	# HandlerData[]
+.LSEH_info_AES_cbc_encrypt:
+	.byte	9,0,0,0
+	.rva	cbc_se_handler
+___
+}
+
+$code =~ s/\`([^\`]*)\`/eval($1)/gem;
+
+print $code;
+
+close STDOUT;