Update to openssl-1.0.1

18 files changed, 4964 insertions, 605 deletions

diff --git a/openssl/crypto/sha/asm/sha1-586.pl b/openssl/crypto/sha/asm/sha1-586.pl
index a1f876281..1084d227f 100644
--- a/openssl/crypto/sha/asm/sha1-586.pl
+++ b/openssl/crypto/sha/asm/sha1-586.pl
@@ -12,6 +12,8 @@
 # commentary below], and in 2006 the rest was rewritten in order to
 # gain freedom to liberate licensing terms.
 
+# January, September 2004.
+#
 # It was noted that Intel IA-32 C compiler generates code which
 # performs ~30% *faster* on P4 CPU than original *hand-coded*
 # SHA1 assembler implementation. To address this problem (and
@@ -31,12 +33,92 @@
 # ----------------------------------------------------------------
 #					<appro@fy.chalmers.se>
 
+# August 2009.
+#
+# George Spelvin has tipped that F_40_59(b,c,d) can be rewritten as
+# '(c&d) + (b&(c^d))', which allows to accumulate partial results
+# and lighten "pressure" on scratch registers. This resulted in
+# >12% performance improvement on contemporary AMD cores (with no
+# degradation on other CPUs:-). Also, the code was revised to maximize
+# "distance" between instructions producing input to 'lea' instruction
+# and the 'lea' instruction itself, which is essential for Intel Atom
+# core and resulted in ~15% improvement.
+
+# October 2010.
+#
+# Add SSSE3, Supplemental[!] SSE3, implementation. The idea behind it
+# is to offload message schedule denoted by Wt in NIST specification,
+# or Xupdate in OpenSSL source, to SIMD unit. The idea is not novel,
+# and in SSE2 context was first explored by Dean Gaudet in 2004, see
+# http://arctic.org/~dean/crypto/sha1.html. Since then several things
+# have changed that made it interesting again:
+#
+# a) XMM units became faster and wider;
+# b) instruction set became more versatile;
+# c) an important observation was made by Max Locktykhin, which made
+#    it possible to reduce amount of instructions required to perform
+#    the operation in question, for further details see
+#    http://software.intel.com/en-us/articles/improving-the-performance-of-the-secure-hash-algorithm-1/.
+
+# April 2011.
+#
+# Add AVX code path, probably most controversial... The thing is that
+# switch to AVX alone improves performance by as little as 4% in
+# comparison to SSSE3 code path. But below result doesn't look like
+# 4% improvement... Trouble is that Sandy Bridge decodes 'ro[rl]' as
+# pair of µ-ops, and it's the additional µ-ops, two per round, that
+# make it run slower than Core2 and Westmere. But 'sh[rl]d' is decoded
+# as single µ-op by Sandy Bridge and it's replacing 'ro[rl]' with
+# equivalent 'sh[rl]d' that is responsible for the impressive 5.1
+# cycles per processed byte. But 'sh[rl]d' is not something that used
+# to be fast, nor does it appear to be fast in upcoming Bulldozer
+# [according to its optimization manual]. Which is why AVX code path
+# is guarded by *both* AVX and synthetic bit denoting Intel CPUs.
+# One can argue that it's unfair to AMD, but without 'sh[rl]d' it
+# makes no sense to keep the AVX code path. If somebody feels that
+# strongly, it's probably more appropriate to discuss possibility of
+# using vector rotate XOP on AMD...
+
+######################################################################
+# Current performance is summarized in following table. Numbers are
+# CPU clock cycles spent to process single byte (less is better).
+#
+#		x86		SSSE3		AVX
+# Pentium	15.7		-
+# PIII		11.5		-
+# P4		10.6		-
+# AMD K8	7.1		-
+# Core2		7.3		6.1/+20%	-
+# Atom		12.5		9.5(*)/+32%	-
+# Westmere	7.3		5.6/+30%	-
+# Sandy Bridge	8.8		6.2/+40%	5.1(**)/+70%
+#
+# (*)	Loop is 1056 instructions long and expected result is ~8.25.
+#	It remains mystery [to me] why ILP is limited to 1.7.
+#
+# (**)	As per above comment, the result is for AVX *plus* sh[rl]d.
+
 $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
 push(@INC,"${dir}","${dir}../../perlasm");
 require "x86asm.pl";
 
 &asm_init($ARGV[0],"sha1-586.pl",$ARGV[$#ARGV] eq "386");
 
+$xmm=$ymm=0;
+for (@ARGV) { $xmm=1 if (/-DOPENSSL_IA32_SSE2/); }
+
+$ymm=1 if ($xmm &&
+		`$ENV{CC} -Wa,-v -c -o /dev/null -x assembler /dev/null 2>&1`
+			=~ /GNU assembler version ([2-9]\.[0-9]+)/ &&
+		$1>=2.19);	# first version supporting AVX
+
+$ymm=1 if ($xmm && !$ymm && $ARGV[0] eq "win32n" && 
+		`nasm -v 2>&1` =~ /NASM version ([2-9]\.[0-9]+)/ &&
+		$1>=2.03);	# first version supporting AVX
+
+&external_label("OPENSSL_ia32cap_P") if ($xmm);
+
+
 $A="eax";
 $B="ebx";
 $C="ecx";
@@ -47,6 +129,10 @@ $tmp1="ebp";
 
 @V=($A,$B,$C,$D,$E,$T);
 
+$alt=0;	# 1 denotes alternative IALU implementation, which performs
+	# 8% *worse* on P4, same on Westmere and Atom, 2% better on
+	# Sandy Bridge...
+
 sub BODY_00_15
 	{
 	local($n,$a,$b,$c,$d,$e,$f)=@_;
@@ -59,16 +145,18 @@ sub BODY_00_15
 	&rotl($tmp1,5);			# tmp1=ROTATE(a,5)
 	 &xor($f,$d);
 	&add($tmp1,$e);			# tmp1+=e;
-	 &and($f,$b);
-	&mov($e,&swtmp($n%16));		# e becomes volatile and is loaded
+	 &mov($e,&swtmp($n%16));	# e becomes volatile and is loaded
 	 				# with xi, also note that e becomes
 					# f in next round...
-	 &xor($f,$d);			# f holds F_00_19(b,c,d)
+	&and($f,$b);
 	&rotr($b,2);			# b=ROTATE(b,30)
-	 &lea($tmp1,&DWP(0x5a827999,$tmp1,$e));	# tmp1+=K_00_19+xi
+	 &xor($f,$d);			# f holds F_00_19(b,c,d)
+	&lea($tmp1,&DWP(0x5a827999,$tmp1,$e));	# tmp1+=K_00_19+xi
 
-	if ($n==15) { &add($f,$tmp1); }	# f+=tmp1
+	if ($n==15) { &mov($e,&swtmp(($n+1)%16));# pre-fetch f for next round
+		      &add($f,$tmp1); }	# f+=tmp1
 	else        { &add($tmp1,$f); }	# f becomes a in next round
+	&mov($tmp1,$a)			if ($alt && $n==15);
 	}
 
 sub BODY_16_19
@@ -77,22 +165,41 @@ sub BODY_16_19
 
 	&comment("16_19 $n");
 
-	&mov($f,&swtmp($n%16));		# f to hold Xupdate(xi,xa,xb,xc,xd)
-	 &mov($tmp1,$c);		# tmp1 to hold F_00_19(b,c,d)
-	&xor($f,&swtmp(($n+2)%16));
-	 &xor($tmp1,$d);
-	&xor($f,&swtmp(($n+8)%16));
-	 &and($tmp1,$b);		# tmp1 holds F_00_19(b,c,d)
-	&rotr($b,2);			# b=ROTATE(b,30)
+if ($alt) {
+	&xor($c,$d);
+	 &xor($f,&swtmp(($n+2)%16));	# f to hold Xupdate(xi,xa,xb,xc,xd)
+	&and($tmp1,$c);			# tmp1 to hold F_00_19(b,c,d), b&=c^d
+	 &xor($f,&swtmp(($n+8)%16));
+	&xor($tmp1,$d);			# tmp1=F_00_19(b,c,d)
+	 &xor($f,&swtmp(($n+13)%16));	# f holds xa^xb^xc^xd
+	&rotl($f,1);			# f=ROTATE(f,1)
+	 &add($e,$tmp1);		# e+=F_00_19(b,c,d)
+	&xor($c,$d);			# restore $c
+	 &mov($tmp1,$a);		# b in next round
+	&rotr($b,$n==16?2:7);		# b=ROTATE(b,30)
+	 &mov(&swtmp($n%16),$f);	# xi=f
+	&rotl($a,5);			# ROTATE(a,5)
+	 &lea($f,&DWP(0x5a827999,$f,$e));# f+=F_00_19(b,c,d)+e
+	&mov($e,&swtmp(($n+1)%16));	# pre-fetch f for next round
+	 &add($f,$a);			# f+=ROTATE(a,5)
+} else {
+	&mov($tmp1,$c);			# tmp1 to hold F_00_19(b,c,d)
+	 &xor($f,&swtmp(($n+2)%16));	# f to hold Xupdate(xi,xa,xb,xc,xd)
+	&xor($tmp1,$d);
+	 &xor($f,&swtmp(($n+8)%16));
+	&and($tmp1,$b);
 	 &xor($f,&swtmp(($n+13)%16));	# f holds xa^xb^xc^xd
 	&rotl($f,1);			# f=ROTATE(f,1)
 	 &xor($tmp1,$d);		# tmp1=F_00_19(b,c,d)
-	&mov(&swtmp($n%16),$f);		# xi=f
-	&lea($f,&DWP(0x5a827999,$f,$e));# f+=K_00_19+e
-	 &mov($e,$a);			# e becomes volatile
-	&rotl($e,5);			# e=ROTATE(a,5)
-	 &add($f,$tmp1);		# f+=F_00_19(b,c,d)
-	&add($f,$e);			# f+=ROTATE(a,5)
+	&add($e,$tmp1);			# e+=F_00_19(b,c,d)
+	 &mov($tmp1,$a);
+	&rotr($b,2);			# b=ROTATE(b,30)
+	 &mov(&swtmp($n%16),$f);	# xi=f
+	&rotl($tmp1,5);			# ROTATE(a,5)
+	 &lea($f,&DWP(0x5a827999,$f,$e));# f+=F_00_19(b,c,d)+e
+	&mov($e,&swtmp(($n+1)%16));	# pre-fetch f for next round
+	 &add($f,$tmp1);		# f+=ROTATE(a,5)
+}
 	}
 
 sub BODY_20_39
@@ -102,21 +209,41 @@ sub BODY_20_39
 
 	&comment("20_39 $n");
 
+if ($alt) {
+	&xor($tmp1,$c);			# tmp1 to hold F_20_39(b,c,d), b^=c
+	 &xor($f,&swtmp(($n+2)%16));	# f to hold Xupdate(xi,xa,xb,xc,xd)
+	&xor($tmp1,$d);			# tmp1 holds F_20_39(b,c,d)
+	 &xor($f,&swtmp(($n+8)%16));
+	&add($e,$tmp1);			# e+=F_20_39(b,c,d)
+	 &xor($f,&swtmp(($n+13)%16));	# f holds xa^xb^xc^xd
+	&rotl($f,1);			# f=ROTATE(f,1)
+	 &mov($tmp1,$a);		# b in next round
+	&rotr($b,7);			# b=ROTATE(b,30)
+	 &mov(&swtmp($n%16),$f)		if($n<77);# xi=f
+	&rotl($a,5);			# ROTATE(a,5)
+	 &xor($b,$c)			if($n==39);# warm up for BODY_40_59
+	&and($tmp1,$b)			if($n==39);
+	 &lea($f,&DWP($K,$f,$e));	# f+=e+K_XX_YY
+	&mov($e,&swtmp(($n+1)%16))	if($n<79);# pre-fetch f for next round
+	 &add($f,$a);			# f+=ROTATE(a,5)
+	&rotr($a,5)			if ($n==79);
+} else {
 	&mov($tmp1,$b);			# tmp1 to hold F_20_39(b,c,d)
-	 &mov($f,&swtmp($n%16));	# f to hold Xupdate(xi,xa,xb,xc,xd)
-	&rotr($b,2);			# b=ROTATE(b,30)
-	 &xor($f,&swtmp(($n+2)%16));
+	 &xor($f,&swtmp(($n+2)%16));	# f to hold Xupdate(xi,xa,xb,xc,xd)
 	&xor($tmp1,$c);
 	 &xor($f,&swtmp(($n+8)%16));
 	&xor($tmp1,$d);			# tmp1 holds F_20_39(b,c,d)
 	 &xor($f,&swtmp(($n+13)%16));	# f holds xa^xb^xc^xd
 	&rotl($f,1);			# f=ROTATE(f,1)
-	 &add($tmp1,$e);
-	&mov(&swtmp($n%16),$f);		# xi=f
-	 &mov($e,$a);			# e becomes volatile
-	&rotl($e,5);			# e=ROTATE(a,5)
-	 &lea($f,&DWP($K,$f,$tmp1));	# f+=K_20_39+e
-	&add($f,$e);			# f+=ROTATE(a,5)
+	 &add($e,$tmp1);		# e+=F_20_39(b,c,d)
+	&rotr($b,2);			# b=ROTATE(b,30)
+	 &mov($tmp1,$a);
+	&rotl($tmp1,5);			# ROTATE(a,5)
+	 &mov(&swtmp($n%16),$f) if($n<77);# xi=f
+	&lea($f,&DWP($K,$f,$e));	# f+=e+K_XX_YY
+	 &mov($e,&swtmp(($n+1)%16)) if($n<79);# pre-fetch f for next round
+	&add($f,$tmp1);			# f+=ROTATE(a,5)
+}
 	}
 
 sub BODY_40_59
@@ -125,41 +252,86 @@ sub BODY_40_59
 
 	&comment("40_59 $n");
 
-	&mov($f,&swtmp($n%16));		# f to hold Xupdate(xi,xa,xb,xc,xd)
-	 &mov($tmp1,&swtmp(($n+2)%16));
-	&xor($f,$tmp1);
-	 &mov($tmp1,&swtmp(($n+8)%16));
-	&xor($f,$tmp1);
-	 &mov($tmp1,&swtmp(($n+13)%16));
-	&xor($f,$tmp1);			# f holds xa^xb^xc^xd
-	 &mov($tmp1,$b);		# tmp1 to hold F_40_59(b,c,d)
+if ($alt) {
+	&add($e,$tmp1);			# e+=b&(c^d)
+	 &xor($f,&swtmp(($n+2)%16));	# f to hold Xupdate(xi,xa,xb,xc,xd)
+	&mov($tmp1,$d);
+	 &xor($f,&swtmp(($n+8)%16));
+	&xor($c,$d);			# restore $c
+	 &xor($f,&swtmp(($n+13)%16));	# f holds xa^xb^xc^xd
 	&rotl($f,1);			# f=ROTATE(f,1)
-	 &or($tmp1,$c);
-	&mov(&swtmp($n%16),$f);		# xi=f
-	 &and($tmp1,$d);
-	&lea($f,&DWP(0x8f1bbcdc,$f,$e));# f+=K_40_59+e
-	 &mov($e,$b);			# e becomes volatile and is used
-					# to calculate F_40_59(b,c,d)
+	 &and($tmp1,$c);
+	&rotr($b,7);			# b=ROTATE(b,30)
+	 &add($e,$tmp1);		# e+=c&d
+	&mov($tmp1,$a);			# b in next round
+	 &mov(&swtmp($n%16),$f);	# xi=f
+	&rotl($a,5);			# ROTATE(a,5)
+	 &xor($b,$c)			if ($n<59);
+	&and($tmp1,$b)			if ($n<59);# tmp1 to hold F_40_59(b,c,d)
+	 &lea($f,&DWP(0x8f1bbcdc,$f,$e));# f+=K_40_59+e+(b&(c^d))
+	&mov($e,&swtmp(($n+1)%16));	# pre-fetch f for next round
+	 &add($f,$a);			# f+=ROTATE(a,5)
+} else {
+	&mov($tmp1,$c);			# tmp1 to hold F_40_59(b,c,d)
+	 &xor($f,&swtmp(($n+2)%16));	# f to hold Xupdate(xi,xa,xb,xc,xd)
+	&xor($tmp1,$d);
+	 &xor($f,&swtmp(($n+8)%16));
+	&and($tmp1,$b);
+	 &xor($f,&swtmp(($n+13)%16));	# f holds xa^xb^xc^xd
+	&rotl($f,1);			# f=ROTATE(f,1)
+	 &add($tmp1,$e);		# b&(c^d)+=e
 	&rotr($b,2);			# b=ROTATE(b,30)
-	 &and($e,$c);
-	&or($tmp1,$e);			# tmp1 holds F_40_59(b,c,d)		
-	 &mov($e,$a);
-	&rotl($e,5);			# e=ROTATE(a,5)
-	 &add($f,$tmp1);		# f+=tmp1;
+	 &mov($e,$a);			# e becomes volatile
+	&rotl($e,5);			# ROTATE(a,5)
+	 &mov(&swtmp($n%16),$f);	# xi=f
+	&lea($f,&DWP(0x8f1bbcdc,$f,$tmp1));# f+=K_40_59+e+(b&(c^d))
+	 &mov($tmp1,$c);
 	&add($f,$e);			# f+=ROTATE(a,5)
+	 &and($tmp1,$d);
+	&mov($e,&swtmp(($n+1)%16));	# pre-fetch f for next round
+	 &add($f,$tmp1);		# f+=c&d
+}
 	}
 
 &function_begin("sha1_block_data_order");
+if ($xmm) {
+  &static_label("ssse3_shortcut");
+  &static_label("avx_shortcut")		if ($ymm);
+  &static_label("K_XX_XX");
+
+	&call	(&label("pic_point"));	# make it PIC!
+  &set_label("pic_point");
+	&blindpop($tmp1);
+	&picmeup($T,"OPENSSL_ia32cap_P",$tmp1,&label("pic_point"));
+	&lea	($tmp1,&DWP(&label("K_XX_XX")."-".&label("pic_point"),$tmp1));
+
+	&mov	($A,&DWP(0,$T));
+	&mov	($D,&DWP(4,$T));
+	&test	($D,1<<9);		# check SSSE3 bit
+	&jz	(&label("x86"));
+	&test	($A,1<<24);		# check FXSR bit
+	&jz	(&label("x86"));
+	if ($ymm) {
+		&and	($D,1<<28);		# mask AVX bit
+		&and	($A,1<<30);		# mask "Intel CPU" bit
+		&or	($A,$D);
+		&cmp	($A,1<<28|1<<30);
+		&je	(&label("avx_shortcut"));
+	}
+	&jmp	(&label("ssse3_shortcut"));
+  &set_label("x86",16);
+}
 	&mov($tmp1,&wparam(0));	# SHA_CTX *c
 	&mov($T,&wparam(1));	# const void *input
 	&mov($A,&wparam(2));	# size_t num
-	&stack_push(16);	# allocate X[16]
+	&stack_push(16+3);	# allocate X[16]
 	&shl($A,6);
 	&add($A,$T);
 	&mov(&wparam(2),$A);	# pointer beyond the end of input
 	&mov($E,&DWP(16,$tmp1));# pre-load E
+	&jmp(&label("loop"));
 
-	&set_label("loop",16);
+&set_label("loop",16);
 
 	# copy input chunk to X, but reversing byte order!
 	for ($i=0; $i<16; $i+=4)
@@ -213,8 +385,845 @@ sub BODY_40_59
 	&mov(&DWP(16,$tmp1),$C);
 	&jb(&label("loop"));
 
-	&stack_pop(16);
+	&stack_pop(16+3);
 &function_end("sha1_block_data_order");
+
+if ($xmm) {
+######################################################################
+# The SSSE3 implementation.
+#
+# %xmm[0-7] are used as ring @X[] buffer containing quadruples of last
+# 32 elements of the message schedule or Xupdate outputs. First 4
+# quadruples are simply byte-swapped input, next 4 are calculated
+# according to method originally suggested by Dean Gaudet (modulo
+# being implemented in SSSE3). Once 8 quadruples or 32 elements are
+# collected, it switches to routine proposed by Max Locktyukhin.
+#
+# Calculations inevitably require temporary reqisters, and there are
+# no %xmm registers left to spare. For this reason part of the ring
+# buffer, X[2..4] to be specific, is offloaded to 3 quadriples ring
+# buffer on the stack. Keep in mind that X[2] is alias X[-6], X[3] -
+# X[-5], and X[4] - X[-4]...
+#
+# Another notable optimization is aggressive stack frame compression
+# aiming to minimize amount of 9-byte instructions...
+#
+# Yet another notable optimization is "jumping" $B variable. It means
+# that there is no register permanently allocated for $B value. This
+# allowed to eliminate one instruction from body_20_39...
+#
+my $Xi=4;			# 4xSIMD Xupdate round, start pre-seeded
+my @X=map("xmm$_",(4..7,0..3));	# pre-seeded for $Xi=4
+my @V=($A,$B,$C,$D,$E);
+my $j=0;			# hash round
+my @T=($T,$tmp1);
+my $inp;
+
+my $_rol=sub { &rol(@_) };
+my $_ror=sub { &ror(@_) };
+
+&function_begin("_sha1_block_data_order_ssse3");
+	&call	(&label("pic_point"));	# make it PIC!
+	&set_label("pic_point");
+	&blindpop($tmp1);
+	&lea	($tmp1,&DWP(&label("K_XX_XX")."-".&label("pic_point"),$tmp1));
+&set_label("ssse3_shortcut");
+
+	&movdqa	(@X[3],&QWP(0,$tmp1));		# K_00_19
+	&movdqa	(@X[4],&QWP(16,$tmp1));		# K_20_39
+	&movdqa	(@X[5],&QWP(32,$tmp1));		# K_40_59
+	&movdqa	(@X[6],&QWP(48,$tmp1));		# K_60_79
+	&movdqa	(@X[2],&QWP(64,$tmp1));		# pbswap mask
+
+	&mov	($E,&wparam(0));		# load argument block
+	&mov	($inp=@T[1],&wparam(1));
+	&mov	($D,&wparam(2));
+	&mov	(@T[0],"esp");
+
+	# stack frame layout
+	#
+	# +0	X[0]+K	X[1]+K	X[2]+K	X[3]+K	# XMM->IALU xfer area
+	#	X[4]+K	X[5]+K	X[6]+K	X[7]+K
+	#	X[8]+K	X[9]+K	X[10]+K	X[11]+K
+	#	X[12]+K	X[13]+K	X[14]+K	X[15]+K
+	#
+	# +64	X[0]	X[1]	X[2]	X[3]	# XMM->XMM backtrace area
+	#	X[4]	X[5]	X[6]	X[7]
+	#	X[8]	X[9]	X[10]	X[11]	# even borrowed for K_00_19
+	#
+	# +112	K_20_39	K_20_39	K_20_39	K_20_39	# constants
+	#	K_40_59	K_40_59	K_40_59	K_40_59
+	#	K_60_79	K_60_79	K_60_79	K_60_79
+	#	K_00_19	K_00_19	K_00_19	K_00_19
+	#	pbswap mask
+	#
+	# +192	ctx				# argument block
+	# +196	inp
+	# +200	end
+	# +204	esp
+	&sub	("esp",208);
+	&and	("esp",-64);
+
+	&movdqa	(&QWP(112+0,"esp"),@X[4]);	# copy constants
+	&movdqa	(&QWP(112+16,"esp"),@X[5]);
+	&movdqa	(&QWP(112+32,"esp"),@X[6]);
+	&shl	($D,6);				# len*64
+	&movdqa	(&QWP(112+48,"esp"),@X[3]);
+	&add	($D,$inp);			# end of input
+	&movdqa	(&QWP(112+64,"esp"),@X[2]);
+	&add	($inp,64);
+	&mov	(&DWP(192+0,"esp"),$E);		# save argument block
+	&mov	(&DWP(192+4,"esp"),$inp);
+	&mov	(&DWP(192+8,"esp"),$D);
+	&mov	(&DWP(192+12,"esp"),@T[0]);	# save original %esp
+
+	&mov	($A,&DWP(0,$E));		# load context
+	&mov	($B,&DWP(4,$E));
+	&mov	($C,&DWP(8,$E));
+	&mov	($D,&DWP(12,$E));
+	&mov	($E,&DWP(16,$E));
+	&mov	(@T[0],$B);			# magic seed
+
+	&movdqu	(@X[-4&7],&QWP(-64,$inp));	# load input to %xmm[0-3]
+	&movdqu	(@X[-3&7],&QWP(-48,$inp));
+	&movdqu	(@X[-2&7],&QWP(-32,$inp));
+	&movdqu	(@X[-1&7],&QWP(-16,$inp));
+	&pshufb	(@X[-4&7],@X[2]);		# byte swap
+	&pshufb	(@X[-3&7],@X[2]);
+	&pshufb	(@X[-2&7],@X[2]);
+	&movdqa	(&QWP(112-16,"esp"),@X[3]);	# borrow last backtrace slot
+	&pshufb	(@X[-1&7],@X[2]);
+	&paddd	(@X[-4&7],@X[3]);		# add K_00_19
+	&paddd	(@X[-3&7],@X[3]);
+	&paddd	(@X[-2&7],@X[3]);
+	&movdqa	(&QWP(0,"esp"),@X[-4&7]);	# X[]+K xfer to IALU
+	&psubd	(@X[-4&7],@X[3]);		# restore X[]
+	&movdqa	(&QWP(0+16,"esp"),@X[-3&7]);
+	&psubd	(@X[-3&7],@X[3]);
+	&movdqa	(&QWP(0+32,"esp"),@X[-2&7]);
+	&psubd	(@X[-2&7],@X[3]);
+	&movdqa	(@X[0],@X[-3&7]);
+	&jmp	(&label("loop"));
+
+######################################################################
+# SSE instruction sequence is first broken to groups of indepentent
+# instructions, independent in respect to their inputs and shifter
+# (not all architectures have more than one). Then IALU instructions
+# are "knitted in" between the SSE groups. Distance is maintained for
+# SSE latency of 2 in hope that it fits better upcoming AMD Bulldozer
+# [which allegedly also implements SSSE3]...
+#
+# Temporary registers usage. X[2] is volatile at the entry and at the
+# end is restored from backtrace ring buffer. X[3] is expected to
+# contain current K_XX_XX constant and is used to caclulate X[-1]+K
+# from previous round, it becomes volatile the moment the value is
+# saved to stack for transfer to IALU. X[4] becomes volatile whenever
+# X[-4] is accumulated and offloaded to backtrace ring buffer, at the
+# end it is loaded with next K_XX_XX [which becomes X[3] in next
+# round]...
+#
+sub Xupdate_ssse3_16_31()		# recall that $Xi starts wtih 4
+{ use integer;
+  my $body = shift;
+  my @insns = (&$body,&$body,&$body,&$body);	# 40 instructions
+  my ($a,$b,$c,$d,$e);
+
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&palignr(@X[0],@X[-4&7],8);	# compose "X[-14]" in "X[0]"
+	&movdqa	(@X[2],@X[-1&7]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	  &paddd	(@X[3],@X[-1&7]);
+	  &movdqa	(&QWP(64+16*(($Xi-4)%3),"esp"),@X[-4&7]);# save X[] to backtrace buffer
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&psrldq	(@X[2],4);		# "X[-3]", 3 dwords
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&pxor	(@X[0],@X[-4&7]);	# "X[0]"^="X[-16]"
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&pxor	(@X[2],@X[-2&7]);	# "X[-3]"^"X[-8]"
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&pxor	(@X[0],@X[2]);		# "X[0]"^="X[-3]"^"X[-8]"
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	  &movdqa	(&QWP(0+16*(($Xi-1)&3),"esp"),@X[3]);	# X[]+K xfer to IALU
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&movdqa	(@X[4],@X[0]);
+	&movdqa	(@X[2],@X[0]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&pslldq	(@X[4],12);		# "X[0]"<<96, extract one dword
+	&paddd	(@X[0],@X[0]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&psrld	(@X[2],31);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&movdqa	(@X[3],@X[4]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&psrld	(@X[4],30);
+	&por	(@X[0],@X[2]);		# "X[0]"<<<=1
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	  &movdqa	(@X[2],&QWP(64+16*(($Xi-6)%3),"esp")) if ($Xi>5);	# restore X[] from backtrace buffer
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&pslld	(@X[3],2);
+	&pxor	(@X[0],@X[4]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	  &movdqa	(@X[4],&QWP(112-16+16*(($Xi)/5),"esp"));	# K_XX_XX
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&pxor	(@X[0],@X[3]);		# "X[0]"^=("X[0]"<<96)<<<2
+	  &movdqa	(@X[1],@X[-2&7])	if ($Xi<7);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	 foreach (@insns) { eval; }	# remaining instructions [if any]
+
+  $Xi++;	push(@X,shift(@X));	# "rotate" X[]
+}
+
+sub Xupdate_ssse3_32_79()
+{ use integer;
+  my $body = shift;
+  my @insns = (&$body,&$body,&$body,&$body);	# 32 to 48 instructions
+  my ($a,$b,$c,$d,$e);
+
+	&movdqa	(@X[2],@X[-1&7])	if ($Xi==8);
+	 eval(shift(@insns));		# body_20_39
+	&pxor	(@X[0],@X[-4&7]);	# "X[0]"="X[-32]"^"X[-16]"
+	&palignr(@X[2],@X[-2&7],8);	# compose "X[-6]"
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# rol
+
+	&pxor	(@X[0],@X[-7&7]);	# "X[0]"^="X[-28]"
+	  &movdqa	(&QWP(64+16*(($Xi-4)%3),"esp"),@X[-4&7]);	# save X[] to backtrace buffer
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 if ($Xi%5) {
+	  &movdqa	(@X[4],@X[3]);	# "perpetuate" K_XX_XX...
+	 } else {			# ... or load next one
+	  &movdqa	(@X[4],&QWP(112-16+16*($Xi/5),"esp"));
+	 }
+	  &paddd	(@X[3],@X[-1&7]);
+	 eval(shift(@insns));		# ror
+	 eval(shift(@insns));
+
+	&pxor	(@X[0],@X[2]);		# "X[0]"^="X[-6]"
+	 eval(shift(@insns));		# body_20_39
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# rol
+
+	&movdqa	(@X[2],@X[0]);
+	  &movdqa	(&QWP(0+16*(($Xi-1)&3),"esp"),@X[3]);	# X[]+K xfer to IALU
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# ror
+	 eval(shift(@insns));
+
+	&pslld	(@X[0],2);
+	 eval(shift(@insns));		# body_20_39
+	 eval(shift(@insns));
+	&psrld	(@X[2],30);
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# rol
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# ror
+	 eval(shift(@insns));
+
+	&por	(@X[0],@X[2]);		# "X[0]"<<<=2
+	 eval(shift(@insns));		# body_20_39
+	 eval(shift(@insns));
+	  &movdqa	(@X[2],&QWP(64+16*(($Xi-6)%3),"esp")) if($Xi<19);	# restore X[] from backtrace buffer
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# rol
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# ror
+	  &movdqa	(@X[3],@X[0])	if ($Xi<19);
+	 eval(shift(@insns));
+
+	 foreach (@insns) { eval; }	# remaining instructions
+
+  $Xi++;	push(@X,shift(@X));	# "rotate" X[]
+}
+
+sub Xuplast_ssse3_80()
+{ use integer;
+  my $body = shift;
+  my @insns = (&$body,&$body,&$body,&$body);	# 32 instructions
+  my ($a,$b,$c,$d,$e);
+
+	 eval(shift(@insns));
+	  &paddd	(@X[3],@X[-1&7]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	  &movdqa	(&QWP(0+16*(($Xi-1)&3),"esp"),@X[3]);	# X[]+K xfer IALU
+
+	 foreach (@insns) { eval; }		# remaining instructions
+
+	&mov	($inp=@T[1],&DWP(192+4,"esp"));
+	&cmp	($inp,&DWP(192+8,"esp"));
+	&je	(&label("done"));
+
+	&movdqa	(@X[3],&QWP(112+48,"esp"));	# K_00_19
+	&movdqa	(@X[2],&QWP(112+64,"esp"));	# pbswap mask
+	&movdqu	(@X[-4&7],&QWP(0,$inp));	# load input
+	&movdqu	(@X[-3&7],&QWP(16,$inp));
+	&movdqu	(@X[-2&7],&QWP(32,$inp));
+	&movdqu	(@X[-1&7],&QWP(48,$inp));
+	&add	($inp,64);
+	&pshufb	(@X[-4&7],@X[2]);		# byte swap
+	&mov	(&DWP(192+4,"esp"),$inp);
+	&movdqa	(&QWP(112-16,"esp"),@X[3]);	# borrow last backtrace slot
+
+  $Xi=0;
+}
+
+sub Xloop_ssse3()
+{ use integer;
+  my $body = shift;
+  my @insns = (&$body,&$body,&$body,&$body);	# 32 instructions
+  my ($a,$b,$c,$d,$e);
+
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&pshufb	(@X[($Xi-3)&7],@X[2]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&paddd	(@X[($Xi-4)&7],@X[3]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&movdqa	(&QWP(0+16*$Xi,"esp"),@X[($Xi-4)&7]);	# X[]+K xfer to IALU
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&psubd	(@X[($Xi-4)&7],@X[3]);
+
+	foreach (@insns) { eval; }
+  $Xi++;
+}
+
+sub Xtail_ssse3()
+{ use integer;
+  my $body = shift;
+  my @insns = (&$body,&$body,&$body,&$body);	# 32 instructions
+  my ($a,$b,$c,$d,$e);
+
+	foreach (@insns) { eval; }
+}
+
+sub body_00_19 () {
+	(
+	'($a,$b,$c,$d,$e)=@V;'.
+	'&add	($e,&DWP(4*($j&15),"esp"));',	# X[]+K xfer
+	'&xor	($c,$d);',
+	'&mov	(@T[1],$a);',	# $b in next round
+	'&$_rol	($a,5);',
+	'&and	(@T[0],$c);',	# ($b&($c^$d))
+	'&xor	($c,$d);',	# restore $c
+	'&xor	(@T[0],$d);',
+	'&add	($e,$a);',
+	'&$_ror	($b,$j?7:2);',	# $b>>>2
+	'&add	($e,@T[0]);'	.'$j++; unshift(@V,pop(@V)); unshift(@T,pop(@T));'
+	);
+}
+
+sub body_20_39 () {
+	(
+	'($a,$b,$c,$d,$e)=@V;'.
+	'&add	($e,&DWP(4*($j++&15),"esp"));',	# X[]+K xfer
+	'&xor	(@T[0],$d);',	# ($b^$d)
+	'&mov	(@T[1],$a);',	# $b in next round
+	'&$_rol	($a,5);',
+	'&xor	(@T[0],$c);',	# ($b^$d^$c)
+	'&add	($e,$a);',
+	'&$_ror	($b,7);',	# $b>>>2
+	'&add	($e,@T[0]);'	.'unshift(@V,pop(@V)); unshift(@T,pop(@T));'
+	);
+}
+
+sub body_40_59 () {
+	(
+	'($a,$b,$c,$d,$e)=@V;'.
+	'&mov	(@T[1],$c);',
+	'&xor	($c,$d);',
+	'&add	($e,&DWP(4*($j++&15),"esp"));',	# X[]+K xfer
+	'&and	(@T[1],$d);',
+	'&and	(@T[0],$c);',	# ($b&($c^$d))
+	'&$_ror	($b,7);',	# $b>>>2
+	'&add	($e,@T[1]);',
+	'&mov	(@T[1],$a);',	# $b in next round
+	'&$_rol	($a,5);',
+	'&add	($e,@T[0]);',
+	'&xor	($c,$d);',	# restore $c
+	'&add	($e,$a);'	.'unshift(@V,pop(@V)); unshift(@T,pop(@T));'
+	);
+}
+
+&set_label("loop",16);
+	&Xupdate_ssse3_16_31(\&body_00_19);
+	&Xupdate_ssse3_16_31(\&body_00_19);
+	&Xupdate_ssse3_16_31(\&body_00_19);
+	&Xupdate_ssse3_16_31(\&body_00_19);
+	&Xupdate_ssse3_32_79(\&body_00_19);
+	&Xupdate_ssse3_32_79(\&body_20_39);
+	&Xupdate_ssse3_32_79(\&body_20_39);
+	&Xupdate_ssse3_32_79(\&body_20_39);
+	&Xupdate_ssse3_32_79(\&body_20_39);
+	&Xupdate_ssse3_32_79(\&body_20_39);
+	&Xupdate_ssse3_32_79(\&body_40_59);
+	&Xupdate_ssse3_32_79(\&body_40_59);
+	&Xupdate_ssse3_32_79(\&body_40_59);
+	&Xupdate_ssse3_32_79(\&body_40_59);
+	&Xupdate_ssse3_32_79(\&body_40_59);
+	&Xupdate_ssse3_32_79(\&body_20_39);
+	&Xuplast_ssse3_80(\&body_20_39);	# can jump to "done"
+
+				$saved_j=$j; @saved_V=@V;
+
+	&Xloop_ssse3(\&body_20_39);
+	&Xloop_ssse3(\&body_20_39);
+	&Xloop_ssse3(\&body_20_39);
+
+	&mov	(@T[1],&DWP(192,"esp"));	# update context
+	&add	($A,&DWP(0,@T[1]));
+	&add	(@T[0],&DWP(4,@T[1]));		# $b
+	&add	($C,&DWP(8,@T[1]));
+	&mov	(&DWP(0,@T[1]),$A);
+	&add	($D,&DWP(12,@T[1]));
+	&mov	(&DWP(4,@T[1]),@T[0]);
+	&add	($E,&DWP(16,@T[1]));
+	&mov	(&DWP(8,@T[1]),$C);
+	&mov	($B,@T[0]);
+	&mov	(&DWP(12,@T[1]),$D);
+	&mov	(&DWP(16,@T[1]),$E);
+	&movdqa	(@X[0],@X[-3&7]);
+
+	&jmp	(&label("loop"));
+
+&set_label("done",16);		$j=$saved_j; @V=@saved_V;
+
+	&Xtail_ssse3(\&body_20_39);
+	&Xtail_ssse3(\&body_20_39);
+	&Xtail_ssse3(\&body_20_39);
+
+	&mov	(@T[1],&DWP(192,"esp"));	# update context
+	&add	($A,&DWP(0,@T[1]));
+	&mov	("esp",&DWP(192+12,"esp"));	# restore %esp
+	&add	(@T[0],&DWP(4,@T[1]));		# $b
+	&add	($C,&DWP(8,@T[1]));
+	&mov	(&DWP(0,@T[1]),$A);
+	&add	($D,&DWP(12,@T[1]));
+	&mov	(&DWP(4,@T[1]),@T[0]);
+	&add	($E,&DWP(16,@T[1]));
+	&mov	(&DWP(8,@T[1]),$C);
+	&mov	(&DWP(12,@T[1]),$D);
+	&mov	(&DWP(16,@T[1]),$E);
+
+&function_end("_sha1_block_data_order_ssse3");
+
+if ($ymm) {
+my $Xi=4;			# 4xSIMD Xupdate round, start pre-seeded
+my @X=map("xmm$_",(4..7,0..3));	# pre-seeded for $Xi=4
+my @V=($A,$B,$C,$D,$E);
+my $j=0;			# hash round
+my @T=($T,$tmp1);
+my $inp;
+
+my $_rol=sub { &shld(@_[0],@_) };
+my $_ror=sub { &shrd(@_[0],@_) };
+
+&function_begin("_sha1_block_data_order_avx");
+	&call	(&label("pic_point"));	# make it PIC!
+	&set_label("pic_point");
+	&blindpop($tmp1);
+	&lea	($tmp1,&DWP(&label("K_XX_XX")."-".&label("pic_point"),$tmp1));
+&set_label("avx_shortcut");
+	&vzeroall();
+
+	&vmovdqa(@X[3],&QWP(0,$tmp1));		# K_00_19
+	&vmovdqa(@X[4],&QWP(16,$tmp1));		# K_20_39
+	&vmovdqa(@X[5],&QWP(32,$tmp1));		# K_40_59
+	&vmovdqa(@X[6],&QWP(48,$tmp1));		# K_60_79
+	&vmovdqa(@X[2],&QWP(64,$tmp1));		# pbswap mask
+
+	&mov	($E,&wparam(0));		# load argument block
+	&mov	($inp=@T[1],&wparam(1));
+	&mov	($D,&wparam(2));
+	&mov	(@T[0],"esp");
+
+	# stack frame layout
+	#
+	# +0	X[0]+K	X[1]+K	X[2]+K	X[3]+K	# XMM->IALU xfer area
+	#	X[4]+K	X[5]+K	X[6]+K	X[7]+K
+	#	X[8]+K	X[9]+K	X[10]+K	X[11]+K
+	#	X[12]+K	X[13]+K	X[14]+K	X[15]+K
+	#
+	# +64	X[0]	X[1]	X[2]	X[3]	# XMM->XMM backtrace area
+	#	X[4]	X[5]	X[6]	X[7]
+	#	X[8]	X[9]	X[10]	X[11]	# even borrowed for K_00_19
+	#
+	# +112	K_20_39	K_20_39	K_20_39	K_20_39	# constants
+	#	K_40_59	K_40_59	K_40_59	K_40_59
+	#	K_60_79	K_60_79	K_60_79	K_60_79
+	#	K_00_19	K_00_19	K_00_19	K_00_19
+	#	pbswap mask
+	#
+	# +192	ctx				# argument block
+	# +196	inp
+	# +200	end
+	# +204	esp
+	&sub	("esp",208);
+	&and	("esp",-64);
+
+	&vmovdqa(&QWP(112+0,"esp"),@X[4]);	# copy constants
+	&vmovdqa(&QWP(112+16,"esp"),@X[5]);
+	&vmovdqa(&QWP(112+32,"esp"),@X[6]);
+	&shl	($D,6);				# len*64
+	&vmovdqa(&QWP(112+48,"esp"),@X[3]);
+	&add	($D,$inp);			# end of input
+	&vmovdqa(&QWP(112+64,"esp"),@X[2]);
+	&add	($inp,64);
+	&mov	(&DWP(192+0,"esp"),$E);		# save argument block
+	&mov	(&DWP(192+4,"esp"),$inp);
+	&mov	(&DWP(192+8,"esp"),$D);
+	&mov	(&DWP(192+12,"esp"),@T[0]);	# save original %esp
+
+	&mov	($A,&DWP(0,$E));		# load context
+	&mov	($B,&DWP(4,$E));
+	&mov	($C,&DWP(8,$E));
+	&mov	($D,&DWP(12,$E));
+	&mov	($E,&DWP(16,$E));
+	&mov	(@T[0],$B);			# magic seed
+
+	&vmovdqu(@X[-4&7],&QWP(-64,$inp));	# load input to %xmm[0-3]
+	&vmovdqu(@X[-3&7],&QWP(-48,$inp));
+	&vmovdqu(@X[-2&7],&QWP(-32,$inp));
+	&vmovdqu(@X[-1&7],&QWP(-16,$inp));
+	&vpshufb(@X[-4&7],@X[-4&7],@X[2]);	# byte swap
+	&vpshufb(@X[-3&7],@X[-3&7],@X[2]);
+	&vpshufb(@X[-2&7],@X[-2&7],@X[2]);
+	&vmovdqa(&QWP(112-16,"esp"),@X[3]);	# borrow last backtrace slot
+	&vpshufb(@X[-1&7],@X[-1&7],@X[2]);
+	&vpaddd	(@X[0],@X[-4&7],@X[3]);		# add K_00_19
+	&vpaddd	(@X[1],@X[-3&7],@X[3]);
+	&vpaddd	(@X[2],@X[-2&7],@X[3]);
+	&vmovdqa(&QWP(0,"esp"),@X[0]);		# X[]+K xfer to IALU
+	&vmovdqa(&QWP(0+16,"esp"),@X[1]);
+	&vmovdqa(&QWP(0+32,"esp"),@X[2]);
+	&jmp	(&label("loop"));
+
+sub Xupdate_avx_16_31()		# recall that $Xi starts wtih 4
+{ use integer;
+  my $body = shift;
+  my @insns = (&$body,&$body,&$body,&$body);	# 40 instructions
+  my ($a,$b,$c,$d,$e);
+
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&vpalignr(@X[0],@X[-3&7],@X[-4&7],8);	# compose "X[-14]" in "X[0]"
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	  &vpaddd	(@X[3],@X[3],@X[-1&7]);
+	  &vmovdqa	(&QWP(64+16*(($Xi-4)%3),"esp"),@X[-4&7]);# save X[] to backtrace buffer
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&vpsrldq(@X[2],@X[-1&7],4);		# "X[-3]", 3 dwords
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&vpxor	(@X[0],@X[0],@X[-4&7]);		# "X[0]"^="X[-16]"
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&vpxor	(@X[2],@X[2],@X[-2&7]);		# "X[-3]"^"X[-8]"
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	  &vmovdqa	(&QWP(0+16*(($Xi-1)&3),"esp"),@X[3]);	# X[]+K xfer to IALU
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&vpxor	(@X[0],@X[0],@X[2]);		# "X[0]"^="X[-3]"^"X[-8]"
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&vpsrld	(@X[2],@X[0],31);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&vpslldq(@X[4],@X[0],12);		# "X[0]"<<96, extract one dword
+	&vpaddd	(@X[0],@X[0],@X[0]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&vpsrld	(@X[3],@X[4],30);
+	&vpor	(@X[0],@X[0],@X[2]);		# "X[0]"<<<=1
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&vpslld	(@X[4],@X[4],2);
+	  &vmovdqa	(@X[2],&QWP(64+16*(($Xi-6)%3),"esp")) if ($Xi>5);	# restore X[] from backtrace buffer
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&vpxor	(@X[0],@X[0],@X[3]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&vpxor	(@X[0],@X[0],@X[4]);		# "X[0]"^=("X[0]"<<96)<<<2
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	  &vmovdqa	(@X[4],&QWP(112-16+16*(($Xi)/5),"esp"));	# K_XX_XX
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	 foreach (@insns) { eval; }	# remaining instructions [if any]
+
+  $Xi++;	push(@X,shift(@X));	# "rotate" X[]
+}
+
+sub Xupdate_avx_32_79()
+{ use integer;
+  my $body = shift;
+  my @insns = (&$body,&$body,&$body,&$body);	# 32 to 48 instructions
+  my ($a,$b,$c,$d,$e);
+
+	&vpalignr(@X[2],@X[-1&7],@X[-2&7],8);	# compose "X[-6]"
+	&vpxor	(@X[0],@X[0],@X[-4&7]);	# "X[0]"="X[-32]"^"X[-16]"
+	 eval(shift(@insns));		# body_20_39
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# rol
+
+	&vpxor	(@X[0],@X[0],@X[-7&7]);	# "X[0]"^="X[-28]"
+	  &vmovdqa	(&QWP(64+16*(($Xi-4)%3),"esp"),@X[-4&7]);	# save X[] to backtrace buffer
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 if ($Xi%5) {
+	  &vmovdqa	(@X[4],@X[3]);	# "perpetuate" K_XX_XX...
+	 } else {			# ... or load next one
+	  &vmovdqa	(@X[4],&QWP(112-16+16*($Xi/5),"esp"));
+	 }
+	  &vpaddd	(@X[3],@X[3],@X[-1&7]);
+	 eval(shift(@insns));		# ror
+	 eval(shift(@insns));
+
+	&vpxor	(@X[0],@X[0],@X[2]);		# "X[0]"^="X[-6]"
+	 eval(shift(@insns));		# body_20_39
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# rol
+
+	&vpsrld	(@X[2],@X[0],30);
+	  &vmovdqa	(&QWP(0+16*(($Xi-1)&3),"esp"),@X[3]);	# X[]+K xfer to IALU
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# ror
+	 eval(shift(@insns));
+
+	&vpslld	(@X[0],@X[0],2);
+	 eval(shift(@insns));		# body_20_39
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# rol
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# ror
+	 eval(shift(@insns));
+
+	&vpor	(@X[0],@X[0],@X[2]);	# "X[0]"<<<=2
+	 eval(shift(@insns));		# body_20_39
+	 eval(shift(@insns));
+	  &vmovdqa	(@X[2],&QWP(64+16*(($Xi-6)%3),"esp")) if($Xi<19);	# restore X[] from backtrace buffer
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# rol
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# ror
+	 eval(shift(@insns));
+
+	 foreach (@insns) { eval; }	# remaining instructions
+
+  $Xi++;	push(@X,shift(@X));	# "rotate" X[]
+}
+
+sub Xuplast_avx_80()
+{ use integer;
+  my $body = shift;
+  my @insns = (&$body,&$body,&$body,&$body);	# 32 instructions
+  my ($a,$b,$c,$d,$e);
+
+	 eval(shift(@insns));
+	  &vpaddd	(@X[3],@X[3],@X[-1&7]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	  &vmovdqa	(&QWP(0+16*(($Xi-1)&3),"esp"),@X[3]);	# X[]+K xfer IALU
+
+	 foreach (@insns) { eval; }		# remaining instructions
+
+	&mov	($inp=@T[1],&DWP(192+4,"esp"));
+	&cmp	($inp,&DWP(192+8,"esp"));
+	&je	(&label("done"));
+
+	&vmovdqa(@X[3],&QWP(112+48,"esp"));	# K_00_19
+	&vmovdqa(@X[2],&QWP(112+64,"esp"));	# pbswap mask
+	&vmovdqu(@X[-4&7],&QWP(0,$inp));	# load input
+	&vmovdqu(@X[-3&7],&QWP(16,$inp));
+	&vmovdqu(@X[-2&7],&QWP(32,$inp));
+	&vmovdqu(@X[-1&7],&QWP(48,$inp));
+	&add	($inp,64);
+	&vpshufb(@X[-4&7],@X[-4&7],@X[2]);		# byte swap
+	&mov	(&DWP(192+4,"esp"),$inp);
+	&vmovdqa(&QWP(112-16,"esp"),@X[3]);	# borrow last backtrace slot
+
+  $Xi=0;
+}
+
+sub Xloop_avx()
+{ use integer;
+  my $body = shift;
+  my @insns = (&$body,&$body,&$body,&$body);	# 32 instructions
+  my ($a,$b,$c,$d,$e);
+
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&vpshufb	(@X[($Xi-3)&7],@X[($Xi-3)&7],@X[2]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&vpaddd	(@X[$Xi&7],@X[($Xi-4)&7],@X[3]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&vmovdqa	(&QWP(0+16*$Xi,"esp"),@X[$Xi&7]);	# X[]+K xfer to IALU
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	foreach (@insns) { eval; }
+  $Xi++;
+}
+
+sub Xtail_avx()
+{ use integer;
+  my $body = shift;
+  my @insns = (&$body,&$body,&$body,&$body);	# 32 instructions
+  my ($a,$b,$c,$d,$e);
+
+	foreach (@insns) { eval; }
+}
+
+&set_label("loop",16);
+	&Xupdate_avx_16_31(\&body_00_19);
+	&Xupdate_avx_16_31(\&body_00_19);
+	&Xupdate_avx_16_31(\&body_00_19);
+	&Xupdate_avx_16_31(\&body_00_19);
+	&Xupdate_avx_32_79(\&body_00_19);
+	&Xupdate_avx_32_79(\&body_20_39);
+	&Xupdate_avx_32_79(\&body_20_39);
+	&Xupdate_avx_32_79(\&body_20_39);
+	&Xupdate_avx_32_79(\&body_20_39);
+	&Xupdate_avx_32_79(\&body_20_39);
+	&Xupdate_avx_32_79(\&body_40_59);
+	&Xupdate_avx_32_79(\&body_40_59);
+	&Xupdate_avx_32_79(\&body_40_59);
+	&Xupdate_avx_32_79(\&body_40_59);
+	&Xupdate_avx_32_79(\&body_40_59);
+	&Xupdate_avx_32_79(\&body_20_39);
+	&Xuplast_avx_80(\&body_20_39);	# can jump to "done"
+
+				$saved_j=$j; @saved_V=@V;
+
+	&Xloop_avx(\&body_20_39);
+	&Xloop_avx(\&body_20_39);
+	&Xloop_avx(\&body_20_39);
+
+	&mov	(@T[1],&DWP(192,"esp"));	# update context
+	&add	($A,&DWP(0,@T[1]));
+	&add	(@T[0],&DWP(4,@T[1]));		# $b
+	&add	($C,&DWP(8,@T[1]));
+	&mov	(&DWP(0,@T[1]),$A);
+	&add	($D,&DWP(12,@T[1]));
+	&mov	(&DWP(4,@T[1]),@T[0]);
+	&add	($E,&DWP(16,@T[1]));
+	&mov	(&DWP(8,@T[1]),$C);
+	&mov	($B,@T[0]);
+	&mov	(&DWP(12,@T[1]),$D);
+	&mov	(&DWP(16,@T[1]),$E);
+
+	&jmp	(&label("loop"));
+
+&set_label("done",16);		$j=$saved_j; @V=@saved_V;
+
+	&Xtail_avx(\&body_20_39);
+	&Xtail_avx(\&body_20_39);
+	&Xtail_avx(\&body_20_39);
+
+	&vzeroall();
+
+	&mov	(@T[1],&DWP(192,"esp"));	# update context
+	&add	($A,&DWP(0,@T[1]));
+	&mov	("esp",&DWP(192+12,"esp"));	# restore %esp
+	&add	(@T[0],&DWP(4,@T[1]));		# $b
+	&add	($C,&DWP(8,@T[1]));
+	&mov	(&DWP(0,@T[1]),$A);
+	&add	($D,&DWP(12,@T[1]));
+	&mov	(&DWP(4,@T[1]),@T[0]);
+	&add	($E,&DWP(16,@T[1]));
+	&mov	(&DWP(8,@T[1]),$C);
+	&mov	(&DWP(12,@T[1]),$D);
+	&mov	(&DWP(16,@T[1]),$E);
+&function_end("_sha1_block_data_order_avx");
+}
+&set_label("K_XX_XX",64);
+&data_word(0x5a827999,0x5a827999,0x5a827999,0x5a827999);	# K_00_19
+&data_word(0x6ed9eba1,0x6ed9eba1,0x6ed9eba1,0x6ed9eba1);	# K_20_39
+&data_word(0x8f1bbcdc,0x8f1bbcdc,0x8f1bbcdc,0x8f1bbcdc);	# K_40_59
+&data_word(0xca62c1d6,0xca62c1d6,0xca62c1d6,0xca62c1d6);	# K_60_79
+&data_word(0x00010203,0x04050607,0x08090a0b,0x0c0d0e0f);	# pbswap mask
+}
 &asciz("SHA1 block transform for x86, CRYPTOGAMS by <appro\@openssl.org>");
 
 &asm_finish();
diff --git a/openssl/crypto/sha/asm/sha1-alpha.pl b/openssl/crypto/sha/asm/sha1-alpha.pl
new file mode 100644
index 000000000..6c4b9251f
--- /dev/null
+++ b/openssl/crypto/sha/asm/sha1-alpha.pl
@@ -0,0 +1,322 @@
+#!/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/.
+# ====================================================================
+
+# SHA1 block procedure for Alpha.
+
+# On 21264 performance is 33% better than code generated by vendor
+# compiler, and 75% better than GCC [3.4], and in absolute terms is
+# 8.7 cycles per processed byte. Implementation features vectorized
+# byte swap, but not Xupdate.
+
+@X=(	"\$0",	"\$1",	"\$2",	"\$3",	"\$4",	"\$5",	"\$6",	"\$7",
+	"\$8",	"\$9",	"\$10",	"\$11",	"\$12",	"\$13",	"\$14",	"\$15");
+$ctx="a0";	# $16
+$inp="a1";
+$num="a2";
+$A="a3";
+$B="a4";	# 20
+$C="a5";
+$D="t8";
+$E="t9";	@V=($A,$B,$C,$D,$E);
+$t0="t10";	# 24
+$t1="t11";
+$t2="ra";
+$t3="t12";
+$K="AT";	# 28
+
+sub BODY_00_19 {
+my ($i,$a,$b,$c,$d,$e)=@_;
+my $j=$i+1;
+$code.=<<___ if ($i==0);
+	ldq_u	@X[0],0+0($inp)
+	ldq_u	@X[1],0+7($inp)
+___
+$code.=<<___ if (!($i&1) && $i<14);
+	ldq_u	@X[$i+2],($i+2)*4+0($inp)
+	ldq_u	@X[$i+3],($i+2)*4+7($inp)
+___
+$code.=<<___ if (!($i&1) && $i<15);
+	extql	@X[$i],$inp,@X[$i]
+	extqh	@X[$i+1],$inp,@X[$i+1]
+
+	or	@X[$i+1],@X[$i],@X[$i]	# pair of 32-bit values are fetched
+
+	srl	@X[$i],24,$t0		# vectorized byte swap
+	srl	@X[$i],8,$t2
+
+	sll	@X[$i],8,$t3
+	sll	@X[$i],24,@X[$i]
+	zapnot	$t0,0x11,$t0
+	zapnot	$t2,0x22,$t2
+
+	zapnot	@X[$i],0x88,@X[$i]
+	or	$t0,$t2,$t0
+	zapnot	$t3,0x44,$t3
+	sll	$a,5,$t1
+
+	or	@X[$i],$t0,@X[$i]
+	addl	$K,$e,$e
+	and	$b,$c,$t2
+	zapnot	$a,0xf,$a
+
+	or	@X[$i],$t3,@X[$i]
+	srl	$a,27,$t0
+	bic	$d,$b,$t3
+	sll	$b,30,$b
+
+	extll	@X[$i],4,@X[$i+1]	# extract upper half
+	or	$t2,$t3,$t2
+	addl	@X[$i],$e,$e
+
+	addl	$t1,$e,$e
+	srl	$b,32,$t3
+	zapnot	@X[$i],0xf,@X[$i]
+
+	addl	$t0,$e,$e
+	addl	$t2,$e,$e
+	or	$t3,$b,$b
+___
+$code.=<<___ if (($i&1) && $i<15);
+	sll	$a,5,$t1
+	addl	$K,$e,$e
+	and	$b,$c,$t2
+	zapnot	$a,0xf,$a
+
+	srl	$a,27,$t0
+	addl	@X[$i%16],$e,$e
+	bic	$d,$b,$t3
+	sll	$b,30,$b
+
+	or	$t2,$t3,$t2
+	addl	$t1,$e,$e
+	srl	$b,32,$t3
+	zapnot	@X[$i],0xf,@X[$i]
+
+	addl	$t0,$e,$e
+	addl	$t2,$e,$e
+	or	$t3,$b,$b
+___
+$code.=<<___ if ($i>=15);	# with forward Xupdate
+	sll	$a,5,$t1
+	addl	$K,$e,$e
+	and	$b,$c,$t2
+	xor	@X[($j+2)%16],@X[$j%16],@X[$j%16]
+
+	zapnot	$a,0xf,$a
+	addl	@X[$i%16],$e,$e
+	bic	$d,$b,$t3
+	xor	@X[($j+8)%16],@X[$j%16],@X[$j%16]
+
+	srl	$a,27,$t0
+	addl	$t1,$e,$e
+	or	$t2,$t3,$t2
+	xor	@X[($j+13)%16],@X[$j%16],@X[$j%16]
+
+	sll	$b,30,$b
+	addl	$t0,$e,$e
+	srl	@X[$j%16],31,$t1
+
+	addl	$t2,$e,$e
+	srl	$b,32,$t3
+	addl	@X[$j%16],@X[$j%16],@X[$j%16]
+
+	or	$t3,$b,$b
+	zapnot	@X[$i%16],0xf,@X[$i%16]
+	or	$t1,@X[$j%16],@X[$j%16]
+___
+}
+
+sub BODY_20_39 {
+my ($i,$a,$b,$c,$d,$e)=@_;
+my $j=$i+1;
+$code.=<<___ if ($i<79);	# with forward Xupdate
+	sll	$a,5,$t1
+	addl	$K,$e,$e
+	zapnot	$a,0xf,$a
+	xor	@X[($j+2)%16],@X[$j%16],@X[$j%16]
+
+	sll	$b,30,$t3
+	addl	$t1,$e,$e
+	xor	$b,$c,$t2
+	xor	@X[($j+8)%16],@X[$j%16],@X[$j%16]
+
+	srl	$b,2,$b
+	addl	@X[$i%16],$e,$e
+	xor	$d,$t2,$t2
+	xor	@X[($j+13)%16],@X[$j%16],@X[$j%16]
+
+	srl	@X[$j%16],31,$t1
+	addl	$t2,$e,$e
+	srl	$a,27,$t0
+	addl	@X[$j%16],@X[$j%16],@X[$j%16]
+
+	or	$t3,$b,$b
+	addl	$t0,$e,$e
+	or	$t1,@X[$j%16],@X[$j%16]
+___
+$code.=<<___ if ($i<77);
+	zapnot	@X[$i%16],0xf,@X[$i%16]
+___
+$code.=<<___ if ($i==79);	# with context fetch
+	sll	$a,5,$t1
+	addl	$K,$e,$e
+	zapnot	$a,0xf,$a
+	ldl	@X[0],0($ctx)
+
+	sll	$b,30,$t3
+	addl	$t1,$e,$e
+	xor	$b,$c,$t2
+	ldl	@X[1],4($ctx)
+
+	srl	$b,2,$b
+	addl	@X[$i%16],$e,$e
+	xor	$d,$t2,$t2
+	ldl	@X[2],8($ctx)
+
+	srl	$a,27,$t0
+	addl	$t2,$e,$e
+	ldl	@X[3],12($ctx)
+
+	or	$t3,$b,$b
+	addl	$t0,$e,$e
+	ldl	@X[4],16($ctx)
+___
+}
+
+sub BODY_40_59 {
+my ($i,$a,$b,$c,$d,$e)=@_;
+my $j=$i+1;
+$code.=<<___;	# with forward Xupdate
+	sll	$a,5,$t1
+	addl	$K,$e,$e
+	zapnot	$a,0xf,$a
+	xor	@X[($j+2)%16],@X[$j%16],@X[$j%16]
+
+	srl	$a,27,$t0
+	and	$b,$c,$t2
+	and	$b,$d,$t3
+	xor	@X[($j+8)%16],@X[$j%16],@X[$j%16]
+
+	sll	$b,30,$b
+	addl	$t1,$e,$e
+	xor	@X[($j+13)%16],@X[$j%16],@X[$j%16]
+
+	srl	@X[$j%16],31,$t1
+	addl	$t0,$e,$e
+	or	$t2,$t3,$t2
+	and	$c,$d,$t3
+
+	or	$t2,$t3,$t2
+	srl	$b,32,$t3
+	addl	@X[$i%16],$e,$e
+	addl	@X[$j%16],@X[$j%16],@X[$j%16]
+
+	or	$t3,$b,$b
+	addl	$t2,$e,$e
+	or	$t1,@X[$j%16],@X[$j%16]
+	zapnot	@X[$i%16],0xf,@X[$i%16]
+___
+}
+
+$code=<<___;
+#ifdef __linux__
+#include <asm/regdef.h>
+#else
+#include <asm.h>
+#include <regdef.h>
+#endif
+
+.text
+
+.set	noat
+.set	noreorder
+.globl	sha1_block_data_order
+.align	5
+.ent	sha1_block_data_order
+sha1_block_data_order:
+	lda	sp,-64(sp)
+	stq	ra,0(sp)
+	stq	s0,8(sp)
+	stq	s1,16(sp)
+	stq	s2,24(sp)
+	stq	s3,32(sp)
+	stq	s4,40(sp)
+	stq	s5,48(sp)
+	stq	fp,56(sp)
+	.mask	0x0400fe00,-64
+	.frame	sp,64,ra
+	.prologue 0
+
+	ldl	$A,0($ctx)
+	ldl	$B,4($ctx)
+	sll	$num,6,$num
+	ldl	$C,8($ctx)
+	ldl	$D,12($ctx)
+	ldl	$E,16($ctx)
+	addq	$inp,$num,$num
+
+.Lloop:
+	.set	noreorder
+	ldah	$K,23170(zero)
+	zapnot	$B,0xf,$B
+	lda	$K,31129($K)	# K_00_19
+___
+for ($i=0;$i<20;$i++) { &BODY_00_19($i,@V); unshift(@V,pop(@V)); }
+
+$code.=<<___;
+	ldah	$K,28378(zero)
+	lda	$K,-5215($K)	# K_20_39
+___
+for (;$i<40;$i++) { &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
+
+$code.=<<___;
+	ldah	$K,-28900(zero)
+	lda	$K,-17188($K)	# K_40_59
+___
+for (;$i<60;$i++) { &BODY_40_59($i,@V); unshift(@V,pop(@V)); }
+
+$code.=<<___;
+	ldah	$K,-13725(zero)
+	lda	$K,-15914($K)	# K_60_79
+___
+for (;$i<80;$i++) { &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
+
+$code.=<<___;
+	addl	@X[0],$A,$A
+	addl	@X[1],$B,$B
+	addl	@X[2],$C,$C
+	addl	@X[3],$D,$D
+	addl	@X[4],$E,$E
+	stl	$A,0($ctx)
+	stl	$B,4($ctx)
+	addq	$inp,64,$inp
+	stl	$C,8($ctx)
+	stl	$D,12($ctx)
+	stl	$E,16($ctx)
+	cmpult	$inp,$num,$t1
+	bne	$t1,.Lloop
+
+	.set	noreorder
+	ldq	ra,0(sp)
+	ldq	s0,8(sp)
+	ldq	s1,16(sp)
+	ldq	s2,24(sp)
+	ldq	s3,32(sp)
+	ldq	s4,40(sp)
+	ldq	s5,48(sp)
+	ldq	fp,56(sp)
+	lda	sp,64(sp)
+	ret	(ra)
+.end	sha1_block_data_order
+.ascii	"SHA1 block transform for Alpha, CRYPTOGAMS by <appro\@openssl.org>"
+.align	2
+___
+$output=shift and open STDOUT,">$output";
+print $code;
+close STDOUT;
diff --git a/openssl/crypto/sha/asm/sha1-armv4-large.pl b/openssl/crypto/sha/asm/sha1-armv4-large.pl
index 6e65fe3e0..fe8207f77 100644
--- a/openssl/crypto/sha/asm/sha1-armv4-large.pl
+++ b/openssl/crypto/sha/asm/sha1-armv4-large.pl
@@ -47,6 +47,10 @@
 # Cortex A8 core and in absolute terms ~870 cycles per input block
 # [or 13.6 cycles per byte].
 
+# February 2011.
+#
+# Profiler-assisted and platform-specific optimization resulted in 10%
+# improvement on Cortex A8 core and 12.2 cycles per byte.
 
 while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {}
 open STDOUT,">$output";
@@ -76,31 +80,41 @@ $code.=<<___;
 	add	$e,$K,$e,ror#2			@ E+=K_xx_xx
 	ldr	$t3,[$Xi,#2*4]
 	eor	$t0,$t0,$t1
-	eor	$t2,$t2,$t3
+	eor	$t2,$t2,$t3			@ 1 cycle stall
 	eor	$t1,$c,$d			@ F_xx_xx
 	mov	$t0,$t0,ror#31
 	add	$e,$e,$a,ror#27			@ E+=ROR(A,27)
 	eor	$t0,$t0,$t2,ror#31
+	str	$t0,[$Xi,#-4]!
 	$opt1					@ F_xx_xx
 	$opt2					@ F_xx_xx
 	add	$e,$e,$t0			@ E+=X[i]
-	str	$t0,[$Xi,#-4]!
 ___
 }
 
 sub BODY_00_15 {
 my ($a,$b,$c,$d,$e)=@_;
 $code.=<<___;
-	ldrb	$t0,[$inp],#4
-	ldrb	$t1,[$inp,#-1]
-	ldrb	$t2,[$inp,#-2]
+#if __ARM_ARCH__<7
+	ldrb	$t1,[$inp,#2]
+	ldrb	$t0,[$inp,#3]
+	ldrb	$t2,[$inp,#1]
 	add	$e,$K,$e,ror#2			@ E+=K_00_19
-	ldrb	$t3,[$inp,#-3]
+	ldrb	$t3,[$inp],#4
+	orr	$t0,$t0,$t1,lsl#8
+	eor	$t1,$c,$d			@ F_xx_xx
+	orr	$t0,$t0,$t2,lsl#16
 	add	$e,$e,$a,ror#27			@ E+=ROR(A,27)
-	orr	$t0,$t1,$t0,lsl#24
+	orr	$t0,$t0,$t3,lsl#24
+#else
+	ldr	$t0,[$inp],#4			@ handles unaligned
+	add	$e,$K,$e,ror#2			@ E+=K_00_19
 	eor	$t1,$c,$d			@ F_xx_xx
-	orr	$t0,$t0,$t2,lsl#8
-	orr	$t0,$t0,$t3,lsl#16
+	add	$e,$e,$a,ror#27			@ E+=ROR(A,27)
+#ifdef __ARMEL__
+	rev	$t0,$t0				@ byte swap
+#endif
+#endif
 	and	$t1,$b,$t1,ror#2
 	add	$e,$e,$t0			@ E+=X[i]
 	eor	$t1,$t1,$d,ror#2		@ F_00_19(B,C,D)
@@ -136,6 +150,8 @@ ___
 }
 
 $code=<<___;
+#include "arm_arch.h"
+
 .text
 
 .global	sha1_block_data_order
@@ -209,10 +225,14 @@ $code.=<<___;
 	teq	$inp,$len
 	bne	.Lloop			@ [+18], total 1307
 
+#if __ARM_ARCH__>=5
+	ldmia	sp!,{r4-r12,pc}
+#else
 	ldmia	sp!,{r4-r12,lr}
 	tst	lr,#1
 	moveq	pc,lr			@ be binary compatible with V4, yet
 	bx	lr			@ interoperable with Thumb ISA:-)
+#endif
 .align	2
 .LK_00_19:	.word	0x5a827999
 .LK_20_39:	.word	0x6ed9eba1
diff --git a/openssl/crypto/sha/asm/sha1-ia64.pl b/openssl/crypto/sha/asm/sha1-ia64.pl
index 51c4f47ec..db28f0805 100644
--- a/openssl/crypto/sha/asm/sha1-ia64.pl
+++ b/openssl/crypto/sha/asm/sha1-ia64.pl
@@ -15,7 +15,7 @@
 # is >50% better than HP C and >2x better than gcc.
 
 $code=<<___;
-.ident  \"sha1-ia64.s, version 1.2\"
+.ident  \"sha1-ia64.s, version 1.3\"
 .ident  \"IA-64 ISA artwork by Andy Polyakov <appro\@fy.chalmers.se>\"
 .explicit
 
@@ -26,14 +26,10 @@ if ($^O eq "hpux") {
     $ADDP="addp4";
     for (@ARGV) { $ADDP="add" if (/[\+DD|\-mlp]64/); }
 } else { $ADDP="add"; }
-for (@ARGV) {	$big_endian=1 if (/\-DB_ENDIAN/);
-		$big_endian=0 if (/\-DL_ENDIAN/);   }
-if (!defined($big_endian))
-	    {	$big_endian=(unpack('L',pack('N',1))==1);   }
 
 #$human=1;
 if ($human) {	# useful for visual code auditing...
-	($A,$B,$C,$D,$E,$T)   = ("A","B","C","D","E","T");
+	($A,$B,$C,$D,$E)   = ("A","B","C","D","E");
 	($h0,$h1,$h2,$h3,$h4) = ("h0","h1","h2","h3","h4");
 	($K_00_19, $K_20_39, $K_40_59, $K_60_79) =
 	    (	"K_00_19","K_20_39","K_40_59","K_60_79"	);
@@ -41,47 +37,50 @@ if ($human) {	# useful for visual code auditing...
 		"X8", "X9","X10","X11","X12","X13","X14","X15"	);
 }
 else {
-	($A,$B,$C,$D,$E,$T)   = ("loc0","loc1","loc2","loc3","loc4","loc5");
-	($h0,$h1,$h2,$h3,$h4) = ("loc6","loc7","loc8","loc9","loc10");
+	($A,$B,$C,$D,$E)   =    ("loc0","loc1","loc2","loc3","loc4");
+	($h0,$h1,$h2,$h3,$h4) = ("loc5","loc6","loc7","loc8","loc9");
 	($K_00_19, $K_20_39, $K_40_59, $K_60_79) =
-	    (	"r14", "r15", "loc11", "loc12"	);
+	    (	"r14", "r15", "loc10", "loc11"	);
 	@X= (	"r16", "r17", "r18", "r19", "r20", "r21", "r22", "r23",
 		"r24", "r25", "r26", "r27", "r28", "r29", "r30", "r31"	);
 }
 
 sub BODY_00_15 {
 local	*code=shift;
-local	($i,$a,$b,$c,$d,$e,$f)=@_;
+my	($i,$a,$b,$c,$d,$e)=@_;
+my	$j=$i+1;
+my	$Xn=@X[$j%16];
 
 $code.=<<___ if ($i==0);
-{ .mmi;	ld1	$X[$i&0xf]=[inp],2	    // MSB
+{ .mmi;	ld1	$X[$i]=[inp],2		    // MSB
 	ld1	tmp2=[tmp3],2		};;
 { .mmi;	ld1	tmp0=[inp],2
 	ld1	tmp4=[tmp3],2		    // LSB
-	dep	$X[$i&0xf]=$X[$i&0xf],tmp2,8,8	};;
+	dep	$X[$i]=$X[$i],tmp2,8,8	};;
 ___
 if ($i<15) {
 	$code.=<<___;
-{ .mmi;	ld1	$X[($i+1)&0xf]=[inp],2	    // +1
+{ .mmi;	ld1	$Xn=[inp],2		    // forward Xload
+	nop.m	0x0
 	dep	tmp1=tmp0,tmp4,8,8	};;
-{ .mmi;	ld1	tmp2=[tmp3],2		    // +1
+{ .mmi;	ld1	tmp2=[tmp3],2		    // forward Xload
 	and	tmp4=$c,$b
-	dep	$X[$i&0xf]=$X[$i&0xf],tmp1,16,16	} //;;
-{ .mmi;	andcm	tmp1=$d,$b
-	add	tmp0=$e,$K_00_19
+	dep	$X[$i]=$X[$i],tmp1,16,16} //;;
+{ .mmi;	add	$e=$e,$K_00_19		    // e+=K_00_19
+	andcm	tmp1=$d,$b
 	dep.z	tmp5=$a,5,27		};; // a<<5
-{ .mmi;	or	tmp4=tmp4,tmp1		    // F_00_19(b,c,d)=(b&c)|(~b&d)
-	add	$f=tmp0,$X[$i&0xf]	    // f=xi+e+K_00_19
+{ .mmi;	add	$e=$e,$X[$i]		    // e+=Xload
+	or	tmp4=tmp4,tmp1		    // F_00_19(b,c,d)=(b&c)|(~b&d)
 	extr.u	tmp1=$a,27,5		};; // a>>27
-{ .mmi;	ld1	tmp0=[inp],2		    // +1
-	add	$f=$f,tmp4		    // f+=F_00_19(b,c,d)
+{ .mmi;	ld1	tmp0=[inp],2		    // forward Xload
+	add	$e=$e,tmp4		    // e+=F_00_19(b,c,d)
 	shrp	$b=tmp6,tmp6,2		}   // b=ROTATE(b,30)
-{ .mmi;	ld1	tmp4=[tmp3],2		    // +1
+{ .mmi;	ld1	tmp4=[tmp3],2		    // forward Xload
 	or	tmp5=tmp1,tmp5		    // ROTATE(a,5)
 	mux2	tmp6=$a,0x44		};; // see b in next iteration
-{ .mii;	add	$f=$f,tmp5		    // f+=ROTATE(a,5)
-	dep	$X[($i+1)&0xf]=$X[($i+1)&0xf],tmp2,8,8	// +1
-	mux2	$X[$i&0xf]=$X[$i&0xf],0x44	} //;;
+{ .mii;	add	$e=$e,tmp5		    // e+=ROTATE(a,5)
+	dep	$Xn=$Xn,tmp2,8,8	    // forward Xload
+	mux2	$X[$i]=$X[$i],0x44	} //;;
 
 ___
 	}
@@ -89,24 +88,24 @@ else	{
 	$code.=<<___;
 { .mii;	and	tmp3=$c,$b
 	dep	tmp1=tmp0,tmp4,8,8;;
-	dep	$X[$i&0xf]=$X[$i&0xf],tmp1,16,16	} //;;
-{ .mmi;	andcm	tmp1=$d,$b
-	add	tmp0=$e,$K_00_19
+	dep	$X[$i]=$X[$i],tmp1,16,16} //;;
+{ .mmi;	add	$e=$e,$K_00_19		    // e+=K_00_19
+	andcm	tmp1=$d,$b
 	dep.z	tmp5=$a,5,27		};; // a<<5
-{ .mmi;	or	tmp4=tmp3,tmp1		    // F_00_19(b,c,d)=(b&c)|(~b&d)
-	add	$f=tmp0,$X[$i&0xf]	    // f=xi+e+K_00_19
+{ .mmi;	add	$e=$e,$X[$i]		    // e+=Xupdate
+	or	tmp4=tmp3,tmp1		    // F_00_19(b,c,d)=(b&c)|(~b&d)
 	extr.u	tmp1=$a,27,5		}   // a>>27
-{ .mmi;	xor	tmp2=$X[($i+0+1)&0xf],$X[($i+2+1)&0xf]	// +1
-	xor	tmp3=$X[($i+8+1)&0xf],$X[($i+13+1)&0xf] // +1
+{ .mmi;	xor	$Xn=$Xn,$X[($j+2)%16]	    // forward Xupdate
+	xor	tmp3=$X[($j+8)%16],$X[($j+13)%16] // forward Xupdate
 	nop.i	0			};;
-{ .mmi;	add	$f=$f,tmp4		    // f+=F_00_19(b,c,d)
-	xor	tmp2=tmp2,tmp3		    // +1
+{ .mmi;	add	$e=$e,tmp4		    // e+=F_00_19(b,c,d)
+	xor	$Xn=$Xn,tmp3		    // forward Xupdate
 	shrp	$b=tmp6,tmp6,2		}   // b=ROTATE(b,30)
 { .mmi; or	tmp1=tmp1,tmp5		    // ROTATE(a,5)
 	mux2	tmp6=$a,0x44		};; // see b in next iteration
-{ .mii;	add	$f=$f,tmp1		    // f+=ROTATE(a,5)
-	shrp	$e=tmp2,tmp2,31		    // f+1=ROTATE(x[0]^x[2]^x[8]^x[13],1)
-	mux2	$X[$i&0xf]=$X[$i&0xf],0x44  };;
+{ .mii;	add	$e=$e,tmp1		    // e+=ROTATE(a,5)
+	shrp	$Xn=$Xn,$Xn,31		    // ROTATE(x[0]^x[2]^x[8]^x[13],1)
+	mux2	$X[$i]=$X[$i],0x44	};;
 
 ___
 	}
@@ -114,27 +113,28 @@ ___
 
 sub BODY_16_19 {
 local	*code=shift;
-local	($i,$a,$b,$c,$d,$e,$f)=@_;
+my	($i,$a,$b,$c,$d,$e)=@_;
+my	$j=$i+1;
+my	$Xn=@X[$j%16];
 
 $code.=<<___;
-{ .mmi;	mov	$X[$i&0xf]=$f		    // Xupdate
-	and	tmp0=$c,$b
+{ .mib;	add	$e=$e,$K_00_19		    // e+=K_00_19
 	dep.z	tmp5=$a,5,27		}   // a<<5
-{ .mmi;	andcm	tmp1=$d,$b
-	add	tmp4=$e,$K_00_19	};;
-{ .mmi;	or	tmp0=tmp0,tmp1		    // F_00_19(b,c,d)=(b&c)|(~b&d)
-	add	$f=$f,tmp4		    // f+=e+K_00_19
+{ .mib;	andcm	tmp1=$d,$b
+	and	tmp0=$c,$b		};;
+{ .mmi;	add	$e=$e,$X[$i%16]		    // e+=Xupdate
+	or	tmp0=tmp0,tmp1		    // F_00_19(b,c,d)=(b&c)|(~b&d)
 	extr.u	tmp1=$a,27,5		}   // a>>27
-{ .mmi;	xor	tmp2=$X[($i+0+1)&0xf],$X[($i+2+1)&0xf]	// +1
-	xor	tmp3=$X[($i+8+1)&0xf],$X[($i+13+1)&0xf]	// +1
+{ .mmi;	xor	$Xn=$Xn,$X[($j+2)%16]	    // forward Xupdate
+	xor	tmp3=$X[($j+8)%16],$X[($j+13)%16]	// forward Xupdate
 	nop.i	0			};;
-{ .mmi;	add	$f=$f,tmp0		    // f+=F_00_19(b,c,d)
-	xor	tmp2=tmp2,tmp3		    // +1
+{ .mmi;	add	$e=$e,tmp0		    // f+=F_00_19(b,c,d)
+	xor	$Xn=$Xn,tmp3		    // forward Xupdate
 	shrp	$b=tmp6,tmp6,2		}   // b=ROTATE(b,30)
 { .mmi;	or	tmp1=tmp1,tmp5		    // ROTATE(a,5)
 	mux2	tmp6=$a,0x44		};; // see b in next iteration
-{ .mii;	add	$f=$f,tmp1		    // f+=ROTATE(a,5)
-	shrp	$e=tmp2,tmp2,31		    // f+1=ROTATE(x[0]^x[2]^x[8]^x[13],1)
+{ .mii;	add	$e=$e,tmp1		    // e+=ROTATE(a,5)
+	shrp	$Xn=$Xn,$Xn,31		    // ROTATE(x[0]^x[2]^x[8]^x[13],1)
 	nop.i	0			};;
 
 ___
@@ -142,49 +142,47 @@ ___
 
 sub BODY_20_39 {
 local	*code=shift;
-local	($i,$a,$b,$c,$d,$e,$f,$Konst)=@_;
+my	($i,$a,$b,$c,$d,$e,$Konst)=@_;
 	$Konst = $K_20_39 if (!defined($Konst));
+my	$j=$i+1;
+my	$Xn=@X[$j%16];
 
 if ($i<79) {
 $code.=<<___;
-{ .mib;	mov	$X[$i&0xf]=$f		    // Xupdate
+{ .mib;	add	$e=$e,$Konst		    // e+=K_XX_XX
 	dep.z	tmp5=$a,5,27		}   // a<<5
 { .mib;	xor	tmp0=$c,$b
-	add	tmp4=$e,$Konst		};;
-{ .mmi;	xor	tmp0=tmp0,$d		    // F_20_39(b,c,d)=b^c^d
-	add	$f=$f,tmp4		    // f+=e+K_20_39
+	xor	$Xn=$Xn,$X[($j+2)%16]	};; // forward Xupdate
+{ .mib;	add	$e=$e,$X[$i%16]		    // e+=Xupdate
 	extr.u	tmp1=$a,27,5		}   // a>>27
-{ .mmi;	xor	tmp2=$X[($i+0+1)&0xf],$X[($i+2+1)&0xf]	// +1
-	xor	tmp3=$X[($i+8+1)&0xf],$X[($i+13+1)&0xf]	// +1
-	nop.i	0			};;
-{ .mmi;	add	$f=$f,tmp0		    // f+=F_20_39(b,c,d)
-	xor	tmp2=tmp2,tmp3		    // +1
+{ .mib;	xor	tmp0=tmp0,$d		    // F_20_39(b,c,d)=b^c^d
+	xor	$Xn=$Xn,$X[($j+8)%16]	};; // forward Xupdate
+{ .mmi;	add	$e=$e,tmp0		    // e+=F_20_39(b,c,d)
+	xor	$Xn=$Xn,$X[($j+13)%16]	    // forward Xupdate
 	shrp	$b=tmp6,tmp6,2		}   // b=ROTATE(b,30)
 { .mmi;	or	tmp1=tmp1,tmp5		    // ROTATE(a,5)
 	mux2	tmp6=$a,0x44		};; // see b in next iteration
-{ .mii;	add	$f=$f,tmp1		    // f+=ROTATE(a,5)
-	shrp	$e=tmp2,tmp2,31		    // f+1=ROTATE(x[0]^x[2]^x[8]^x[13],1)
+{ .mii;	add	$e=$e,tmp1		    // e+=ROTATE(a,5)
+	shrp	$Xn=$Xn,$Xn,31		    // ROTATE(x[0]^x[2]^x[8]^x[13],1)
 	nop.i	0			};;
 
 ___
 }
 else {
 $code.=<<___;
-{ .mib;	mov	$X[$i&0xf]=$f		    // Xupdate
+{ .mib;	add	$e=$e,$Konst		    // e+=K_60_79
 	dep.z	tmp5=$a,5,27		}   // a<<5
 { .mib;	xor	tmp0=$c,$b
-	add	tmp4=$e,$Konst		};;
-{ .mib;	xor	tmp0=tmp0,$d		    // F_20_39(b,c,d)=b^c^d
-	extr.u	tmp1=$a,27,5		}   // a>>27
-{ .mib;	add	$f=$f,tmp4		    // f+=e+K_20_39
 	add	$h1=$h1,$a		};; // wrap up
-{ .mmi;	add	$f=$f,tmp0		    // f+=F_20_39(b,c,d)
-	shrp	$b=tmp6,tmp6,2		}   // b=ROTATE(b,30) ;;?
-{ .mmi;	or	tmp1=tmp1,tmp5		    // ROTATE(a,5)
+{ .mib;	add	$e=$e,$X[$i%16]		    // e+=Xupdate
+	extr.u	tmp1=$a,27,5		}   // a>>27
+{ .mib;	xor	tmp0=tmp0,$d		    // F_20_39(b,c,d)=b^c^d
 	add	$h3=$h3,$c		};; // wrap up
-{ .mib;	add	tmp3=1,inp		    // used in unaligned codepath
-	add	$f=$f,tmp1		}   // f+=ROTATE(a,5)
-{ .mib;	add	$h2=$h2,$b		    // wrap up
+{ .mmi;	add	$e=$e,tmp0		    // e+=F_20_39(b,c,d)
+	or	tmp1=tmp1,tmp5		    // ROTATE(a,5)
+	shrp	$b=tmp6,tmp6,2		};; // b=ROTATE(b,30) ;;?
+{ .mmi;	add	$e=$e,tmp1		    // e+=ROTATE(a,5)
+	add	tmp3=1,inp		    // used in unaligned codepath
 	add	$h4=$h4,$d		};; // wrap up
 
 ___
@@ -193,29 +191,29 @@ ___
 
 sub BODY_40_59 {
 local	*code=shift;
-local	($i,$a,$b,$c,$d,$e,$f)=@_;
+my	($i,$a,$b,$c,$d,$e)=@_;
+my	$j=$i+1;
+my	$Xn=@X[$j%16];
 
 $code.=<<___;
-{ .mmi;	mov	$X[$i&0xf]=$f		    // Xupdate
-	and	tmp0=$c,$b
+{ .mib;	add	$e=$e,$K_40_59		    // e+=K_40_59
 	dep.z	tmp5=$a,5,27		}   // a<<5
-{ .mmi;	and	tmp1=$d,$b
-	add	tmp4=$e,$K_40_59	};;
-{ .mmi;	or	tmp0=tmp0,tmp1		    // (b&c)|(b&d)
-	add	$f=$f,tmp4		    // f+=e+K_40_59
+{ .mib;	and	tmp1=$c,$d
+	xor	tmp0=$c,$d		};;
+{ .mmi;	add	$e=$e,$X[$i%16]		    // e+=Xupdate
+	add	tmp5=tmp5,tmp1		    // a<<5+(c&d)
 	extr.u	tmp1=$a,27,5		}   // a>>27
-{ .mmi;	and	tmp4=$c,$d
-	xor	tmp2=$X[($i+0+1)&0xf],$X[($i+2+1)&0xf]	// +1
-	xor	tmp3=$X[($i+8+1)&0xf],$X[($i+13+1)&0xf]	// +1
-	};;
-{ .mmi;	or	tmp1=tmp1,tmp5		    // ROTATE(a,5)
-	xor	tmp2=tmp2,tmp3		    // +1
+{ .mmi;	and	tmp0=tmp0,$b
+	xor	$Xn=$Xn,$X[($j+2)%16]	    // forward Xupdate
+	xor	tmp3=$X[($j+8)%16],$X[($j+13)%16] };;	// forward Xupdate
+{ .mmi;	add	$e=$e,tmp0		    // e+=b&(c^d)
+	add	tmp5=tmp5,tmp1		    // ROTATE(a,5)+(c&d)
 	shrp	$b=tmp6,tmp6,2		}   // b=ROTATE(b,30)
-{ .mmi;	or	tmp0=tmp0,tmp4		    // F_40_59(b,c,d)=(b&c)|(b&d)|(c&d)
+{ .mmi;	xor	$Xn=$Xn,tmp3
 	mux2	tmp6=$a,0x44		};; // see b in next iteration
-{ .mii;	add	$f=$f,tmp0		    // f+=F_40_59(b,c,d)
-	shrp	$e=tmp2,tmp2,31;;	    // f+1=ROTATE(x[0]^x[2]^x[8]^x[13],1)
-	add	$f=$f,tmp1		};; // f+=ROTATE(a,5)
+{ .mii;	add	$e=$e,tmp5		    // e+=ROTATE(a,5)+(c&d)
+	shrp	$Xn=$Xn,$Xn,31		    // ROTATE(x[0]^x[2]^x[8]^x[13],1)
+	nop.i	0x0			};;
 
 ___
 }
@@ -237,7 +235,7 @@ inp=r33;	// in1
 .align	32
 sha1_block_data_order:
 	.prologue
-{ .mmi;	alloc	tmp1=ar.pfs,3,15,0,0
+{ .mmi;	alloc	tmp1=ar.pfs,3,14,0,0
 	$ADDP	tmp0=4,ctx
 	.save	ar.lc,r3
 	mov	r3=ar.lc		}
@@ -245,8 +243,8 @@ sha1_block_data_order:
 	$ADDP	inp=0,inp
 	mov	r2=pr			};;
 tmp4=in2;
-tmp5=loc13;
-tmp6=loc14;
+tmp5=loc12;
+tmp6=loc13;
 	.body
 { .mlx;	ld4	$h0=[ctx],8
 	movl	$K_00_19=0x5a827999	}
@@ -273,7 +271,7 @@ tmp6=loc14;
 
 ___
 
-{ my $i,@V=($A,$B,$C,$D,$E,$T);
+{ my $i,@V=($A,$B,$C,$D,$E);
 
 	for($i=0;$i<16;$i++)	{ &BODY_00_15(\$code,$i,@V); unshift(@V,pop(@V)); }
 	for(;$i<20;$i++)	{ &BODY_16_19(\$code,$i,@V); unshift(@V,pop(@V)); }
@@ -281,12 +279,12 @@ ___
 	for(;$i<60;$i++)	{ &BODY_40_59(\$code,$i,@V); unshift(@V,pop(@V)); }
 	for(;$i<80;$i++)	{ &BODY_60_79(\$code,$i,@V); unshift(@V,pop(@V)); }
 
-	(($V[5] eq $D) and ($V[0] eq $E)) or die;	# double-check
+	(($V[0] eq $A) and ($V[4] eq $E)) or die;	# double-check
 }
 
 $code.=<<___;
-{ .mmb;	add	$h0=$h0,$E
-	nop.m	0
+{ .mmb;	add	$h0=$h0,$A
+	add	$h2=$h2,$C
 	br.ctop.dptk.many	.Ldtop	};;
 .Ldend:
 { .mmi;	add	tmp0=4,ctx
diff --git a/openssl/crypto/sha/asm/sha1-mips.pl b/openssl/crypto/sha/asm/sha1-mips.pl
new file mode 100644
index 000000000..f1a702f38
--- /dev/null
+++ b/openssl/crypto/sha/asm/sha1-mips.pl
@@ -0,0 +1,354 @@
+#!/usr/bin/env perl
+
+# ====================================================================
+# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+
+# SHA1 block procedure for MIPS.
+
+# Performance improvement is 30% on unaligned input. The "secret" is
+# to deploy lwl/lwr pair to load unaligned input. One could have
+# vectorized Xupdate on MIPSIII/IV, but the goal was to code MIPS32-
+# compatible subroutine. There is room for minor optimization on
+# little-endian platforms...
+
+######################################################################
+# There is a number of MIPS ABI in use, O32 and N32/64 are most
+# widely used. Then there is a new contender: NUBI. It appears that if
+# one picks the latter, it's possible to arrange code in ABI neutral
+# manner. Therefore let's stick to NUBI register layout:
+#
+($zero,$at,$t0,$t1,$t2)=map("\$$_",(0..2,24,25));
+($a0,$a1,$a2,$a3,$a4,$a5,$a6,$a7)=map("\$$_",(4..11));
+($s0,$s1,$s2,$s3,$s4,$s5,$s6,$s7,$s8,$s9,$s10,$s11)=map("\$$_",(12..23));
+($gp,$tp,$sp,$fp,$ra)=map("\$$_",(3,28..31));
+#
+# The return value is placed in $a0. Following coding rules facilitate
+# interoperability:
+#
+# - never ever touch $tp, "thread pointer", former $gp;
+# - copy return value to $t0, former $v0 [or to $a0 if you're adapting
+#   old code];
+# - on O32 populate $a4-$a7 with 'lw $aN,4*N($sp)' if necessary;
+#
+# For reference here is register layout for N32/64 MIPS ABIs:
+#
+# ($zero,$at,$v0,$v1)=map("\$$_",(0..3));
+# ($a0,$a1,$a2,$a3,$a4,$a5,$a6,$a7)=map("\$$_",(4..11));
+# ($t0,$t1,$t2,$t3,$t8,$t9)=map("\$$_",(12..15,24,25));
+# ($s0,$s1,$s2,$s3,$s4,$s5,$s6,$s7)=map("\$$_",(16..23));
+# ($gp,$sp,$fp,$ra)=map("\$$_",(28..31));
+#
+$flavour = shift; # supported flavours are o32,n32,64,nubi32,nubi64
+
+if ($flavour =~ /64|n32/i) {
+	$PTR_ADD="dadd";	# incidentally works even on n32
+	$PTR_SUB="dsub";	# incidentally works even on n32
+	$REG_S="sd";
+	$REG_L="ld";
+	$PTR_SLL="dsll";	# incidentally works even on n32
+	$SZREG=8;
+} else {
+	$PTR_ADD="add";
+	$PTR_SUB="sub";
+	$REG_S="sw";
+	$REG_L="lw";
+	$PTR_SLL="sll";
+	$SZREG=4;
+}
+#
+# <appro@openssl.org>
+#
+######################################################################
+
+$big_endian=(`echo MIPSEL | $ENV{CC} -E -P -`=~/MIPSEL/)?1:0;
+
+for (@ARGV) {	$output=$_ if (/^\w[\w\-]*\.\w+$/);   }
+open STDOUT,">$output";
+
+if (!defined($big_endian))
+            {   $big_endian=(unpack('L',pack('N',1))==1);   }
+
+# offsets of the Most and Least Significant Bytes
+$MSB=$big_endian?0:3;
+$LSB=3&~$MSB;
+
+@X=map("\$$_",(8..23));	# a4-a7,s0-s11
+
+$ctx=$a0;
+$inp=$a1;
+$num=$a2;
+$A="\$1";
+$B="\$2";
+$C="\$3";
+$D="\$7";
+$E="\$24";	@V=($A,$B,$C,$D,$E);
+$t0="\$25";
+$t1=$num;	# $num is offloaded to stack
+$t2="\$30";	# fp
+$K="\$31";	# ra
+
+sub BODY_00_14 {
+my ($i,$a,$b,$c,$d,$e)=@_;
+my $j=$i+1;
+$code.=<<___	if (!$big_endian);
+	srl	$t0,@X[$i],24	# byte swap($i)
+	srl	$t1,@X[$i],8
+	andi	$t2,@X[$i],0xFF00
+	sll	@X[$i],@X[$i],24
+	andi	$t1,0xFF00
+	sll	$t2,$t2,8
+	or	@X[$i],$t0
+	or	$t1,$t2
+	or	@X[$i],$t1
+___
+$code.=<<___;
+	 lwl	@X[$j],$j*4+$MSB($inp)
+	sll	$t0,$a,5	# $i
+	addu	$e,$K
+	 lwr	@X[$j],$j*4+$LSB($inp)
+	srl	$t1,$a,27
+	addu	$e,$t0
+	xor	$t0,$c,$d
+	addu	$e,$t1
+	sll	$t2,$b,30
+	and	$t0,$b
+	srl	$b,$b,2
+	xor	$t0,$d
+	addu	$e,@X[$i]
+	or	$b,$t2
+	addu	$e,$t0
+___
+}
+
+sub BODY_15_19 {
+my ($i,$a,$b,$c,$d,$e)=@_;
+my $j=$i+1;
+
+$code.=<<___	if (!$big_endian && $i==15);
+	srl	$t0,@X[$i],24	# byte swap($i)
+	srl	$t1,@X[$i],8
+	andi	$t2,@X[$i],0xFF00
+	sll	@X[$i],@X[$i],24
+	andi	$t1,0xFF00
+	sll	$t2,$t2,8
+	or	@X[$i],$t0
+	or	@X[$i],$t1
+	or	@X[$i],$t2
+___
+$code.=<<___;
+	 xor	@X[$j%16],@X[($j+2)%16]
+	sll	$t0,$a,5	# $i
+	addu	$e,$K
+	srl	$t1,$a,27
+	addu	$e,$t0
+	 xor	@X[$j%16],@X[($j+8)%16]
+	xor	$t0,$c,$d
+	addu	$e,$t1
+	 xor	@X[$j%16],@X[($j+13)%16]
+	sll	$t2,$b,30
+	and	$t0,$b
+	 srl	$t1,@X[$j%16],31
+	 addu	@X[$j%16],@X[$j%16]
+	srl	$b,$b,2
+	xor	$t0,$d
+	 or	@X[$j%16],$t1
+	addu	$e,@X[$i%16]
+	or	$b,$t2
+	addu	$e,$t0
+___
+}
+
+sub BODY_20_39 {
+my ($i,$a,$b,$c,$d,$e)=@_;
+my $j=$i+1;
+$code.=<<___ if ($i<79);
+	 xor	@X[$j%16],@X[($j+2)%16]
+	sll	$t0,$a,5	# $i
+	addu	$e,$K
+	srl	$t1,$a,27
+	addu	$e,$t0
+	 xor	@X[$j%16],@X[($j+8)%16]
+	xor	$t0,$c,$d
+	addu	$e,$t1
+	 xor	@X[$j%16],@X[($j+13)%16]
+	sll	$t2,$b,30
+	xor	$t0,$b
+	 srl	$t1,@X[$j%16],31
+	 addu	@X[$j%16],@X[$j%16]
+	srl	$b,$b,2
+	addu	$e,@X[$i%16]
+	 or	@X[$j%16],$t1
+	or	$b,$t2
+	addu	$e,$t0
+___
+$code.=<<___ if ($i==79);
+	 lw	@X[0],0($ctx)
+	sll	$t0,$a,5	# $i
+	addu	$e,$K
+	 lw	@X[1],4($ctx)
+	srl	$t1,$a,27
+	addu	$e,$t0
+	 lw	@X[2],8($ctx)
+	xor	$t0,$c,$d
+	addu	$e,$t1
+	 lw	@X[3],12($ctx)
+	sll	$t2,$b,30
+	xor	$t0,$b
+	 lw	@X[4],16($ctx)
+	srl	$b,$b,2
+	addu	$e,@X[$i%16]
+	or	$b,$t2
+	addu	$e,$t0
+___
+}
+
+sub BODY_40_59 {
+my ($i,$a,$b,$c,$d,$e)=@_;
+my $j=$i+1;
+$code.=<<___ if ($i<79);
+	 xor	@X[$j%16],@X[($j+2)%16]
+	sll	$t0,$a,5	# $i
+	addu	$e,$K
+	srl	$t1,$a,27
+	addu	$e,$t0
+	 xor	@X[$j%16],@X[($j+8)%16]
+	and	$t0,$c,$d
+	addu	$e,$t1
+	 xor	@X[$j%16],@X[($j+13)%16]
+	sll	$t2,$b,30
+	addu	$e,$t0
+	 srl	$t1,@X[$j%16],31
+	xor	$t0,$c,$d
+	 addu	@X[$j%16],@X[$j%16]
+	and	$t0,$b
+	srl	$b,$b,2
+	 or	@X[$j%16],$t1
+	addu	$e,@X[$i%16]
+	or	$b,$t2
+	addu	$e,$t0
+___
+}
+
+$FRAMESIZE=16;	# large enough to accomodate NUBI saved registers
+$SAVED_REGS_MASK = ($flavour =~ /nubi/i) ? 0xc0fff008 : 0xc0ff0000;
+
+$code=<<___;
+#ifdef OPENSSL_FIPSCANISTER
+# include <openssl/fipssyms.h>
+#endif
+
+.text
+
+.set	noat
+.set	noreorder
+.align	5
+.globl	sha1_block_data_order
+.ent	sha1_block_data_order
+sha1_block_data_order:
+	.frame	$sp,$FRAMESIZE*$SZREG,$ra
+	.mask	$SAVED_REGS_MASK,-$SZREG
+	.set	noreorder
+	$PTR_SUB $sp,$FRAMESIZE*$SZREG
+	$REG_S	$ra,($FRAMESIZE-1)*$SZREG($sp)
+	$REG_S	$fp,($FRAMESIZE-2)*$SZREG($sp)
+	$REG_S	$s11,($FRAMESIZE-3)*$SZREG($sp)
+	$REG_S	$s10,($FRAMESIZE-4)*$SZREG($sp)
+	$REG_S	$s9,($FRAMESIZE-5)*$SZREG($sp)
+	$REG_S	$s8,($FRAMESIZE-6)*$SZREG($sp)
+	$REG_S	$s7,($FRAMESIZE-7)*$SZREG($sp)
+	$REG_S	$s6,($FRAMESIZE-8)*$SZREG($sp)
+	$REG_S	$s5,($FRAMESIZE-9)*$SZREG($sp)
+	$REG_S	$s4,($FRAMESIZE-10)*$SZREG($sp)
+___
+$code.=<<___ if ($flavour =~ /nubi/i);	# optimize non-nubi prologue
+	$REG_S	$s3,($FRAMESIZE-11)*$SZREG($sp)
+	$REG_S	$s2,($FRAMESIZE-12)*$SZREG($sp)
+	$REG_S	$s1,($FRAMESIZE-13)*$SZREG($sp)
+	$REG_S	$s0,($FRAMESIZE-14)*$SZREG($sp)
+	$REG_S	$gp,($FRAMESIZE-15)*$SZREG($sp)
+___
+$code.=<<___;
+	$PTR_SLL $num,6
+	$PTR_ADD $num,$inp
+	$REG_S	$num,0($sp)
+	lw	$A,0($ctx)
+	lw	$B,4($ctx)
+	lw	$C,8($ctx)
+	lw	$D,12($ctx)
+	b	.Loop
+	lw	$E,16($ctx)
+.align	4
+.Loop:
+	.set	reorder
+	lwl	@X[0],$MSB($inp)
+	lui	$K,0x5a82
+	lwr	@X[0],$LSB($inp)
+	ori	$K,0x7999	# K_00_19
+___
+for ($i=0;$i<15;$i++)	{ &BODY_00_14($i,@V); unshift(@V,pop(@V)); }
+for (;$i<20;$i++)	{ &BODY_15_19($i,@V); unshift(@V,pop(@V)); }
+$code.=<<___;
+	lui	$K,0x6ed9
+	ori	$K,0xeba1	# K_20_39
+___
+for (;$i<40;$i++)	{ &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
+$code.=<<___;
+	lui	$K,0x8f1b
+	ori	$K,0xbcdc	# K_40_59
+___
+for (;$i<60;$i++)	{ &BODY_40_59($i,@V); unshift(@V,pop(@V)); }
+$code.=<<___;
+	lui	$K,0xca62
+	ori	$K,0xc1d6	# K_60_79
+___
+for (;$i<80;$i++)	{ &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
+$code.=<<___;
+	$PTR_ADD $inp,64
+	$REG_L	$num,0($sp)
+
+	addu	$A,$X[0]
+	addu	$B,$X[1]
+	sw	$A,0($ctx)
+	addu	$C,$X[2]
+	addu	$D,$X[3]
+	sw	$B,4($ctx)
+	addu	$E,$X[4]
+	sw	$C,8($ctx)
+	sw	$D,12($ctx)
+	sw	$E,16($ctx)
+	.set	noreorder
+	bne	$inp,$num,.Loop
+	nop
+
+	.set	noreorder
+	$REG_L	$ra,($FRAMESIZE-1)*$SZREG($sp)
+	$REG_L	$fp,($FRAMESIZE-2)*$SZREG($sp)
+	$REG_L	$s11,($FRAMESIZE-3)*$SZREG($sp)
+	$REG_L	$s10,($FRAMESIZE-4)*$SZREG($sp)
+	$REG_L	$s9,($FRAMESIZE-5)*$SZREG($sp)
+	$REG_L	$s8,($FRAMESIZE-6)*$SZREG($sp)
+	$REG_L	$s7,($FRAMESIZE-7)*$SZREG($sp)
+	$REG_L	$s6,($FRAMESIZE-8)*$SZREG($sp)
+	$REG_L	$s5,($FRAMESIZE-9)*$SZREG($sp)
+	$REG_L	$s4,($FRAMESIZE-10)*$SZREG($sp)
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+	$REG_L	$s3,($FRAMESIZE-11)*$SZREG($sp)
+	$REG_L	$s2,($FRAMESIZE-12)*$SZREG($sp)
+	$REG_L	$s1,($FRAMESIZE-13)*$SZREG($sp)
+	$REG_L	$s0,($FRAMESIZE-14)*$SZREG($sp)
+	$REG_L	$gp,($FRAMESIZE-15)*$SZREG($sp)
+___
+$code.=<<___;
+	jr	$ra
+	$PTR_ADD $sp,$FRAMESIZE*$SZREG
+.end	sha1_block_data_order
+.rdata
+.asciiz	"SHA1 for MIPS, CRYPTOGAMS by <appro\@openssl.org>"
+___
+print $code;
+close STDOUT;
diff --git a/openssl/crypto/sha/asm/sha1-parisc.pl b/openssl/crypto/sha/asm/sha1-parisc.pl
new file mode 100644
index 000000000..6d7bf495b
--- /dev/null
+++ b/openssl/crypto/sha/asm/sha1-parisc.pl
@@ -0,0 +1,259 @@
+#!/usr/bin/env perl
+
+# ====================================================================
+# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+
+# SHA1 block procedure for PA-RISC.
+
+# June 2009.
+#
+# On PA-7100LC performance is >30% better than gcc 3.2 generated code
+# for aligned input and >50% better for unaligned. Compared to vendor
+# compiler on PA-8600 it's almost 60% faster in 64-bit build and just
+# few percent faster in 32-bit one (this for aligned input, data for
+# unaligned input is not available).
+#
+# Special thanks to polarhome.com for providing HP-UX account.
+
+$flavour = shift;
+$output = shift;
+open STDOUT,">$output";
+
+if ($flavour =~ /64/) {
+	$LEVEL		="2.0W";
+	$SIZE_T		=8;
+	$FRAME_MARKER	=80;
+	$SAVED_RP	=16;
+	$PUSH		="std";
+	$PUSHMA		="std,ma";
+	$POP		="ldd";
+	$POPMB		="ldd,mb";
+} else {
+	$LEVEL		="1.0";
+	$SIZE_T		=4;
+	$FRAME_MARKER	=48;
+	$SAVED_RP	=20;
+	$PUSH		="stw";
+	$PUSHMA		="stwm";
+	$POP		="ldw";
+	$POPMB		="ldwm";
+}
+
+$FRAME=14*$SIZE_T+$FRAME_MARKER;# 14 saved regs + frame marker
+				#                 [+ argument transfer]
+$ctx="%r26";		# arg0
+$inp="%r25";		# arg1
+$num="%r24";		# arg2
+
+$t0="%r28";
+$t1="%r29";
+$K="%r31";
+
+@X=("%r1", "%r2", "%r3", "%r4", "%r5", "%r6", "%r7", "%r8",
+    "%r9", "%r10","%r11","%r12","%r13","%r14","%r15","%r16",$t0);
+
+@V=($A,$B,$C,$D,$E)=("%r19","%r20","%r21","%r22","%r23");
+
+sub BODY_00_19 {
+my ($i,$a,$b,$c,$d,$e)=@_;
+my $j=$i+1;
+$code.=<<___ if ($i<15);
+	addl	$K,$e,$e	; $i
+	shd	$a,$a,27,$t1
+	addl	@X[$i],$e,$e
+	and	$c,$b,$t0
+	addl	$t1,$e,$e
+	andcm	$d,$b,$t1
+	shd	$b,$b,2,$b
+	or	$t1,$t0,$t0
+	addl	$t0,$e,$e
+___
+$code.=<<___ if ($i>=15);	# with forward Xupdate
+	addl	$K,$e,$e	; $i
+	shd	$a,$a,27,$t1
+	xor	@X[($j+2)%16],@X[$j%16],@X[$j%16]
+	addl	@X[$i%16],$e,$e
+	and	$c,$b,$t0
+	xor	@X[($j+8)%16],@X[$j%16],@X[$j%16]
+	addl	$t1,$e,$e
+	andcm	$d,$b,$t1
+	shd	$b,$b,2,$b
+	or	$t1,$t0,$t0
+	xor	@X[($j+13)%16],@X[$j%16],@X[$j%16]
+	add	$t0,$e,$e
+	shd	@X[$j%16],@X[$j%16],31,@X[$j%16]
+___
+}
+
+sub BODY_20_39 {
+my ($i,$a,$b,$c,$d,$e)=@_;
+my $j=$i+1;
+$code.=<<___ if ($i<79);
+	xor	@X[($j+2)%16],@X[$j%16],@X[$j%16]	; $i
+	addl	$K,$e,$e
+	shd	$a,$a,27,$t1
+	xor	@X[($j+8)%16],@X[$j%16],@X[$j%16]
+	addl	@X[$i%16],$e,$e
+	xor	$b,$c,$t0
+	xor	@X[($j+13)%16],@X[$j%16],@X[$j%16]
+	addl	$t1,$e,$e
+	shd	$b,$b,2,$b
+	xor	$d,$t0,$t0
+	shd	@X[$j%16],@X[$j%16],31,@X[$j%16]
+	addl	$t0,$e,$e
+___
+$code.=<<___ if ($i==79);	# with context load
+	ldw	0($ctx),@X[0]	; $i
+	addl	$K,$e,$e
+	shd	$a,$a,27,$t1
+	ldw	4($ctx),@X[1]
+	addl	@X[$i%16],$e,$e
+	xor	$b,$c,$t0
+	ldw	8($ctx),@X[2]
+	addl	$t1,$e,$e
+	shd	$b,$b,2,$b
+	xor	$d,$t0,$t0
+	ldw	12($ctx),@X[3]
+	addl	$t0,$e,$e
+	ldw	16($ctx),@X[4]
+___
+}
+
+sub BODY_40_59 {
+my ($i,$a,$b,$c,$d,$e)=@_;
+my $j=$i+1;
+$code.=<<___;
+	shd	$a,$a,27,$t1	; $i
+	addl	$K,$e,$e
+	xor	@X[($j+2)%16],@X[$j%16],@X[$j%16]
+	xor	$d,$c,$t0
+	addl	@X[$i%16],$e,$e
+	xor	@X[($j+8)%16],@X[$j%16],@X[$j%16]
+	and	$b,$t0,$t0
+	addl	$t1,$e,$e
+	shd	$b,$b,2,$b
+	xor	@X[($j+13)%16],@X[$j%16],@X[$j%16]
+	addl	$t0,$e,$e
+	and	$d,$c,$t1
+	shd	@X[$j%16],@X[$j%16],31,@X[$j%16]
+	addl	$t1,$e,$e
+___
+}
+
+$code=<<___;
+	.LEVEL	$LEVEL
+	.SPACE	\$TEXT\$
+	.SUBSPA	\$CODE\$,QUAD=0,ALIGN=8,ACCESS=0x2C,CODE_ONLY
+
+	.EXPORT	sha1_block_data_order,ENTRY,ARGW0=GR,ARGW1=GR,ARGW2=GR
+sha1_block_data_order
+	.PROC
+	.CALLINFO	FRAME=`$FRAME-14*$SIZE_T`,NO_CALLS,SAVE_RP,ENTRY_GR=16
+	.ENTRY
+	$PUSH	%r2,-$SAVED_RP(%sp)	; standard prologue
+	$PUSHMA	%r3,$FRAME(%sp)
+	$PUSH	%r4,`-$FRAME+1*$SIZE_T`(%sp)
+	$PUSH	%r5,`-$FRAME+2*$SIZE_T`(%sp)
+	$PUSH	%r6,`-$FRAME+3*$SIZE_T`(%sp)
+	$PUSH	%r7,`-$FRAME+4*$SIZE_T`(%sp)
+	$PUSH	%r8,`-$FRAME+5*$SIZE_T`(%sp)
+	$PUSH	%r9,`-$FRAME+6*$SIZE_T`(%sp)
+	$PUSH	%r10,`-$FRAME+7*$SIZE_T`(%sp)
+	$PUSH	%r11,`-$FRAME+8*$SIZE_T`(%sp)
+	$PUSH	%r12,`-$FRAME+9*$SIZE_T`(%sp)
+	$PUSH	%r13,`-$FRAME+10*$SIZE_T`(%sp)
+	$PUSH	%r14,`-$FRAME+11*$SIZE_T`(%sp)
+	$PUSH	%r15,`-$FRAME+12*$SIZE_T`(%sp)
+	$PUSH	%r16,`-$FRAME+13*$SIZE_T`(%sp)
+
+	ldw	0($ctx),$A
+	ldw	4($ctx),$B
+	ldw	8($ctx),$C
+	ldw	12($ctx),$D
+	ldw	16($ctx),$E
+
+	extru	$inp,31,2,$t0		; t0=inp&3;
+	sh3addl	$t0,%r0,$t0		; t0*=8;
+	subi	32,$t0,$t0		; t0=32-t0;
+	mtctl	$t0,%cr11		; %sar=t0;
+
+L\$oop
+	ldi	3,$t0
+	andcm	$inp,$t0,$t0		; 64-bit neutral
+___
+	for ($i=0;$i<15;$i++) {		# load input block
+	$code.="\tldw	`4*$i`($t0),@X[$i]\n";		}
+$code.=<<___;
+	cmpb,*=	$inp,$t0,L\$aligned
+	ldw	60($t0),@X[15]
+	ldw	64($t0),@X[16]
+___
+	for ($i=0;$i<16;$i++) {		# align input
+	$code.="\tvshd	@X[$i],@X[$i+1],@X[$i]\n";	}
+$code.=<<___;
+L\$aligned
+	ldil	L'0x5a827000,$K		; K_00_19
+	ldo	0x999($K),$K
+___
+for ($i=0;$i<20;$i++)   { &BODY_00_19($i,@V); unshift(@V,pop(@V)); }
+$code.=<<___;
+	ldil	L'0x6ed9e000,$K		; K_20_39
+	ldo	0xba1($K),$K
+___
+
+for (;$i<40;$i++)       { &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
+$code.=<<___;
+	ldil	L'0x8f1bb000,$K		; K_40_59
+	ldo	0xcdc($K),$K
+___
+
+for (;$i<60;$i++)       { &BODY_40_59($i,@V); unshift(@V,pop(@V)); }
+$code.=<<___;
+	ldil	L'0xca62c000,$K		; K_60_79
+	ldo	0x1d6($K),$K
+___
+for (;$i<80;$i++)       { &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
+
+$code.=<<___;
+	addl	@X[0],$A,$A
+	addl	@X[1],$B,$B
+	addl	@X[2],$C,$C
+	addl	@X[3],$D,$D
+	addl	@X[4],$E,$E
+	stw	$A,0($ctx)
+	stw	$B,4($ctx)
+	stw	$C,8($ctx)
+	stw	$D,12($ctx)
+	stw	$E,16($ctx)
+	addib,*<> -1,$num,L\$oop
+	ldo	64($inp),$inp
+
+	$POP	`-$FRAME-$SAVED_RP`(%sp),%r2	; standard epilogue
+	$POP	`-$FRAME+1*$SIZE_T`(%sp),%r4
+	$POP	`-$FRAME+2*$SIZE_T`(%sp),%r5
+	$POP	`-$FRAME+3*$SIZE_T`(%sp),%r6
+	$POP	`-$FRAME+4*$SIZE_T`(%sp),%r7
+	$POP	`-$FRAME+5*$SIZE_T`(%sp),%r8
+	$POP	`-$FRAME+6*$SIZE_T`(%sp),%r9
+	$POP	`-$FRAME+7*$SIZE_T`(%sp),%r10
+	$POP	`-$FRAME+8*$SIZE_T`(%sp),%r11
+	$POP	`-$FRAME+9*$SIZE_T`(%sp),%r12
+	$POP	`-$FRAME+10*$SIZE_T`(%sp),%r13
+	$POP	`-$FRAME+11*$SIZE_T`(%sp),%r14
+	$POP	`-$FRAME+12*$SIZE_T`(%sp),%r15
+	$POP	`-$FRAME+13*$SIZE_T`(%sp),%r16
+	bv	(%r2)
+	.EXIT
+	$POPMB	-$FRAME(%sp),%r3
+	.PROCEND
+	.STRINGZ "SHA1 block transform for PA-RISC, CRYPTOGAMS by <appro\@openssl.org>"
+___
+
+$code =~ s/\`([^\`]*)\`/eval $1/gem;
+$code =~ s/,\*/,/gm if ($SIZE_T==4);
+print $code;
+close STDOUT;
diff --git a/openssl/crypto/sha/asm/sha1-ppc.pl b/openssl/crypto/sha/asm/sha1-ppc.pl
index dcd0fcdfc..2140dd2f8 100644
--- a/openssl/crypto/sha/asm/sha1-ppc.pl
+++ b/openssl/crypto/sha/asm/sha1-ppc.pl
@@ -24,12 +24,14 @@ $flavour = shift;
 
 if ($flavour =~ /64/) {
 	$SIZE_T	=8;
+	$LRSAVE	=2*$SIZE_T;
 	$UCMP	="cmpld";
 	$STU	="stdu";
 	$POP	="ld";
 	$PUSH	="std";
 } elsif ($flavour =~ /32/) {
 	$SIZE_T	=4;
+	$LRSAVE	=$SIZE_T;
 	$UCMP	="cmplw";
 	$STU	="stwu";
 	$POP	="lwz";
@@ -43,7 +45,8 @@ die "can't locate ppc-xlate.pl";
 
 open STDOUT,"| $^X $xlate $flavour ".shift || die "can't call $xlate: $!";
 
-$FRAME=24*$SIZE_T;
+$FRAME=24*$SIZE_T+64;
+$LOCALS=6*$SIZE_T;
 
 $K  ="r0";
 $sp ="r1";
@@ -162,9 +165,8 @@ $code=<<___;
 .globl	.sha1_block_data_order
 .align	4
 .sha1_block_data_order:
+	$STU	$sp,-$FRAME($sp)
 	mflr	r0
-	$STU	$sp,`-($FRAME+64)`($sp)
-	$PUSH	r0,`$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)
@@ -182,6 +184,7 @@ $code=<<___;
 	$PUSH	r29,`$FRAME-$SIZE_T*3`($sp)
 	$PUSH	r30,`$FRAME-$SIZE_T*2`($sp)
 	$PUSH	r31,`$FRAME-$SIZE_T*1`($sp)
+	$PUSH	r0,`$FRAME+$LRSAVE`($sp)
 	lwz	$A,0($ctx)
 	lwz	$B,4($ctx)
 	lwz	$C,8($ctx)
@@ -192,37 +195,14 @@ $code=<<___;
 Laligned:
 	mtctr	$num
 	bl	Lsha1_block_private
-Ldone:
-	$POP	r0,`$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+64`
-	blr
-___
+	b	Ldone
 
-# PowerPC specification allows an implementation to be ill-behaved
-# upon unaligned access which crosses page boundary. "Better safe
-# than sorry" principle makes me treat it specially. But I don't
-# look for particular offending word, but rather for 64-byte input
-# block which crosses the boundary. Once found that block is aligned
-# and hashed separately...
-$code.=<<___;
+; PowerPC specification allows an implementation to be ill-behaved
+; upon unaligned access which crosses page boundary. "Better safe
+; than sorry" principle makes me treat it specially. But I don't
+; look for particular offending word, but rather for 64-byte input
+; block which crosses the boundary. Once found that block is aligned
+; and hashed separately...
 .align	4
 Lunaligned:
 	subfic	$t1,$inp,4096
@@ -237,7 +217,7 @@ Lunaligned:
 Lcross_page:
 	li	$t1,16
 	mtctr	$t1
-	addi	r20,$sp,$FRAME	; spot below the frame
+	addi	r20,$sp,$LOCALS	; spot within the frame
 Lmemcpy:
 	lbz	r16,0($inp)
 	lbz	r17,1($inp)
@@ -251,15 +231,40 @@ Lmemcpy:
 	addi	r20,r20,4
 	bdnz	Lmemcpy
 
-	$PUSH	$inp,`$FRAME-$SIZE_T*19`($sp)
+	$PUSH	$inp,`$FRAME-$SIZE_T*18`($sp)
 	li	$t1,1
-	addi	$inp,$sp,$FRAME
+	addi	$inp,$sp,$LOCALS
 	mtctr	$t1
 	bl	Lsha1_block_private
-	$POP	$inp,`$FRAME-$SIZE_T*19`($sp)
+	$POP	$inp,`$FRAME-$SIZE_T*18`($sp)
 	addic.	$num,$num,-1
 	bne-	Lunaligned
-	b	Ldone
+
+Ldone:
+	$POP	r0,`$FRAME+$LRSAVE`($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
+	.long	0
+	.byte	0,12,4,1,0x80,18,3,0
+	.long	0
 ___
 
 # This is private block function, which uses tailored calling
@@ -309,6 +314,8 @@ $code.=<<___;
 	addi	$inp,$inp,`16*4`
 	bdnz-	Lsha1_block_private
 	blr
+	.long	0
+	.byte	0,12,0x14,0,0,0,0,0
 ___
 $code.=<<___;
 .asciz	"SHA1 block transform for PPC, CRYPTOGAMS by <appro\@fy.chalmers.se>"
diff --git a/openssl/crypto/sha/asm/sha1-s390x.pl b/openssl/crypto/sha/asm/sha1-s390x.pl
index 4b1784828..9193dda45 100644
--- a/openssl/crypto/sha/asm/sha1-s390x.pl
+++ b/openssl/crypto/sha/asm/sha1-s390x.pl
@@ -21,9 +21,28 @@
 # instructions to favour dual-issue z10 pipeline. On z10 hardware is
 # "only" ~2.3x faster than software.
 
+# November 2010.
+#
+# Adapt for -m31 build. If kernel supports what's called "highgprs"
+# feature on Linux [see /proc/cpuinfo], it's possible to use 64-bit
+# instructions and achieve "64-bit" performance even in 31-bit legacy
+# application context. The feature is not specific to any particular
+# processor, as long as it's "z-CPU". Latter implies that the code
+# remains z/Architecture specific.
+
 $kimdfunc=1;	# magic function code for kimd instruction
 
-$output=shift;
+$flavour = shift;
+
+if ($flavour =~ /3[12]/) {
+	$SIZE_T=4;
+	$g="";
+} else {
+	$SIZE_T=8;
+	$g="g";
+}
+
+while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {}
 open STDOUT,">$output";
 
 $K_00_39="%r0"; $K=$K_00_39;
@@ -42,13 +61,14 @@ $t1="%r11";
 @X=("%r12","%r13","%r14");
 $sp="%r15";
 
-$frame=160+16*4;
+$stdframe=16*$SIZE_T+4*8;
+$frame=$stdframe+16*4;
 
 sub Xupdate {
 my $i=shift;
 
 $code.=<<___ if ($i==15);
-	lg	$prefetch,160($sp)	### Xupdate(16) warm-up
+	lg	$prefetch,$stdframe($sp)	### Xupdate(16) warm-up
 	lr	$X[0],$X[2]
 ___
 return if ($i&1);	# Xupdate is vectorized and executed every 2nd cycle
@@ -58,8 +78,8 @@ $code.=<<___ if ($i<16);
 ___
 $code.=<<___ if ($i>=16);
 	xgr	$X[0],$prefetch		### Xupdate($i)
-	lg	$prefetch,`160+4*(($i+2)%16)`($sp)
-	xg	$X[0],`160+4*(($i+8)%16)`($sp)
+	lg	$prefetch,`$stdframe+4*(($i+2)%16)`($sp)
+	xg	$X[0],`$stdframe+4*(($i+8)%16)`($sp)
 	xgr	$X[0],$prefetch
 	rll	$X[0],$X[0],1
 	rllg	$X[1],$X[0],32
@@ -68,7 +88,7 @@ $code.=<<___ if ($i>=16);
 	lr	$X[2],$X[1]		# feedback
 ___
 $code.=<<___ if ($i<=70);
-	stg	$X[0],`160+4*($i%16)`($sp)
+	stg	$X[0],`$stdframe+4*($i%16)`($sp)
 ___
 unshift(@X,pop(@X));
 }
@@ -148,9 +168,9 @@ $code.=<<___ if ($kimdfunc);
 	tmhl	%r0,0x4000	# check for message-security assist
 	jz	.Lsoftware
 	lghi	%r0,0
-	la	%r1,16($sp)
+	la	%r1,`2*$SIZE_T`($sp)
 	.long	0xb93e0002	# kimd %r0,%r2
-	lg	%r0,16($sp)
+	lg	%r0,`2*$SIZE_T`($sp)
 	tmhh	%r0,`0x8000>>$kimdfunc`
 	jz	.Lsoftware
 	lghi	%r0,$kimdfunc
@@ -165,11 +185,11 @@ $code.=<<___ if ($kimdfunc);
 ___
 $code.=<<___;
 	lghi	%r1,-$frame
-	stg	$ctx,16($sp)
-	stmg	%r6,%r15,48($sp)
+	st${g}	$ctx,`2*$SIZE_T`($sp)
+	stm${g}	%r6,%r15,`6*$SIZE_T`($sp)
 	lgr	%r0,$sp
 	la	$sp,0(%r1,$sp)
-	stg	%r0,0($sp)
+	st${g}	%r0,0($sp)
 
 	larl	$t0,Ktable
 	llgf	$A,0($ctx)
@@ -199,7 +219,7 @@ ___
 for (;$i<80;$i++)	{ &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
 $code.=<<___;
 
-	lg	$ctx,`$frame+16`($sp)
+	l${g}	$ctx,`$frame+2*$SIZE_T`($sp)
 	la	$inp,64($inp)
 	al	$A,0($ctx)
 	al	$B,4($ctx)
@@ -211,13 +231,13 @@ $code.=<<___;
 	st	$C,8($ctx)
 	st	$D,12($ctx)
 	st	$E,16($ctx)
-	brct	$len,.Lloop
+	brct${g} $len,.Lloop
 
-	lmg	%r6,%r15,`$frame+48`($sp)
+	lm${g}	%r6,%r15,`$frame+6*$SIZE_T`($sp)
 	br	%r14
 .size	sha1_block_data_order,.-sha1_block_data_order
 .string	"SHA1 block transform for s390x, CRYPTOGAMS by <appro\@openssl.org>"
-.comm	OPENSSL_s390xcap_P,8,8
+.comm	OPENSSL_s390xcap_P,16,8
 ___
 
 $code =~ s/\`([^\`]*)\`/eval $1/gem;
diff --git a/openssl/crypto/sha/asm/sha1-x86_64.pl b/openssl/crypto/sha/asm/sha1-x86_64.pl
index 4edc5ea9a..f27c1e3fb 100644
--- a/openssl/crypto/sha/asm/sha1-x86_64.pl
+++ b/openssl/crypto/sha/asm/sha1-x86_64.pl
@@ -16,7 +16,7 @@
 # There was suggestion to mechanically translate 32-bit code, but I
 # dismissed it, reasoning that x86_64 offers enough register bank
 # capacity to fully utilize SHA-1 parallelism. Therefore this fresh
-# implementation:-) However! While 64-bit code does performs better
+# implementation:-) However! While 64-bit code does perform better
 # on Opteron, I failed to beat 32-bit assembler on EM64T core. Well,
 # x86_64 does offer larger *addressable* bank, but out-of-order core
 # reaches for even more registers through dynamic aliasing, and EM64T
@@ -29,6 +29,38 @@
 # Xeon P4	+65%		+0%		9.9
 # Core2		+60%		+10%		7.0
 
+# August 2009.
+#
+# The code was revised to minimize code size and to maximize
+# "distance" between instructions producing input to 'lea'
+# instruction and the 'lea' instruction itself, which is essential
+# for Intel Atom core.
+
+# October 2010.
+#
+# Add SSSE3, Supplemental[!] SSE3, implementation. The idea behind it
+# is to offload message schedule denoted by Wt in NIST specification,
+# or Xupdate in OpenSSL source, to SIMD unit. See sha1-586.pl module
+# for background and implementation details. The only difference from
+# 32-bit code is that 64-bit code doesn't have to spill @X[] elements
+# to free temporary registers.
+
+# April 2011.
+#
+# Add AVX code path. See sha1-586.pl for further information.
+
+######################################################################
+# Current performance is summarized in following table. Numbers are
+# CPU clock cycles spent to process single byte (less is better).
+#
+#		x86_64		SSSE3		AVX
+# P4		9.8		-
+# Opteron	6.6		-
+# Core2		6.7		6.1/+10%	-
+# Atom		11.0		9.7/+13%	-
+# Westmere	7.1		5.6/+27%	-
+# Sandy Bridge	7.9		6.3/+25%	5.2/+51%
+
 $flavour = shift;
 $output  = shift;
 if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
@@ -40,6 +72,16 @@ $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
 ( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or
 die "can't locate x86_64-xlate.pl";
 
+$avx=1 if (`$ENV{CC} -Wa,-v -c -o /dev/null -x assembler /dev/null 2>&1`
+		=~ /GNU assembler version ([2-9]\.[0-9]+)/ &&
+	   $1>=2.19);
+$avx=1 if (!$avx && $win64 && ($flavour =~ /nasm/ || $ENV{ASM} =~ /nasm/) &&
+	   `nasm -v 2>&1` =~ /NASM version ([2-9]\.[0-9]+)/ &&
+	   $1>=2.09);
+$avx=1 if (!$avx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
+	   `ml64 2>&1` =~ /Version ([0-9]+)\./ &&
+	   $1>=10);
+
 open STDOUT,"| $^X $xlate $flavour $output";
 
 $ctx="%rdi";	# 1st arg
@@ -51,196 +93,994 @@ $ctx="%r8";
 $inp="%r9";
 $num="%r10";
 
-$xi="%eax";
-$t0="%ebx";
-$t1="%ecx";
-$A="%edx";
-$B="%esi";
-$C="%edi";
-$D="%ebp";
-$E="%r11d";
-$T="%r12d";
-
-@V=($A,$B,$C,$D,$E,$T);
+$t0="%eax";
+$t1="%ebx";
+$t2="%ecx";
+@xi=("%edx","%ebp");
+$A="%esi";
+$B="%edi";
+$C="%r11d";
+$D="%r12d";
+$E="%r13d";
 
-sub PROLOGUE {
-my $func=shift;
-$code.=<<___;
-.globl	$func
-.type	$func,\@function,3
-.align	16
-$func:
-	push	%rbx
-	push	%rbp
-	push	%r12
-	mov	%rsp,%r11
-	mov	%rdi,$ctx	# reassigned argument
-	sub	\$`8+16*4`,%rsp
-	mov	%rsi,$inp	# reassigned argument
-	and	\$-64,%rsp
-	mov	%rdx,$num	# reassigned argument
-	mov	%r11,`16*4`(%rsp)
-.Lprologue:
-
-	mov	0($ctx),$A
-	mov	4($ctx),$B
-	mov	8($ctx),$C
-	mov	12($ctx),$D
-	mov	16($ctx),$E
-___
-}
-
-sub EPILOGUE {
-my $func=shift;
-$code.=<<___;
-	mov	`16*4`(%rsp),%rsi
-	mov	(%rsi),%r12
-	mov	8(%rsi),%rbp
-	mov	16(%rsi),%rbx
-	lea	24(%rsi),%rsp
-.Lepilogue:
-	ret
-.size	$func,.-$func
-___
-}
+@V=($A,$B,$C,$D,$E);
 
 sub BODY_00_19 {
-my ($i,$a,$b,$c,$d,$e,$f,$host)=@_;
+my ($i,$a,$b,$c,$d,$e)=@_;
 my $j=$i+1;
 $code.=<<___ if ($i==0);
-	mov	`4*$i`($inp),$xi	
-	`"bswap	$xi"	if(!defined($host))`
-	mov	$xi,`4*$i`(%rsp)
+	mov	`4*$i`($inp),$xi[0]
+	bswap	$xi[0]
+	mov	$xi[0],`4*$i`(%rsp)
 ___
 $code.=<<___ if ($i<15);
-	lea	0x5a827999($xi,$e),$f
 	mov	$c,$t0
-	mov	`4*$j`($inp),$xi
-	mov	$a,$e
+	mov	`4*$j`($inp),$xi[1]
+	mov	$a,$t2
 	xor	$d,$t0
-	`"bswap	$xi"	if(!defined($host))`	
-	rol	\$5,$e
+	bswap	$xi[1]
+	rol	\$5,$t2
+	lea	0x5a827999($xi[0],$e),$e
 	and	$b,$t0
-	mov	$xi,`4*$j`(%rsp)
-	add	$e,$f
+	mov	$xi[1],`4*$j`(%rsp)
+	add	$t2,$e
 	xor	$d,$t0
 	rol	\$30,$b
-	add	$t0,$f
+	add	$t0,$e
 ___
 $code.=<<___ if ($i>=15);
-	lea	0x5a827999($xi,$e),$f
-	mov	`4*($j%16)`(%rsp),$xi
+	mov	`4*($j%16)`(%rsp),$xi[1]
 	mov	$c,$t0
-	mov	$a,$e
-	xor	`4*(($j+2)%16)`(%rsp),$xi
+	mov	$a,$t2
+	xor	`4*(($j+2)%16)`(%rsp),$xi[1]
 	xor	$d,$t0
-	rol	\$5,$e
-	xor	`4*(($j+8)%16)`(%rsp),$xi
+	rol	\$5,$t2
+	xor	`4*(($j+8)%16)`(%rsp),$xi[1]
 	and	$b,$t0
-	add	$e,$f
-	xor	`4*(($j+13)%16)`(%rsp),$xi
+	lea	0x5a827999($xi[0],$e),$e
+	xor	`4*(($j+13)%16)`(%rsp),$xi[1]
 	xor	$d,$t0
+	rol	\$1,$xi[1]
+	add	$t2,$e
 	rol	\$30,$b
-	add	$t0,$f
-	rol	\$1,$xi
-	mov	$xi,`4*($j%16)`(%rsp)
+	mov	$xi[1],`4*($j%16)`(%rsp)
+	add	$t0,$e
 ___
+unshift(@xi,pop(@xi));
 }
 
 sub BODY_20_39 {
-my ($i,$a,$b,$c,$d,$e,$f)=@_;
+my ($i,$a,$b,$c,$d,$e)=@_;
 my $j=$i+1;
 my $K=($i<40)?0x6ed9eba1:0xca62c1d6;
 $code.=<<___ if ($i<79);
-	lea	$K($xi,$e),$f
-	mov	`4*($j%16)`(%rsp),$xi
+	mov	`4*($j%16)`(%rsp),$xi[1]
 	mov	$c,$t0
-	mov	$a,$e
-	xor	`4*(($j+2)%16)`(%rsp),$xi
+	mov	$a,$t2
+	xor	`4*(($j+2)%16)`(%rsp),$xi[1]
 	xor	$b,$t0
-	rol	\$5,$e
-	xor	`4*(($j+8)%16)`(%rsp),$xi
+	rol	\$5,$t2
+	lea	$K($xi[0],$e),$e
+	xor	`4*(($j+8)%16)`(%rsp),$xi[1]
 	xor	$d,$t0
-	add	$e,$f
-	xor	`4*(($j+13)%16)`(%rsp),$xi
+	add	$t2,$e
+	xor	`4*(($j+13)%16)`(%rsp),$xi[1]
 	rol	\$30,$b
-	add	$t0,$f
-	rol	\$1,$xi
+	add	$t0,$e
+	rol	\$1,$xi[1]
 ___
 $code.=<<___ if ($i<76);
-	mov	$xi,`4*($j%16)`(%rsp)
+	mov	$xi[1],`4*($j%16)`(%rsp)
 ___
 $code.=<<___ if ($i==79);
-	lea	$K($xi,$e),$f
 	mov	$c,$t0
-	mov	$a,$e
+	mov	$a,$t2
 	xor	$b,$t0
-	rol	\$5,$e
+	lea	$K($xi[0],$e),$e
+	rol	\$5,$t2
 	xor	$d,$t0
-	add	$e,$f
+	add	$t2,$e
 	rol	\$30,$b
-	add	$t0,$f
+	add	$t0,$e
 ___
+unshift(@xi,pop(@xi));
 }
 
 sub BODY_40_59 {
-my ($i,$a,$b,$c,$d,$e,$f)=@_;
+my ($i,$a,$b,$c,$d,$e)=@_;
 my $j=$i+1;
 $code.=<<___;
-	lea	0x8f1bbcdc($xi,$e),$f
-	mov	`4*($j%16)`(%rsp),$xi
-	mov	$b,$t0
-	mov	$b,$t1
-	xor	`4*(($j+2)%16)`(%rsp),$xi
-	mov	$a,$e
-	and	$c,$t0
-	xor	`4*(($j+8)%16)`(%rsp),$xi
-	or	$c,$t1
-	rol	\$5,$e
-	xor	`4*(($j+13)%16)`(%rsp),$xi
-	and	$d,$t1
-	add	$e,$f
-	rol	\$1,$xi
-	or	$t1,$t0
+	mov	`4*($j%16)`(%rsp),$xi[1]
+	mov	$c,$t0
+	mov	$c,$t1
+	xor	`4*(($j+2)%16)`(%rsp),$xi[1]
+	and	$d,$t0
+	mov	$a,$t2
+	xor	`4*(($j+8)%16)`(%rsp),$xi[1]
+	xor	$d,$t1
+	lea	0x8f1bbcdc($xi[0],$e),$e
+	rol	\$5,$t2
+	xor	`4*(($j+13)%16)`(%rsp),$xi[1]
+	add	$t0,$e
+	and	$b,$t1
+	rol	\$1,$xi[1]
+	add	$t1,$e
 	rol	\$30,$b
-	mov	$xi,`4*($j%16)`(%rsp)
-	add	$t0,$f
+	mov	$xi[1],`4*($j%16)`(%rsp)
+	add	$t2,$e
 ___
+unshift(@xi,pop(@xi));
 }
 
-$code=".text\n";
+$code.=<<___;
+.text
+.extern	OPENSSL_ia32cap_P
 
-&PROLOGUE("sha1_block_data_order");
-$code.=".align	4\n.Lloop:\n";
+.globl	sha1_block_data_order
+.type	sha1_block_data_order,\@function,3
+.align	16
+sha1_block_data_order:
+	mov	OPENSSL_ia32cap_P+0(%rip),%r9d
+	mov	OPENSSL_ia32cap_P+4(%rip),%r8d
+	test	\$`1<<9`,%r8d		# check SSSE3 bit
+	jz	.Lialu
+___
+$code.=<<___ if ($avx);
+	and	\$`1<<28`,%r8d		# mask AVX bit
+	and	\$`1<<30`,%r9d		# mask "Intel CPU" bit
+	or	%r9d,%r8d
+	cmp	\$`1<<28|1<<30`,%r8d
+	je	_avx_shortcut
+___
+$code.=<<___;
+	jmp	_ssse3_shortcut
+
+.align	16
+.Lialu:
+	push	%rbx
+	push	%rbp
+	push	%r12
+	push	%r13
+	mov	%rsp,%r11
+	mov	%rdi,$ctx	# reassigned argument
+	sub	\$`8+16*4`,%rsp
+	mov	%rsi,$inp	# reassigned argument
+	and	\$-64,%rsp
+	mov	%rdx,$num	# reassigned argument
+	mov	%r11,`16*4`(%rsp)
+.Lprologue:
+
+	mov	0($ctx),$A
+	mov	4($ctx),$B
+	mov	8($ctx),$C
+	mov	12($ctx),$D
+	mov	16($ctx),$E
+	jmp	.Lloop
+
+.align	16
+.Lloop:
+___
 for($i=0;$i<20;$i++)	{ &BODY_00_19($i,@V); unshift(@V,pop(@V)); }
 for(;$i<40;$i++)	{ &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
 for(;$i<60;$i++)	{ &BODY_40_59($i,@V); unshift(@V,pop(@V)); }
 for(;$i<80;$i++)	{ &BODY_20_39($i,@V); unshift(@V,pop(@V)); }
 $code.=<<___;
-	add	0($ctx),$E
-	add	4($ctx),$T
-	add	8($ctx),$A
-	add	12($ctx),$B
-	add	16($ctx),$C
-	mov	$E,0($ctx)
-	mov	$T,4($ctx)
-	mov	$A,8($ctx)
-	mov	$B,12($ctx)
-	mov	$C,16($ctx)
-
-	xchg	$E,$A	# mov	$E,$A
-	xchg	$T,$B	# mov	$T,$B
-	xchg	$E,$C	# mov	$A,$C
-	xchg	$T,$D	# mov	$B,$D
-			# mov	$C,$E
-	lea	`16*4`($inp),$inp
+	add	0($ctx),$A
+	add	4($ctx),$B
+	add	8($ctx),$C
+	add	12($ctx),$D
+	add	16($ctx),$E
+	mov	$A,0($ctx)
+	mov	$B,4($ctx)
+	mov	$C,8($ctx)
+	mov	$D,12($ctx)
+	mov	$E,16($ctx)
+
 	sub	\$1,$num
+	lea	`16*4`($inp),$inp
 	jnz	.Lloop
+
+	mov	`16*4`(%rsp),%rsi
+	mov	(%rsi),%r13
+	mov	8(%rsi),%r12
+	mov	16(%rsi),%rbp
+	mov	24(%rsi),%rbx
+	lea	32(%rsi),%rsp
+.Lepilogue:
+	ret
+.size	sha1_block_data_order,.-sha1_block_data_order
 ___
-&EPILOGUE("sha1_block_data_order");
+{{{
+my $Xi=4;
+my @X=map("%xmm$_",(4..7,0..3));
+my @Tx=map("%xmm$_",(8..10));
+my @V=($A,$B,$C,$D,$E)=("%eax","%ebx","%ecx","%edx","%ebp");	# size optimization
+my @T=("%esi","%edi");
+my $j=0;
+my $K_XX_XX="%r11";
+
+my $_rol=sub { &rol(@_) };
+my $_ror=sub { &ror(@_) };
+
+$code.=<<___;
+.type	sha1_block_data_order_ssse3,\@function,3
+.align	16
+sha1_block_data_order_ssse3:
+_ssse3_shortcut:
+	push	%rbx
+	push	%rbp
+	push	%r12
+	lea	`-64-($win64?5*16:0)`(%rsp),%rsp
+___
+$code.=<<___ if ($win64);
+	movaps	%xmm6,64+0(%rsp)
+	movaps	%xmm7,64+16(%rsp)
+	movaps	%xmm8,64+32(%rsp)
+	movaps	%xmm9,64+48(%rsp)
+	movaps	%xmm10,64+64(%rsp)
+.Lprologue_ssse3:
+___
+$code.=<<___;
+	mov	%rdi,$ctx	# reassigned argument
+	mov	%rsi,$inp	# reassigned argument
+	mov	%rdx,$num	# reassigned argument
+
+	shl	\$6,$num
+	add	$inp,$num
+	lea	K_XX_XX(%rip),$K_XX_XX
+
+	mov	0($ctx),$A		# load context
+	mov	4($ctx),$B
+	mov	8($ctx),$C
+	mov	12($ctx),$D
+	mov	$B,@T[0]		# magic seed
+	mov	16($ctx),$E
+
+	movdqa	64($K_XX_XX),@X[2]	# pbswap mask
+	movdqa	0($K_XX_XX),@Tx[1]	# K_00_19
+	movdqu	0($inp),@X[-4&7]	# load input to %xmm[0-3]
+	movdqu	16($inp),@X[-3&7]
+	movdqu	32($inp),@X[-2&7]
+	movdqu	48($inp),@X[-1&7]
+	pshufb	@X[2],@X[-4&7]		# byte swap
+	add	\$64,$inp
+	pshufb	@X[2],@X[-3&7]
+	pshufb	@X[2],@X[-2&7]
+	pshufb	@X[2],@X[-1&7]
+	paddd	@Tx[1],@X[-4&7]		# add K_00_19
+	paddd	@Tx[1],@X[-3&7]
+	paddd	@Tx[1],@X[-2&7]
+	movdqa	@X[-4&7],0(%rsp)	# X[]+K xfer to IALU
+	psubd	@Tx[1],@X[-4&7]		# restore X[]
+	movdqa	@X[-3&7],16(%rsp)
+	psubd	@Tx[1],@X[-3&7]
+	movdqa	@X[-2&7],32(%rsp)
+	psubd	@Tx[1],@X[-2&7]
+	jmp	.Loop_ssse3
+___
+
+sub AUTOLOAD()		# thunk [simplified] 32-bit style perlasm
+{ my $opcode = $AUTOLOAD; $opcode =~ s/.*:://;
+  my $arg = pop;
+    $arg = "\$$arg" if ($arg*1 eq $arg);
+    $code .= "\t$opcode\t".join(',',$arg,reverse @_)."\n";
+}
+
+sub Xupdate_ssse3_16_31()		# recall that $Xi starts wtih 4
+{ use integer;
+  my $body = shift;
+  my @insns = (&$body,&$body,&$body,&$body);	# 40 instructions
+  my ($a,$b,$c,$d,$e);
+
+	&movdqa	(@X[0],@X[-3&7]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&movdqa	(@Tx[0],@X[-1&7]);
+	&palignr(@X[0],@X[-4&7],8);	# compose "X[-14]" in "X[0]"
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	  &paddd	(@Tx[1],@X[-1&7]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&psrldq	(@Tx[0],4);		# "X[-3]", 3 dwords
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&pxor	(@X[0],@X[-4&7]);	# "X[0]"^="X[-16]"
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&pxor	(@Tx[0],@X[-2&7]);	# "X[-3]"^"X[-8]"
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&pxor	(@X[0],@Tx[0]);		# "X[0]"^="X[-3]"^"X[-8]"
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	  &movdqa	(eval(16*(($Xi-1)&3))."(%rsp)",@Tx[1]);	# X[]+K xfer to IALU
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&movdqa	(@Tx[2],@X[0]);
+	&movdqa	(@Tx[0],@X[0]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&pslldq	(@Tx[2],12);		# "X[0]"<<96, extract one dword
+	&paddd	(@X[0],@X[0]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&psrld	(@Tx[0],31);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&movdqa	(@Tx[1],@Tx[2]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&psrld	(@Tx[2],30);
+	&por	(@X[0],@Tx[0]);		# "X[0]"<<<=1
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&pslld	(@Tx[1],2);
+	&pxor	(@X[0],@Tx[2]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	  &movdqa	(@Tx[2],eval(16*(($Xi)/5))."($K_XX_XX)");	# K_XX_XX
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&pxor	(@X[0],@Tx[1]);		# "X[0]"^=("X[0]">>96)<<<2
+
+	 foreach (@insns) { eval; }	# remaining instructions [if any]
+
+  $Xi++;	push(@X,shift(@X));	# "rotate" X[]
+		push(@Tx,shift(@Tx));
+}
+
+sub Xupdate_ssse3_32_79()
+{ use integer;
+  my $body = shift;
+  my @insns = (&$body,&$body,&$body,&$body);	# 32 to 48 instructions
+  my ($a,$b,$c,$d,$e);
+
+	&movdqa	(@Tx[0],@X[-1&7])	if ($Xi==8);
+	 eval(shift(@insns));		# body_20_39
+	&pxor	(@X[0],@X[-4&7]);	# "X[0]"="X[-32]"^"X[-16]"
+	&palignr(@Tx[0],@X[-2&7],8);	# compose "X[-6]"
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# rol
+
+	&pxor	(@X[0],@X[-7&7]);	# "X[0]"^="X[-28]"
+	 eval(shift(@insns));
+	 eval(shift(@insns))	if (@insns[0] !~ /&ro[rl]/);
+	if ($Xi%5) {
+	  &movdqa	(@Tx[2],@Tx[1]);# "perpetuate" K_XX_XX...
+	} else {			# ... or load next one
+	  &movdqa	(@Tx[2],eval(16*($Xi/5))."($K_XX_XX)");
+	}
+	  &paddd	(@Tx[1],@X[-1&7]);
+	 eval(shift(@insns));		# ror
+	 eval(shift(@insns));
+
+	&pxor	(@X[0],@Tx[0]);		# "X[0]"^="X[-6]"
+	 eval(shift(@insns));		# body_20_39
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# rol
+
+	&movdqa	(@Tx[0],@X[0]);
+	  &movdqa	(eval(16*(($Xi-1)&3))."(%rsp)",@Tx[1]);	# X[]+K xfer to IALU
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# ror
+	 eval(shift(@insns));
+
+	&pslld	(@X[0],2);
+	 eval(shift(@insns));		# body_20_39
+	 eval(shift(@insns));
+	&psrld	(@Tx[0],30);
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# rol
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# ror
+	 eval(shift(@insns));
+
+	&por	(@X[0],@Tx[0]);		# "X[0]"<<<=2
+	 eval(shift(@insns));		# body_20_39
+	 eval(shift(@insns));
+	  &movdqa	(@Tx[1],@X[0])	if ($Xi<19);
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# rol
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# rol
+	 eval(shift(@insns));
+
+	 foreach (@insns) { eval; }	# remaining instructions
+
+  $Xi++;	push(@X,shift(@X));	# "rotate" X[]
+		push(@Tx,shift(@Tx));
+}
+
+sub Xuplast_ssse3_80()
+{ use integer;
+  my $body = shift;
+  my @insns = (&$body,&$body,&$body,&$body);	# 32 instructions
+  my ($a,$b,$c,$d,$e);
+
+	 eval(shift(@insns));
+	  &paddd	(@Tx[1],@X[-1&7]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	  &movdqa	(eval(16*(($Xi-1)&3))."(%rsp)",@Tx[1]);	# X[]+K xfer IALU
+
+	 foreach (@insns) { eval; }		# remaining instructions
+
+	&cmp	($inp,$num);
+	&je	(".Ldone_ssse3");
+
+	unshift(@Tx,pop(@Tx));
+
+	&movdqa	(@X[2],"64($K_XX_XX)");		# pbswap mask
+	&movdqa	(@Tx[1],"0($K_XX_XX)");		# K_00_19
+	&movdqu	(@X[-4&7],"0($inp)");		# load input
+	&movdqu	(@X[-3&7],"16($inp)");
+	&movdqu	(@X[-2&7],"32($inp)");
+	&movdqu	(@X[-1&7],"48($inp)");
+	&pshufb	(@X[-4&7],@X[2]);		# byte swap
+	&add	($inp,64);
+
+  $Xi=0;
+}
+
+sub Xloop_ssse3()
+{ use integer;
+  my $body = shift;
+  my @insns = (&$body,&$body,&$body,&$body);	# 32 instructions
+  my ($a,$b,$c,$d,$e);
+
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&pshufb	(@X[($Xi-3)&7],@X[2]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&paddd	(@X[($Xi-4)&7],@Tx[1]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&movdqa	(eval(16*$Xi)."(%rsp)",@X[($Xi-4)&7]);	# X[]+K xfer to IALU
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&psubd	(@X[($Xi-4)&7],@Tx[1]);
+
+	foreach (@insns) { eval; }
+  $Xi++;
+}
+
+sub Xtail_ssse3()
+{ use integer;
+  my $body = shift;
+  my @insns = (&$body,&$body,&$body,&$body);	# 32 instructions
+  my ($a,$b,$c,$d,$e);
+
+	foreach (@insns) { eval; }
+}
+
+sub body_00_19 () {
+	(
+	'($a,$b,$c,$d,$e)=@V;'.
+	'&add	($e,eval(4*($j&15))."(%rsp)");',	# X[]+K xfer
+	'&xor	($c,$d);',
+	'&mov	(@T[1],$a);',	# $b in next round
+	'&$_rol	($a,5);',
+	'&and	(@T[0],$c);',	# ($b&($c^$d))
+	'&xor	($c,$d);',	# restore $c
+	'&xor	(@T[0],$d);',
+	'&add	($e,$a);',
+	'&$_ror	($b,$j?7:2);',	# $b>>>2
+	'&add	($e,@T[0]);'	.'$j++; unshift(@V,pop(@V)); unshift(@T,pop(@T));'
+	);
+}
+
+sub body_20_39 () {
+	(
+	'($a,$b,$c,$d,$e)=@V;'.
+	'&add	($e,eval(4*($j++&15))."(%rsp)");',	# X[]+K xfer
+	'&xor	(@T[0],$d);',	# ($b^$d)
+	'&mov	(@T[1],$a);',	# $b in next round
+	'&$_rol	($a,5);',
+	'&xor	(@T[0],$c);',	# ($b^$d^$c)
+	'&add	($e,$a);',
+	'&$_ror	($b,7);',	# $b>>>2
+	'&add	($e,@T[0]);'	.'unshift(@V,pop(@V)); unshift(@T,pop(@T));'
+	);
+}
+
+sub body_40_59 () {
+	(
+	'($a,$b,$c,$d,$e)=@V;'.
+	'&mov	(@T[1],$c);',
+	'&xor	($c,$d);',
+	'&add	($e,eval(4*($j++&15))."(%rsp)");',	# X[]+K xfer
+	'&and	(@T[1],$d);',
+	'&and	(@T[0],$c);',	# ($b&($c^$d))
+	'&$_ror	($b,7);',	# $b>>>2
+	'&add	($e,@T[1]);',
+	'&mov	(@T[1],$a);',	# $b in next round
+	'&$_rol	($a,5);',
+	'&add	($e,@T[0]);',
+	'&xor	($c,$d);',	# restore $c
+	'&add	($e,$a);'	.'unshift(@V,pop(@V)); unshift(@T,pop(@T));'
+	);
+}
 $code.=<<___;
-.asciz	"SHA1 block transform for x86_64, CRYPTOGAMS by <appro\@openssl.org>"
 .align	16
+.Loop_ssse3:
+___
+	&Xupdate_ssse3_16_31(\&body_00_19);
+	&Xupdate_ssse3_16_31(\&body_00_19);
+	&Xupdate_ssse3_16_31(\&body_00_19);
+	&Xupdate_ssse3_16_31(\&body_00_19);
+	&Xupdate_ssse3_32_79(\&body_00_19);
+	&Xupdate_ssse3_32_79(\&body_20_39);
+	&Xupdate_ssse3_32_79(\&body_20_39);
+	&Xupdate_ssse3_32_79(\&body_20_39);
+	&Xupdate_ssse3_32_79(\&body_20_39);
+	&Xupdate_ssse3_32_79(\&body_20_39);
+	&Xupdate_ssse3_32_79(\&body_40_59);
+	&Xupdate_ssse3_32_79(\&body_40_59);
+	&Xupdate_ssse3_32_79(\&body_40_59);
+	&Xupdate_ssse3_32_79(\&body_40_59);
+	&Xupdate_ssse3_32_79(\&body_40_59);
+	&Xupdate_ssse3_32_79(\&body_20_39);
+	&Xuplast_ssse3_80(\&body_20_39);	# can jump to "done"
+
+				$saved_j=$j; @saved_V=@V;
+
+	&Xloop_ssse3(\&body_20_39);
+	&Xloop_ssse3(\&body_20_39);
+	&Xloop_ssse3(\&body_20_39);
+
+$code.=<<___;
+	add	0($ctx),$A			# update context
+	add	4($ctx),@T[0]
+	add	8($ctx),$C
+	add	12($ctx),$D
+	mov	$A,0($ctx)
+	add	16($ctx),$E
+	mov	@T[0],4($ctx)
+	mov	@T[0],$B			# magic seed
+	mov	$C,8($ctx)
+	mov	$D,12($ctx)
+	mov	$E,16($ctx)
+	jmp	.Loop_ssse3
+
+.align	16
+.Ldone_ssse3:
+___
+				$j=$saved_j; @V=@saved_V;
+
+	&Xtail_ssse3(\&body_20_39);
+	&Xtail_ssse3(\&body_20_39);
+	&Xtail_ssse3(\&body_20_39);
+
+$code.=<<___;
+	add	0($ctx),$A			# update context
+	add	4($ctx),@T[0]
+	add	8($ctx),$C
+	mov	$A,0($ctx)
+	add	12($ctx),$D
+	mov	@T[0],4($ctx)
+	add	16($ctx),$E
+	mov	$C,8($ctx)
+	mov	$D,12($ctx)
+	mov	$E,16($ctx)
+___
+$code.=<<___ if ($win64);
+	movaps	64+0(%rsp),%xmm6
+	movaps	64+16(%rsp),%xmm7
+	movaps	64+32(%rsp),%xmm8
+	movaps	64+48(%rsp),%xmm9
+	movaps	64+64(%rsp),%xmm10
+___
+$code.=<<___;
+	lea	`64+($win64?5*16:0)`(%rsp),%rsi
+	mov	0(%rsi),%r12
+	mov	8(%rsi),%rbp
+	mov	16(%rsi),%rbx
+	lea	24(%rsi),%rsp
+.Lepilogue_ssse3:
+	ret
+.size	sha1_block_data_order_ssse3,.-sha1_block_data_order_ssse3
+___
+
+if ($avx) {
+my $Xi=4;
+my @X=map("%xmm$_",(4..7,0..3));
+my @Tx=map("%xmm$_",(8..10));
+my @V=($A,$B,$C,$D,$E)=("%eax","%ebx","%ecx","%edx","%ebp");	# size optimization
+my @T=("%esi","%edi");
+my $j=0;
+my $K_XX_XX="%r11";
+
+my $_rol=sub { &shld(@_[0],@_) };
+my $_ror=sub { &shrd(@_[0],@_) };
+
+$code.=<<___;
+.type	sha1_block_data_order_avx,\@function,3
+.align	16
+sha1_block_data_order_avx:
+_avx_shortcut:
+	push	%rbx
+	push	%rbp
+	push	%r12
+	lea	`-64-($win64?5*16:0)`(%rsp),%rsp
+___
+$code.=<<___ if ($win64);
+	movaps	%xmm6,64+0(%rsp)
+	movaps	%xmm7,64+16(%rsp)
+	movaps	%xmm8,64+32(%rsp)
+	movaps	%xmm9,64+48(%rsp)
+	movaps	%xmm10,64+64(%rsp)
+.Lprologue_avx:
+___
+$code.=<<___;
+	mov	%rdi,$ctx	# reassigned argument
+	mov	%rsi,$inp	# reassigned argument
+	mov	%rdx,$num	# reassigned argument
+	vzeroall
+
+	shl	\$6,$num
+	add	$inp,$num
+	lea	K_XX_XX(%rip),$K_XX_XX
+
+	mov	0($ctx),$A		# load context
+	mov	4($ctx),$B
+	mov	8($ctx),$C
+	mov	12($ctx),$D
+	mov	$B,@T[0]		# magic seed
+	mov	16($ctx),$E
+
+	vmovdqa	64($K_XX_XX),@X[2]	# pbswap mask
+	vmovdqa	0($K_XX_XX),@Tx[1]	# K_00_19
+	vmovdqu	0($inp),@X[-4&7]	# load input to %xmm[0-3]
+	vmovdqu	16($inp),@X[-3&7]
+	vmovdqu	32($inp),@X[-2&7]
+	vmovdqu	48($inp),@X[-1&7]
+	vpshufb	@X[2],@X[-4&7],@X[-4&7]	# byte swap
+	add	\$64,$inp
+	vpshufb	@X[2],@X[-3&7],@X[-3&7]
+	vpshufb	@X[2],@X[-2&7],@X[-2&7]
+	vpshufb	@X[2],@X[-1&7],@X[-1&7]
+	vpaddd	@Tx[1],@X[-4&7],@X[0]	# add K_00_19
+	vpaddd	@Tx[1],@X[-3&7],@X[1]
+	vpaddd	@Tx[1],@X[-2&7],@X[2]
+	vmovdqa	@X[0],0(%rsp)		# X[]+K xfer to IALU
+	vmovdqa	@X[1],16(%rsp)
+	vmovdqa	@X[2],32(%rsp)
+	jmp	.Loop_avx
+___
+
+sub Xupdate_avx_16_31()		# recall that $Xi starts wtih 4
+{ use integer;
+  my $body = shift;
+  my @insns = (&$body,&$body,&$body,&$body);	# 40 instructions
+  my ($a,$b,$c,$d,$e);
+
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&vpalignr(@X[0],@X[-3&7],@X[-4&7],8);	# compose "X[-14]" in "X[0]"
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	  &vpaddd	(@Tx[1],@Tx[1],@X[-1&7]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&vpsrldq(@Tx[0],@X[-1&7],4);	# "X[-3]", 3 dwords
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&vpxor	(@X[0],@X[0],@X[-4&7]);		# "X[0]"^="X[-16]"
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&vpxor	(@Tx[0],@Tx[0],@X[-2&7]);	# "X[-3]"^"X[-8]"
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&vpxor	(@X[0],@X[0],@Tx[0]);		# "X[0]"^="X[-3]"^"X[-8]"
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	  &vmovdqa	(eval(16*(($Xi-1)&3))."(%rsp)",@Tx[1]);	# X[]+K xfer to IALU
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&vpsrld	(@Tx[0],@X[0],31);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&vpslldq(@Tx[2],@X[0],12);		# "X[0]"<<96, extract one dword
+	&vpaddd	(@X[0],@X[0],@X[0]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&vpsrld	(@Tx[1],@Tx[2],30);
+	&vpor	(@X[0],@X[0],@Tx[0]);		# "X[0]"<<<=1
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&vpslld	(@Tx[2],@Tx[2],2);
+	&vpxor	(@X[0],@X[0],@Tx[1]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	&vpxor	(@X[0],@X[0],@Tx[2]);		# "X[0]"^=("X[0]">>96)<<<2
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	  &vmovdqa	(@Tx[2],eval(16*(($Xi)/5))."($K_XX_XX)");	# K_XX_XX
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+
+	 foreach (@insns) { eval; }	# remaining instructions [if any]
+
+  $Xi++;	push(@X,shift(@X));	# "rotate" X[]
+		push(@Tx,shift(@Tx));
+}
+
+sub Xupdate_avx_32_79()
+{ use integer;
+  my $body = shift;
+  my @insns = (&$body,&$body,&$body,&$body);	# 32 to 48 instructions
+  my ($a,$b,$c,$d,$e);
+
+	&vpalignr(@Tx[0],@X[-1&7],@X[-2&7],8);	# compose "X[-6]"
+	&vpxor	(@X[0],@X[0],@X[-4&7]);		# "X[0]"="X[-32]"^"X[-16]"
+	 eval(shift(@insns));		# body_20_39
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# rol
+
+	&vpxor	(@X[0],@X[0],@X[-7&7]);		# "X[0]"^="X[-28]"
+	 eval(shift(@insns));
+	 eval(shift(@insns))	if (@insns[0] !~ /&ro[rl]/);
+	if ($Xi%5) {
+	  &vmovdqa	(@Tx[2],@Tx[1]);# "perpetuate" K_XX_XX...
+	} else {			# ... or load next one
+	  &vmovdqa	(@Tx[2],eval(16*($Xi/5))."($K_XX_XX)");
+	}
+	  &vpaddd	(@Tx[1],@Tx[1],@X[-1&7]);
+	 eval(shift(@insns));		# ror
+	 eval(shift(@insns));
+
+	&vpxor	(@X[0],@X[0],@Tx[0]);		# "X[0]"^="X[-6]"
+	 eval(shift(@insns));		# body_20_39
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# rol
+
+	&vpsrld	(@Tx[0],@X[0],30);
+	  &vmovdqa	(eval(16*(($Xi-1)&3))."(%rsp)",@Tx[1]);	# X[]+K xfer to IALU
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# ror
+	 eval(shift(@insns));
+
+	&vpslld	(@X[0],@X[0],2);
+	 eval(shift(@insns));		# body_20_39
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# rol
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# ror
+	 eval(shift(@insns));
+
+	&vpor	(@X[0],@X[0],@Tx[0]);		# "X[0]"<<<=2
+	 eval(shift(@insns));		# body_20_39
+	 eval(shift(@insns));
+	  &vmovdqa	(@Tx[1],@X[0])	if ($Xi<19);
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# rol
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));		# rol
+	 eval(shift(@insns));
+
+	 foreach (@insns) { eval; }	# remaining instructions
+
+  $Xi++;	push(@X,shift(@X));	# "rotate" X[]
+		push(@Tx,shift(@Tx));
+}
+
+sub Xuplast_avx_80()
+{ use integer;
+  my $body = shift;
+  my @insns = (&$body,&$body,&$body,&$body);	# 32 instructions
+  my ($a,$b,$c,$d,$e);
+
+	 eval(shift(@insns));
+	  &vpaddd	(@Tx[1],@Tx[1],@X[-1&7]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	  &movdqa	(eval(16*(($Xi-1)&3))."(%rsp)",@Tx[1]);	# X[]+K xfer IALU
+
+	 foreach (@insns) { eval; }		# remaining instructions
+
+	&cmp	($inp,$num);
+	&je	(".Ldone_avx");
+
+	unshift(@Tx,pop(@Tx));
+
+	&vmovdqa(@X[2],"64($K_XX_XX)");		# pbswap mask
+	&vmovdqa(@Tx[1],"0($K_XX_XX)");		# K_00_19
+	&vmovdqu(@X[-4&7],"0($inp)");		# load input
+	&vmovdqu(@X[-3&7],"16($inp)");
+	&vmovdqu(@X[-2&7],"32($inp)");
+	&vmovdqu(@X[-1&7],"48($inp)");
+	&vpshufb(@X[-4&7],@X[-4&7],@X[2]);	# byte swap
+	&add	($inp,64);
+
+  $Xi=0;
+}
+
+sub Xloop_avx()
+{ use integer;
+  my $body = shift;
+  my @insns = (&$body,&$body,&$body,&$body);	# 32 instructions
+  my ($a,$b,$c,$d,$e);
+
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&vpshufb(@X[($Xi-3)&7],@X[($Xi-3)&7],@X[2]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&vpaddd	(@X[$Xi&7],@X[($Xi-4)&7],@Tx[1]);
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+	&vmovdqa(eval(16*$Xi)."(%rsp)",@X[$Xi&7]);	# X[]+K xfer to IALU
+	 eval(shift(@insns));
+	 eval(shift(@insns));
+
+	foreach (@insns) { eval; }
+  $Xi++;
+}
+
+sub Xtail_avx()
+{ use integer;
+  my $body = shift;
+  my @insns = (&$body,&$body,&$body,&$body);	# 32 instructions
+  my ($a,$b,$c,$d,$e);
+
+	foreach (@insns) { eval; }
+}
+
+$code.=<<___;
+.align	16
+.Loop_avx:
+___
+	&Xupdate_avx_16_31(\&body_00_19);
+	&Xupdate_avx_16_31(\&body_00_19);
+	&Xupdate_avx_16_31(\&body_00_19);
+	&Xupdate_avx_16_31(\&body_00_19);
+	&Xupdate_avx_32_79(\&body_00_19);
+	&Xupdate_avx_32_79(\&body_20_39);
+	&Xupdate_avx_32_79(\&body_20_39);
+	&Xupdate_avx_32_79(\&body_20_39);
+	&Xupdate_avx_32_79(\&body_20_39);
+	&Xupdate_avx_32_79(\&body_20_39);
+	&Xupdate_avx_32_79(\&body_40_59);
+	&Xupdate_avx_32_79(\&body_40_59);
+	&Xupdate_avx_32_79(\&body_40_59);
+	&Xupdate_avx_32_79(\&body_40_59);
+	&Xupdate_avx_32_79(\&body_40_59);
+	&Xupdate_avx_32_79(\&body_20_39);
+	&Xuplast_avx_80(\&body_20_39);	# can jump to "done"
+
+				$saved_j=$j; @saved_V=@V;
+
+	&Xloop_avx(\&body_20_39);
+	&Xloop_avx(\&body_20_39);
+	&Xloop_avx(\&body_20_39);
+
+$code.=<<___;
+	add	0($ctx),$A			# update context
+	add	4($ctx),@T[0]
+	add	8($ctx),$C
+	add	12($ctx),$D
+	mov	$A,0($ctx)
+	add	16($ctx),$E
+	mov	@T[0],4($ctx)
+	mov	@T[0],$B			# magic seed
+	mov	$C,8($ctx)
+	mov	$D,12($ctx)
+	mov	$E,16($ctx)
+	jmp	.Loop_avx
+
+.align	16
+.Ldone_avx:
+___
+				$j=$saved_j; @V=@saved_V;
+
+	&Xtail_avx(\&body_20_39);
+	&Xtail_avx(\&body_20_39);
+	&Xtail_avx(\&body_20_39);
+
+$code.=<<___;
+	vzeroall
+
+	add	0($ctx),$A			# update context
+	add	4($ctx),@T[0]
+	add	8($ctx),$C
+	mov	$A,0($ctx)
+	add	12($ctx),$D
+	mov	@T[0],4($ctx)
+	add	16($ctx),$E
+	mov	$C,8($ctx)
+	mov	$D,12($ctx)
+	mov	$E,16($ctx)
+___
+$code.=<<___ if ($win64);
+	movaps	64+0(%rsp),%xmm6
+	movaps	64+16(%rsp),%xmm7
+	movaps	64+32(%rsp),%xmm8
+	movaps	64+48(%rsp),%xmm9
+	movaps	64+64(%rsp),%xmm10
+___
+$code.=<<___;
+	lea	`64+($win64?5*16:0)`(%rsp),%rsi
+	mov	0(%rsi),%r12
+	mov	8(%rsi),%rbp
+	mov	16(%rsi),%rbx
+	lea	24(%rsi),%rsp
+.Lepilogue_avx:
+	ret
+.size	sha1_block_data_order_avx,.-sha1_block_data_order_avx
+___
+}
+$code.=<<___;
+.align	64
+K_XX_XX:
+.long	0x5a827999,0x5a827999,0x5a827999,0x5a827999	# K_00_19
+.long	0x6ed9eba1,0x6ed9eba1,0x6ed9eba1,0x6ed9eba1	# K_20_39
+.long	0x8f1bbcdc,0x8f1bbcdc,0x8f1bbcdc,0x8f1bbcdc	# K_40_59
+.long	0xca62c1d6,0xca62c1d6,0xca62c1d6,0xca62c1d6	# K_60_79
+.long	0x00010203,0x04050607,0x08090a0b,0x0c0d0e0f	# pbswap mask
+___
+}}}
+$code.=<<___;
+.asciz	"SHA1 block transform for x86_64, CRYPTOGAMS by <appro\@openssl.org>"
+.align	64
 ___
 
 # EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame,
@@ -272,25 +1112,75 @@ se_handler:
 
 	lea	.Lprologue(%rip),%r10
 	cmp	%r10,%rbx		# context->Rip<.Lprologue
-	jb	.Lin_prologue
+	jb	.Lcommon_seh_tail
 
 	mov	152($context),%rax	# pull context->Rsp
 
 	lea	.Lepilogue(%rip),%r10
 	cmp	%r10,%rbx		# context->Rip>=.Lepilogue
-	jae	.Lin_prologue
+	jae	.Lcommon_seh_tail
 
 	mov	`16*4`(%rax),%rax	# pull saved stack pointer
-	lea	24(%rax),%rax
+	lea	32(%rax),%rax
 
 	mov	-8(%rax),%rbx
 	mov	-16(%rax),%rbp
 	mov	-24(%rax),%r12
+	mov	-32(%rax),%r13
 	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
+
+	jmp	.Lcommon_seh_tail
+.size	se_handler,.-se_handler
 
-.Lin_prologue:
+.type	ssse3_handler,\@abi-omnipotent
+.align	16
+ssse3_handler:
+	push	%rsi
+	push	%rdi
+	push	%rbx
+	push	%rbp
+	push	%r12
+	push	%r13
+	push	%r14
+	push	%r15
+	pushfq
+	sub	\$64,%rsp
+
+	mov	120($context),%rax	# pull context->Rax
+	mov	248($context),%rbx	# pull context->Rip
+
+	mov	8($disp),%rsi		# disp->ImageBase
+	mov	56($disp),%r11		# disp->HandlerData
+
+	mov	0(%r11),%r10d		# HandlerData[0]
+	lea	(%rsi,%r10),%r10	# prologue label
+	cmp	%r10,%rbx		# context->Rip<prologue label
+	jb	.Lcommon_seh_tail
+
+	mov	152($context),%rax	# pull context->Rsp
+
+	mov	4(%r11),%r10d		# HandlerData[1]
+	lea	(%rsi,%r10),%r10	# epilogue label
+	cmp	%r10,%rbx		# context->Rip>=epilogue label
+	jae	.Lcommon_seh_tail
+
+	lea	64(%rax),%rsi
+	lea	512($context),%rdi	# &context.Xmm6
+	mov	\$10,%ecx
+	.long	0xa548f3fc		# cld; rep movsq
+	lea	`24+64+5*16`(%rax),%rax	# adjust stack pointer
+
+	mov	-8(%rax),%rbx
+	mov	-16(%rax),%rbp
+	mov	-24(%rax),%r12
+	mov	%rbx,144($context)	# restore context->Rbx
+	mov	%rbp,160($context)	# restore context->Rbp
+	mov	%r12,216($context)	# restore cotnext->R12
+
+.Lcommon_seh_tail:
 	mov	8(%rax),%rdi
 	mov	16(%rax),%rsi
 	mov	%rax,152($context)	# restore context->Rsp
@@ -328,19 +1218,38 @@ se_handler:
 	pop	%rdi
 	pop	%rsi
 	ret
-.size	se_handler,.-se_handler
+.size	ssse3_handler,.-ssse3_handler
 
 .section	.pdata
 .align	4
 	.rva	.LSEH_begin_sha1_block_data_order
 	.rva	.LSEH_end_sha1_block_data_order
 	.rva	.LSEH_info_sha1_block_data_order
-
+	.rva	.LSEH_begin_sha1_block_data_order_ssse3
+	.rva	.LSEH_end_sha1_block_data_order_ssse3
+	.rva	.LSEH_info_sha1_block_data_order_ssse3
+___
+$code.=<<___ if ($avx);
+	.rva	.LSEH_begin_sha1_block_data_order_avx
+	.rva	.LSEH_end_sha1_block_data_order_avx
+	.rva	.LSEH_info_sha1_block_data_order_avx
+___
+$code.=<<___;
 .section	.xdata
 .align	8
 .LSEH_info_sha1_block_data_order:
 	.byte	9,0,0,0
 	.rva	se_handler
+.LSEH_info_sha1_block_data_order_ssse3:
+	.byte	9,0,0,0
+	.rva	ssse3_handler
+	.rva	.Lprologue_ssse3,.Lepilogue_ssse3	# HandlerData[]
+___
+$code.=<<___ if ($avx);
+.LSEH_info_sha1_block_data_order_avx:
+	.byte	9,0,0,0
+	.rva	ssse3_handler
+	.rva	.Lprologue_avx,.Lepilogue_avx		# HandlerData[]
 ___
 }
 
diff --git a/openssl/crypto/sha/asm/sha256-586.pl b/openssl/crypto/sha/asm/sha256-586.pl
index ecc8b69c7..928ec5312 100644
--- a/openssl/crypto/sha/asm/sha256-586.pl
+++ b/openssl/crypto/sha/asm/sha256-586.pl
@@ -14,8 +14,8 @@
 #		Pentium	PIII	P4	AMD K8	Core2
 # gcc		46	36	41	27	26
 # icc		57	33	38	25	23	
-# x86 asm	40	30	35	20	20
-# x86_64 asm(*)	-	-	21	15.8	16.5
+# x86 asm	40	30	33	20	18
+# x86_64 asm(*)	-	-	21	16	16
 #
 # (*) x86_64 assembler performance is presented for reference
 #     purposes.
@@ -48,20 +48,19 @@ sub BODY_00_15() {
     my $in_16_63=shift;
 
 	&mov	("ecx",$E);
-	 &add	($T,&DWP(4*(8+15+16-9),"esp"))	if ($in_16_63);	# T += X[-7]
-	&ror	("ecx",6);
-	&mov	("edi",$E);
-	&ror	("edi",11);
+	 &add	($T,"edi")			if ($in_16_63);	# T += sigma1(X[-2])
+	&ror	("ecx",25-11);
 	 &mov	("esi",$Foff);
-	&xor	("ecx","edi");
-	&ror	("edi",25-11);
+	&xor	("ecx",$E);
+	&ror	("ecx",11-6);
 	 &mov	(&DWP(4*(8+15),"esp"),$T)	if ($in_16_63);	# save X[0]
-	&xor	("ecx","edi");	# Sigma1(e)
+	&xor	("ecx",$E);
+	&ror	("ecx",6);	# Sigma1(e)
 	 &mov	("edi",$Goff);
 	&add	($T,"ecx");	# T += Sigma1(e)
-	 &mov	($Eoff,$E);	# modulo-scheduled
 
 	&xor	("esi","edi");
+	 &mov	($Eoff,$E);	# modulo-scheduled
 	 &mov	("ecx",$A);
 	&and	("esi",$E);
 	 &mov	($E,$Doff);	# e becomes d, which is e in next iteration
@@ -69,14 +68,14 @@ sub BODY_00_15() {
 	 &mov	("edi",$A);
 	&add	($T,"esi");	# T += Ch(e,f,g)
 
-	&ror	("ecx",2);
+	&ror	("ecx",22-13);
 	 &add	($T,$Hoff);	# T += h
-	&ror	("edi",13);
+	&xor	("ecx",$A);
+	&ror	("ecx",13-2);
 	 &mov	("esi",$Boff);
-	&xor	("ecx","edi");
-	&ror	("edi",22-13);
+	&xor	("ecx",$A);
+	&ror	("ecx",2);	# Sigma0(a)
 	 &add	($E,$T);	# d += T
-	&xor	("ecx","edi");	# Sigma0(a)
 	 &mov	("edi",$Coff);
 
 	&add	($T,"ecx");	# T += Sigma0(a)
@@ -168,23 +167,22 @@ sub BODY_00_15() {
 &set_label("16_63",16);
 	&mov	("esi",$T);
 	 &mov	("ecx",&DWP(4*(8+15+16-14),"esp"));
-	&shr	($T,3);
-	&ror	("esi",7);
-	&xor	($T,"esi");
 	&ror	("esi",18-7);
 	 &mov	("edi","ecx");
-	&xor	($T,"esi");			# T = sigma0(X[-15])
+	&xor	("esi",$T);
+	&ror	("esi",7);
+	&shr	($T,3);
 
-	&shr	("ecx",10);
-	 &mov	("esi",&DWP(4*(8+15+16),"esp"));
-	&ror	("edi",17);
-	&xor	("ecx","edi");
 	&ror	("edi",19-17);
-	 &add	($T,"esi");			# T += X[-16]
-	&xor	("edi","ecx")			# sigma1(X[-2])
+	 &xor	($T,"esi");			# T = sigma0(X[-15])
+	&xor	("edi","ecx");
+	&ror	("edi",17);
+	&shr	("ecx",10);
+	 &add	($T,&DWP(4*(8+15+16),"esp"));	# T += X[-16]
+	&xor	("edi","ecx");			# sigma1(X[-2])
 
-	&add	($T,"edi");			# T += sigma1(X[-2])
-	# &add	($T,&DWP(4*(8+15+16-9),"esp"));	# T += X[-7], moved to BODY_00_15(1)
+	 &add	($T,&DWP(4*(8+15+16-9),"esp"));	# T += X[-7]
+	# &add	($T,"edi");			# T += sigma1(X[-2])
 	# &mov	(&DWP(4*(8+15),"esp"),$T);	# save X[0]
 
 	&BODY_00_15(1);
diff --git a/openssl/crypto/sha/asm/sha256-armv4.pl b/openssl/crypto/sha/asm/sha256-armv4.pl
index 492cb62bc..9c84e8d93 100644
--- a/openssl/crypto/sha/asm/sha256-armv4.pl
+++ b/openssl/crypto/sha/asm/sha256-armv4.pl
@@ -18,11 +18,16 @@
 # Rescheduling for dual-issue pipeline resulted in 22% improvement on
 # Cortex A8 core and ~20 cycles per processed byte.
 
+# February 2011.
+#
+# Profiler-assisted and platform-specific optimization resulted in 16%
+# improvement on Cortex A8 core and ~17 cycles per processed byte.
+
 while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {}
 open STDOUT,">$output";
 
 $ctx="r0";	$t0="r0";
-$inp="r1";
+$inp="r1";	$t3="r1";
 $len="r2";	$t1="r2";
 $T1="r3";
 $A="r4";
@@ -46,6 +51,9 @@ sub BODY_00_15 {
 my ($i,$a,$b,$c,$d,$e,$f,$g,$h) = @_;
 
 $code.=<<___ if ($i<16);
+#if __ARM_ARCH__>=7
+	ldr	$T1,[$inp],#4
+#else
 	ldrb	$T1,[$inp,#3]			@ $i
 	ldrb	$t2,[$inp,#2]
 	ldrb	$t1,[$inp,#1]
@@ -53,16 +61,24 @@ $code.=<<___ if ($i<16);
 	orr	$T1,$T1,$t2,lsl#8
 	orr	$T1,$T1,$t1,lsl#16
 	orr	$T1,$T1,$t0,lsl#24
-	`"str	$inp,[sp,#17*4]"	if ($i==15)`
+#endif
 ___
 $code.=<<___;
-	ldr	$t2,[$Ktbl],#4			@ *K256++
 	mov	$t0,$e,ror#$Sigma1[0]
-	str	$T1,[sp,#`$i%16`*4]
+	ldr	$t2,[$Ktbl],#4			@ *K256++
 	eor	$t0,$t0,$e,ror#$Sigma1[1]
 	eor	$t1,$f,$g
+#if $i>=16
+	add	$T1,$T1,$t3			@ from BODY_16_xx
+#elif __ARM_ARCH__>=7 && defined(__ARMEL__)
+	rev	$T1,$T1
+#endif
+#if $i==15
+	str	$inp,[sp,#17*4]			@ leave room for $t3
+#endif
 	eor	$t0,$t0,$e,ror#$Sigma1[2]	@ Sigma1(e)
 	and	$t1,$t1,$e
+	str	$T1,[sp,#`$i%16`*4]
 	add	$T1,$T1,$t0
 	eor	$t1,$t1,$g			@ Ch(e,f,g)
 	add	$T1,$T1,$h
@@ -71,6 +87,9 @@ $code.=<<___;
 	eor	$h,$h,$a,ror#$Sigma0[1]
 	add	$T1,$T1,$t2
 	eor	$h,$h,$a,ror#$Sigma0[2]		@ Sigma0(a)
+#if $i>=15
+	ldr	$t3,[sp,#`($i+2)%16`*4]		@ from BODY_16_xx
+#endif
 	orr	$t0,$a,$b
 	and	$t1,$a,$b
 	and	$t0,$t0,$c
@@ -85,24 +104,26 @@ sub BODY_16_XX {
 my ($i,$a,$b,$c,$d,$e,$f,$g,$h) = @_;
 
 $code.=<<___;
-	ldr	$t1,[sp,#`($i+1)%16`*4]		@ $i
+	@ ldr	$t3,[sp,#`($i+1)%16`*4]		@ $i
 	ldr	$t2,[sp,#`($i+14)%16`*4]
+	mov	$t0,$t3,ror#$sigma0[0]
 	ldr	$T1,[sp,#`($i+0)%16`*4]
-	mov	$t0,$t1,ror#$sigma0[0]
-	ldr	$inp,[sp,#`($i+9)%16`*4]
-	eor	$t0,$t0,$t1,ror#$sigma0[1]
-	eor	$t0,$t0,$t1,lsr#$sigma0[2]	@ sigma0(X[i+1])
-	mov	$t1,$t2,ror#$sigma1[0]
+	eor	$t0,$t0,$t3,ror#$sigma0[1]
+	ldr	$t1,[sp,#`($i+9)%16`*4]
+	eor	$t0,$t0,$t3,lsr#$sigma0[2]	@ sigma0(X[i+1])
+	mov	$t3,$t2,ror#$sigma1[0]
 	add	$T1,$T1,$t0
-	eor	$t1,$t1,$t2,ror#$sigma1[1]
-	add	$T1,$T1,$inp
-	eor	$t1,$t1,$t2,lsr#$sigma1[2]	@ sigma1(X[i+14])
+	eor	$t3,$t3,$t2,ror#$sigma1[1]
 	add	$T1,$T1,$t1
+	eor	$t3,$t3,$t2,lsr#$sigma1[2]	@ sigma1(X[i+14])
+	@ add	$T1,$T1,$t3
 ___
 	&BODY_00_15(@_);
 }
 
 $code=<<___;
+#include "arm_arch.h"
+
 .text
 .code	32
 
@@ -132,7 +153,7 @@ K256:
 sha256_block_data_order:
 	sub	r3,pc,#8		@ sha256_block_data_order
 	add	$len,$inp,$len,lsl#6	@ len to point at the end of inp
-	stmdb	sp!,{$ctx,$inp,$len,r4-r12,lr}
+	stmdb	sp!,{$ctx,$inp,$len,r4-r11,lr}
 	ldmia	$ctx,{$A,$B,$C,$D,$E,$F,$G,$H}
 	sub	$Ktbl,r3,#256		@ K256
 	sub	sp,sp,#16*4		@ alloca(X[16])
@@ -171,10 +192,14 @@ $code.=<<___;
 	bne	.Loop
 
 	add	sp,sp,#`16+3`*4	@ destroy frame
-	ldmia	sp!,{r4-r12,lr}
+#if __ARM_ARCH__>=5
+	ldmia	sp!,{r4-r11,pc}
+#else
+	ldmia	sp!,{r4-r11,lr}
 	tst	lr,#1
 	moveq	pc,lr			@ be binary compatible with V4, yet
 	bx	lr			@ interoperable with Thumb ISA:-)
+#endif
 .size   sha256_block_data_order,.-sha256_block_data_order
 .asciz  "SHA256 block transform for ARMv4, CRYPTOGAMS by <appro\@openssl.org>"
 .align	2
diff --git a/openssl/crypto/sha/asm/sha512-armv4.pl b/openssl/crypto/sha/asm/sha512-armv4.pl
index 3a35861ac..7faf37b14 100644
--- a/openssl/crypto/sha/asm/sha512-armv4.pl
+++ b/openssl/crypto/sha/asm/sha512-armv4.pl
@@ -18,22 +18,33 @@
 # Rescheduling for dual-issue pipeline resulted in 6% improvement on
 # Cortex A8 core and ~40 cycles per processed byte.
 
+# February 2011.
+#
+# Profiler-assisted and platform-specific optimization resulted in 7%
+# improvement on Coxtex A8 core and ~38 cycles per byte.
+
+# March 2011.
+#
+# Add NEON implementation. On Cortex A8 it was measured to process
+# one byte in 25.5 cycles or 47% faster than integer-only code.
+
 # Byte order [in]dependence. =========================================
 #
-# Caller is expected to maintain specific *dword* order in h[0-7],
-# namely with most significant dword at *lower* address, which is
-# reflected in below two parameters. *Byte* order within these dwords
-# in turn is whatever *native* byte order on current platform.
-$hi=0;
-$lo=4;
+# Originally caller was expected to maintain specific *dword* order in
+# h[0-7], namely with most significant dword at *lower* address, which
+# was reflected in below two parameters as 0 and 4. Now caller is
+# expected to maintain native byte order for whole 64-bit values.
+$hi="HI";
+$lo="LO";
 # ====================================================================
 
 while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {}
 open STDOUT,">$output";
 
-$ctx="r0";
+$ctx="r0";	# parameter block
 $inp="r1";
 $len="r2";
+
 $Tlo="r3";
 $Thi="r4";
 $Alo="r5";
@@ -61,15 +72,17 @@ $Xoff=8*8;
 sub BODY_00_15() {
 my $magic = shift;
 $code.=<<___;
-	ldr	$t2,[sp,#$Hoff+0]	@ h.lo
-	ldr	$t3,[sp,#$Hoff+4]	@ h.hi
 	@ Sigma1(x)	(ROTR((x),14) ^ ROTR((x),18)  ^ ROTR((x),41))
 	@ LO		lo>>14^hi<<18 ^ lo>>18^hi<<14 ^ hi>>9^lo<<23
 	@ HI		hi>>14^lo<<18 ^ hi>>18^lo<<14 ^ lo>>9^hi<<23
 	mov	$t0,$Elo,lsr#14
+	str	$Tlo,[sp,#$Xoff+0]
 	mov	$t1,$Ehi,lsr#14
+	str	$Thi,[sp,#$Xoff+4]
 	eor	$t0,$t0,$Ehi,lsl#18
+	ldr	$t2,[sp,#$Hoff+0]	@ h.lo
 	eor	$t1,$t1,$Elo,lsl#18
+	ldr	$t3,[sp,#$Hoff+4]	@ h.hi
 	eor	$t0,$t0,$Elo,lsr#18
 	eor	$t1,$t1,$Ehi,lsr#18
 	eor	$t0,$t0,$Ehi,lsl#14
@@ -96,25 +109,24 @@ $code.=<<___;
 	and	$t1,$t1,$Ehi
 	str	$Ahi,[sp,#$Aoff+4]
 	eor	$t0,$t0,$t2
-	ldr	$t2,[$Ktbl,#4]		@ K[i].lo
+	ldr	$t2,[$Ktbl,#$lo]	@ K[i].lo
 	eor	$t1,$t1,$t3		@ Ch(e,f,g)
-	ldr	$t3,[$Ktbl,#0]		@ K[i].hi
+	ldr	$t3,[$Ktbl,#$hi]	@ K[i].hi
 
 	adds	$Tlo,$Tlo,$t0
 	ldr	$Elo,[sp,#$Doff+0]	@ d.lo
 	adc	$Thi,$Thi,$t1		@ T += Ch(e,f,g)
 	ldr	$Ehi,[sp,#$Doff+4]	@ d.hi
 	adds	$Tlo,$Tlo,$t2
+	and	$t0,$t2,#0xff
 	adc	$Thi,$Thi,$t3		@ T += K[i]
 	adds	$Elo,$Elo,$Tlo
+	ldr	$t2,[sp,#$Boff+0]	@ b.lo
 	adc	$Ehi,$Ehi,$Thi		@ d += T
-
-	and	$t0,$t2,#0xff
 	teq	$t0,#$magic
-	orreq	$Ktbl,$Ktbl,#1
 
-	ldr	$t2,[sp,#$Boff+0]	@ b.lo
 	ldr	$t3,[sp,#$Coff+0]	@ c.lo
+	orreq	$Ktbl,$Ktbl,#1
 	@ Sigma0(x)	(ROTR((x),28) ^ ROTR((x),34) ^ ROTR((x),39))
 	@ LO		lo>>28^hi<<4  ^ hi>>2^lo<<30 ^ hi>>7^lo<<25
 	@ HI		hi>>28^lo<<4  ^ lo>>2^hi<<30 ^ lo>>7^hi<<25
@@ -131,80 +143,100 @@ $code.=<<___;
 	eor	$t0,$t0,$Alo,lsl#25
 	eor	$t1,$t1,$Ahi,lsl#25	@ Sigma0(a)
 	adds	$Tlo,$Tlo,$t0
+	and	$t0,$Alo,$t2
 	adc	$Thi,$Thi,$t1		@ T += Sigma0(a)
 
-	and	$t0,$Alo,$t2
-	orr	$Alo,$Alo,$t2
 	ldr	$t1,[sp,#$Boff+4]	@ b.hi
+	orr	$Alo,$Alo,$t2
 	ldr	$t2,[sp,#$Coff+4]	@ c.hi
 	and	$Alo,$Alo,$t3
-	orr	$Alo,$Alo,$t0		@ Maj(a,b,c).lo
 	and	$t3,$Ahi,$t1
 	orr	$Ahi,$Ahi,$t1
+	orr	$Alo,$Alo,$t0		@ Maj(a,b,c).lo
 	and	$Ahi,$Ahi,$t2
-	orr	$Ahi,$Ahi,$t3		@ Maj(a,b,c).hi
 	adds	$Alo,$Alo,$Tlo
-	adc	$Ahi,$Ahi,$Thi		@ h += T
-
+	orr	$Ahi,$Ahi,$t3		@ Maj(a,b,c).hi
 	sub	sp,sp,#8
+	adc	$Ahi,$Ahi,$Thi		@ h += T
+	tst	$Ktbl,#1
 	add	$Ktbl,$Ktbl,#8
 ___
 }
 $code=<<___;
+#include "arm_arch.h"
+#ifdef __ARMEL__
+# define LO 0
+# define HI 4
+# define WORD64(hi0,lo0,hi1,lo1)	.word	lo0,hi0, lo1,hi1
+#else
+# define HI 0
+# define LO 4
+# define WORD64(hi0,lo0,hi1,lo1)	.word	hi0,lo0, hi1,lo1
+#endif
+
 .text
 .code	32
 .type	K512,%object
 .align	5
 K512:
-.word	0x428a2f98,0xd728ae22, 0x71374491,0x23ef65cd
-.word	0xb5c0fbcf,0xec4d3b2f, 0xe9b5dba5,0x8189dbbc
-.word	0x3956c25b,0xf348b538, 0x59f111f1,0xb605d019
-.word	0x923f82a4,0xaf194f9b, 0xab1c5ed5,0xda6d8118
-.word	0xd807aa98,0xa3030242, 0x12835b01,0x45706fbe
-.word	0x243185be,0x4ee4b28c, 0x550c7dc3,0xd5ffb4e2
-.word	0x72be5d74,0xf27b896f, 0x80deb1fe,0x3b1696b1
-.word	0x9bdc06a7,0x25c71235, 0xc19bf174,0xcf692694
-.word	0xe49b69c1,0x9ef14ad2, 0xefbe4786,0x384f25e3
-.word	0x0fc19dc6,0x8b8cd5b5, 0x240ca1cc,0x77ac9c65
-.word	0x2de92c6f,0x592b0275, 0x4a7484aa,0x6ea6e483
-.word	0x5cb0a9dc,0xbd41fbd4, 0x76f988da,0x831153b5
-.word	0x983e5152,0xee66dfab, 0xa831c66d,0x2db43210
-.word	0xb00327c8,0x98fb213f, 0xbf597fc7,0xbeef0ee4
-.word	0xc6e00bf3,0x3da88fc2, 0xd5a79147,0x930aa725
-.word	0x06ca6351,0xe003826f, 0x14292967,0x0a0e6e70
-.word	0x27b70a85,0x46d22ffc, 0x2e1b2138,0x5c26c926
-.word	0x4d2c6dfc,0x5ac42aed, 0x53380d13,0x9d95b3df
-.word	0x650a7354,0x8baf63de, 0x766a0abb,0x3c77b2a8
-.word	0x81c2c92e,0x47edaee6, 0x92722c85,0x1482353b
-.word	0xa2bfe8a1,0x4cf10364, 0xa81a664b,0xbc423001
-.word	0xc24b8b70,0xd0f89791, 0xc76c51a3,0x0654be30
-.word	0xd192e819,0xd6ef5218, 0xd6990624,0x5565a910
-.word	0xf40e3585,0x5771202a, 0x106aa070,0x32bbd1b8
-.word	0x19a4c116,0xb8d2d0c8, 0x1e376c08,0x5141ab53
-.word	0x2748774c,0xdf8eeb99, 0x34b0bcb5,0xe19b48a8
-.word	0x391c0cb3,0xc5c95a63, 0x4ed8aa4a,0xe3418acb
-.word	0x5b9cca4f,0x7763e373, 0x682e6ff3,0xd6b2b8a3
-.word	0x748f82ee,0x5defb2fc, 0x78a5636f,0x43172f60
-.word	0x84c87814,0xa1f0ab72, 0x8cc70208,0x1a6439ec
-.word	0x90befffa,0x23631e28, 0xa4506ceb,0xde82bde9
-.word	0xbef9a3f7,0xb2c67915, 0xc67178f2,0xe372532b
-.word	0xca273ece,0xea26619c, 0xd186b8c7,0x21c0c207
-.word	0xeada7dd6,0xcde0eb1e, 0xf57d4f7f,0xee6ed178
-.word	0x06f067aa,0x72176fba, 0x0a637dc5,0xa2c898a6
-.word	0x113f9804,0xbef90dae, 0x1b710b35,0x131c471b
-.word	0x28db77f5,0x23047d84, 0x32caab7b,0x40c72493
-.word	0x3c9ebe0a,0x15c9bebc, 0x431d67c4,0x9c100d4c
-.word	0x4cc5d4be,0xcb3e42b6, 0x597f299c,0xfc657e2a
-.word	0x5fcb6fab,0x3ad6faec, 0x6c44198c,0x4a475817
+WORD64(0x428a2f98,0xd728ae22, 0x71374491,0x23ef65cd)
+WORD64(0xb5c0fbcf,0xec4d3b2f, 0xe9b5dba5,0x8189dbbc)
+WORD64(0x3956c25b,0xf348b538, 0x59f111f1,0xb605d019)
+WORD64(0x923f82a4,0xaf194f9b, 0xab1c5ed5,0xda6d8118)
+WORD64(0xd807aa98,0xa3030242, 0x12835b01,0x45706fbe)
+WORD64(0x243185be,0x4ee4b28c, 0x550c7dc3,0xd5ffb4e2)
+WORD64(0x72be5d74,0xf27b896f, 0x80deb1fe,0x3b1696b1)
+WORD64(0x9bdc06a7,0x25c71235, 0xc19bf174,0xcf692694)
+WORD64(0xe49b69c1,0x9ef14ad2, 0xefbe4786,0x384f25e3)
+WORD64(0x0fc19dc6,0x8b8cd5b5, 0x240ca1cc,0x77ac9c65)
+WORD64(0x2de92c6f,0x592b0275, 0x4a7484aa,0x6ea6e483)
+WORD64(0x5cb0a9dc,0xbd41fbd4, 0x76f988da,0x831153b5)
+WORD64(0x983e5152,0xee66dfab, 0xa831c66d,0x2db43210)
+WORD64(0xb00327c8,0x98fb213f, 0xbf597fc7,0xbeef0ee4)
+WORD64(0xc6e00bf3,0x3da88fc2, 0xd5a79147,0x930aa725)
+WORD64(0x06ca6351,0xe003826f, 0x14292967,0x0a0e6e70)
+WORD64(0x27b70a85,0x46d22ffc, 0x2e1b2138,0x5c26c926)
+WORD64(0x4d2c6dfc,0x5ac42aed, 0x53380d13,0x9d95b3df)
+WORD64(0x650a7354,0x8baf63de, 0x766a0abb,0x3c77b2a8)
+WORD64(0x81c2c92e,0x47edaee6, 0x92722c85,0x1482353b)
+WORD64(0xa2bfe8a1,0x4cf10364, 0xa81a664b,0xbc423001)
+WORD64(0xc24b8b70,0xd0f89791, 0xc76c51a3,0x0654be30)
+WORD64(0xd192e819,0xd6ef5218, 0xd6990624,0x5565a910)
+WORD64(0xf40e3585,0x5771202a, 0x106aa070,0x32bbd1b8)
+WORD64(0x19a4c116,0xb8d2d0c8, 0x1e376c08,0x5141ab53)
+WORD64(0x2748774c,0xdf8eeb99, 0x34b0bcb5,0xe19b48a8)
+WORD64(0x391c0cb3,0xc5c95a63, 0x4ed8aa4a,0xe3418acb)
+WORD64(0x5b9cca4f,0x7763e373, 0x682e6ff3,0xd6b2b8a3)
+WORD64(0x748f82ee,0x5defb2fc, 0x78a5636f,0x43172f60)
+WORD64(0x84c87814,0xa1f0ab72, 0x8cc70208,0x1a6439ec)
+WORD64(0x90befffa,0x23631e28, 0xa4506ceb,0xde82bde9)
+WORD64(0xbef9a3f7,0xb2c67915, 0xc67178f2,0xe372532b)
+WORD64(0xca273ece,0xea26619c, 0xd186b8c7,0x21c0c207)
+WORD64(0xeada7dd6,0xcde0eb1e, 0xf57d4f7f,0xee6ed178)
+WORD64(0x06f067aa,0x72176fba, 0x0a637dc5,0xa2c898a6)
+WORD64(0x113f9804,0xbef90dae, 0x1b710b35,0x131c471b)
+WORD64(0x28db77f5,0x23047d84, 0x32caab7b,0x40c72493)
+WORD64(0x3c9ebe0a,0x15c9bebc, 0x431d67c4,0x9c100d4c)
+WORD64(0x4cc5d4be,0xcb3e42b6, 0x597f299c,0xfc657e2a)
+WORD64(0x5fcb6fab,0x3ad6faec, 0x6c44198c,0x4a475817)
 .size	K512,.-K512
+.LOPENSSL_armcap:
+.word	OPENSSL_armcap_P-sha512_block_data_order
+.skip	32-4
 
 .global	sha512_block_data_order
 .type	sha512_block_data_order,%function
 sha512_block_data_order:
 	sub	r3,pc,#8		@ sha512_block_data_order
 	add	$len,$inp,$len,lsl#7	@ len to point at the end of inp
+#if __ARM_ARCH__>=7
+	ldr	r12,.LOPENSSL_armcap
+	ldr	r12,[r3,r12]		@ OPENSSL_armcap_P
+	tst	r12,#1
+	bne	.LNEON
+#endif
 	stmdb	sp!,{r4-r12,lr}
-	sub	$Ktbl,r3,#640		@ K512
+	sub	$Ktbl,r3,#672		@ K512
 	sub	sp,sp,#9*8
 
 	ldr	$Elo,[$ctx,#$Eoff+$lo]
@@ -238,6 +270,7 @@ sha512_block_data_order:
 	str	$Thi,[sp,#$Foff+4]
 
 .L00_15:
+#if __ARM_ARCH__<7
 	ldrb	$Tlo,[$inp,#7]
 	ldrb	$t0, [$inp,#6]
 	ldrb	$t1, [$inp,#5]
@@ -252,26 +285,30 @@ sha512_block_data_order:
 	orr	$Thi,$Thi,$t3,lsl#8
 	orr	$Thi,$Thi,$t0,lsl#16
 	orr	$Thi,$Thi,$t1,lsl#24
-	str	$Tlo,[sp,#$Xoff+0]
-	str	$Thi,[sp,#$Xoff+4]
+#else
+	ldr	$Tlo,[$inp,#4]
+	ldr	$Thi,[$inp],#8
+#ifdef __ARMEL__
+	rev	$Tlo,$Tlo
+	rev	$Thi,$Thi
+#endif
+#endif
 ___
 	&BODY_00_15(0x94);
 $code.=<<___;
 	tst	$Ktbl,#1
 	beq	.L00_15
-	bic	$Ktbl,$Ktbl,#1
-
-.L16_79:
 	ldr	$t0,[sp,#`$Xoff+8*(16-1)`+0]
 	ldr	$t1,[sp,#`$Xoff+8*(16-1)`+4]
-	ldr	$t2,[sp,#`$Xoff+8*(16-14)`+0]
-	ldr	$t3,[sp,#`$Xoff+8*(16-14)`+4]
-
+	bic	$Ktbl,$Ktbl,#1
+.L16_79:
 	@ sigma0(x)	(ROTR((x),1)  ^ ROTR((x),8)  ^ ((x)>>7))
 	@ LO		lo>>1^hi<<31  ^ lo>>8^hi<<24 ^ lo>>7^hi<<25
 	@ HI		hi>>1^lo<<31  ^ hi>>8^lo<<24 ^ hi>>7
 	mov	$Tlo,$t0,lsr#1
+	ldr	$t2,[sp,#`$Xoff+8*(16-14)`+0]
 	mov	$Thi,$t1,lsr#1
+	ldr	$t3,[sp,#`$Xoff+8*(16-14)`+4]
 	eor	$Tlo,$Tlo,$t1,lsl#31
 	eor	$Thi,$Thi,$t0,lsl#31
 	eor	$Tlo,$Tlo,$t0,lsr#8
@@ -295,25 +332,24 @@ $code.=<<___;
 	eor	$t1,$t1,$t3,lsl#3
 	eor	$t0,$t0,$t2,lsr#6
 	eor	$t1,$t1,$t3,lsr#6
+	ldr	$t2,[sp,#`$Xoff+8*(16-9)`+0]
 	eor	$t0,$t0,$t3,lsl#26
 
-	ldr	$t2,[sp,#`$Xoff+8*(16-9)`+0]
 	ldr	$t3,[sp,#`$Xoff+8*(16-9)`+4]
 	adds	$Tlo,$Tlo,$t0
+	ldr	$t0,[sp,#`$Xoff+8*16`+0]
 	adc	$Thi,$Thi,$t1
 
-	ldr	$t0,[sp,#`$Xoff+8*16`+0]
 	ldr	$t1,[sp,#`$Xoff+8*16`+4]
 	adds	$Tlo,$Tlo,$t2
 	adc	$Thi,$Thi,$t3
 	adds	$Tlo,$Tlo,$t0
 	adc	$Thi,$Thi,$t1
-	str	$Tlo,[sp,#$Xoff+0]
-	str	$Thi,[sp,#$Xoff+4]
 ___
 	&BODY_00_15(0x17);
 $code.=<<___;
-	tst	$Ktbl,#1
+	ldreq	$t0,[sp,#`$Xoff+8*(16-1)`+0]
+	ldreq	$t1,[sp,#`$Xoff+8*(16-1)`+4]
 	beq	.L16_79
 	bic	$Ktbl,$Ktbl,#1
 
@@ -324,12 +360,12 @@ $code.=<<___;
 	ldr	$t2, [$ctx,#$Boff+$lo]
 	ldr	$t3, [$ctx,#$Boff+$hi]
 	adds	$t0,$Alo,$t0
-	adc	$t1,$Ahi,$t1
-	adds	$t2,$Tlo,$t2
-	adc	$t3,$Thi,$t3
 	str	$t0, [$ctx,#$Aoff+$lo]
+	adc	$t1,$Ahi,$t1
 	str	$t1, [$ctx,#$Aoff+$hi]
+	adds	$t2,$Tlo,$t2
 	str	$t2, [$ctx,#$Boff+$lo]
+	adc	$t3,$Thi,$t3
 	str	$t3, [$ctx,#$Boff+$hi]
 
 	ldr	$Alo,[sp,#$Coff+0]
@@ -341,12 +377,12 @@ $code.=<<___;
 	ldr	$t2, [$ctx,#$Doff+$lo]
 	ldr	$t3, [$ctx,#$Doff+$hi]
 	adds	$t0,$Alo,$t0
-	adc	$t1,$Ahi,$t1
-	adds	$t2,$Tlo,$t2
-	adc	$t3,$Thi,$t3
 	str	$t0, [$ctx,#$Coff+$lo]
+	adc	$t1,$Ahi,$t1
 	str	$t1, [$ctx,#$Coff+$hi]
+	adds	$t2,$Tlo,$t2
 	str	$t2, [$ctx,#$Doff+$lo]
+	adc	$t3,$Thi,$t3
 	str	$t3, [$ctx,#$Doff+$hi]
 
 	ldr	$Tlo,[sp,#$Foff+0]
@@ -356,12 +392,12 @@ $code.=<<___;
 	ldr	$t2, [$ctx,#$Foff+$lo]
 	ldr	$t3, [$ctx,#$Foff+$hi]
 	adds	$Elo,$Elo,$t0
-	adc	$Ehi,$Ehi,$t1
-	adds	$t2,$Tlo,$t2
-	adc	$t3,$Thi,$t3
 	str	$Elo,[$ctx,#$Eoff+$lo]
+	adc	$Ehi,$Ehi,$t1
 	str	$Ehi,[$ctx,#$Eoff+$hi]
+	adds	$t2,$Tlo,$t2
 	str	$t2, [$ctx,#$Foff+$lo]
+	adc	$t3,$Thi,$t3
 	str	$t3, [$ctx,#$Foff+$hi]
 
 	ldr	$Alo,[sp,#$Goff+0]
@@ -373,12 +409,12 @@ $code.=<<___;
 	ldr	$t2, [$ctx,#$Hoff+$lo]
 	ldr	$t3, [$ctx,#$Hoff+$hi]
 	adds	$t0,$Alo,$t0
-	adc	$t1,$Ahi,$t1
-	adds	$t2,$Tlo,$t2
-	adc	$t3,$Thi,$t3
 	str	$t0, [$ctx,#$Goff+$lo]
+	adc	$t1,$Ahi,$t1
 	str	$t1, [$ctx,#$Goff+$hi]
+	adds	$t2,$Tlo,$t2
 	str	$t2, [$ctx,#$Hoff+$lo]
+	adc	$t3,$Thi,$t3
 	str	$t3, [$ctx,#$Hoff+$hi]
 
 	add	sp,sp,#640
@@ -388,13 +424,156 @@ $code.=<<___;
 	bne	.Loop
 
 	add	sp,sp,#8*9		@ destroy frame
+#if __ARM_ARCH__>=5
+	ldmia	sp!,{r4-r12,pc}
+#else
 	ldmia	sp!,{r4-r12,lr}
 	tst	lr,#1
 	moveq	pc,lr			@ be binary compatible with V4, yet
 	bx	lr			@ interoperable with Thumb ISA:-)
-.size   sha512_block_data_order,.-sha512_block_data_order
-.asciz  "SHA512 block transform for ARMv4, CRYPTOGAMS by <appro\@openssl.org>"
+#endif
+___
+
+{
+my @Sigma0=(28,34,39);
+my @Sigma1=(14,18,41);
+my @sigma0=(1, 8, 7);
+my @sigma1=(19,61,6);
+
+my $Ktbl="r3";
+my $cnt="r12";	# volatile register known as ip, intra-procedure-call scratch
+
+my @X=map("d$_",(0..15));
+my @V=($A,$B,$C,$D,$E,$F,$G,$H)=map("d$_",(16..23));
+
+sub NEON_00_15() {
+my $i=shift;
+my ($a,$b,$c,$d,$e,$f,$g,$h)=@_;
+my ($t0,$t1,$t2,$T1,$K,$Ch,$Maj)=map("d$_",(24..31));	# temps
+
+$code.=<<___ if ($i<16 || $i&1);
+	vshr.u64	$t0,$e,#@Sigma1[0]	@ $i
+#if $i<16
+	vld1.64		{@X[$i%16]},[$inp]!	@ handles unaligned
+#endif
+	vshr.u64	$t1,$e,#@Sigma1[1]
+	vshr.u64	$t2,$e,#@Sigma1[2]
+___
+$code.=<<___;
+	vld1.64		{$K},[$Ktbl,:64]!	@ K[i++]
+	vsli.64		$t0,$e,#`64-@Sigma1[0]`
+	vsli.64		$t1,$e,#`64-@Sigma1[1]`
+	vsli.64		$t2,$e,#`64-@Sigma1[2]`
+#if $i<16 && defined(__ARMEL__)
+	vrev64.8	@X[$i],@X[$i]
+#endif
+	vadd.i64	$T1,$K,$h
+	veor		$Ch,$f,$g
+	veor		$t0,$t1
+	vand		$Ch,$e
+	veor		$t0,$t2			@ Sigma1(e)
+	veor		$Ch,$g			@ Ch(e,f,g)
+	vadd.i64	$T1,$t0
+	vshr.u64	$t0,$a,#@Sigma0[0]
+	vadd.i64	$T1,$Ch
+	vshr.u64	$t1,$a,#@Sigma0[1]
+	vshr.u64	$t2,$a,#@Sigma0[2]
+	vsli.64		$t0,$a,#`64-@Sigma0[0]`
+	vsli.64		$t1,$a,#`64-@Sigma0[1]`
+	vsli.64		$t2,$a,#`64-@Sigma0[2]`
+	vadd.i64	$T1,@X[$i%16]
+	vorr		$Maj,$a,$c
+	vand		$Ch,$a,$c
+	veor		$h,$t0,$t1
+	vand		$Maj,$b
+	veor		$h,$t2			@ Sigma0(a)
+	vorr		$Maj,$Ch		@ Maj(a,b,c)
+	vadd.i64	$h,$T1
+	vadd.i64	$d,$T1
+	vadd.i64	$h,$Maj
+___
+}
+
+sub NEON_16_79() {
+my $i=shift;
+
+if ($i&1)	{ &NEON_00_15($i,@_); return; }
+
+# 2x-vectorized, therefore runs every 2nd round
+my @X=map("q$_",(0..7));			# view @X as 128-bit vector
+my ($t0,$t1,$s0,$s1) = map("q$_",(12..15));	# temps
+my ($d0,$d1,$d2) = map("d$_",(24..26));		# temps from NEON_00_15
+my $e=@_[4];					# $e from NEON_00_15
+$i /= 2;
+$code.=<<___;
+	vshr.u64	$t0,@X[($i+7)%8],#@sigma1[0]
+	vshr.u64	$t1,@X[($i+7)%8],#@sigma1[1]
+	vshr.u64	$s1,@X[($i+7)%8],#@sigma1[2]
+	vsli.64		$t0,@X[($i+7)%8],#`64-@sigma1[0]`
+	vext.8		$s0,@X[$i%8],@X[($i+1)%8],#8	@ X[i+1]
+	vsli.64		$t1,@X[($i+7)%8],#`64-@sigma1[1]`
+	veor		$s1,$t0
+	vshr.u64	$t0,$s0,#@sigma0[0]
+	veor		$s1,$t1				@ sigma1(X[i+14])
+	vshr.u64	$t1,$s0,#@sigma0[1]
+	vadd.i64	@X[$i%8],$s1
+	vshr.u64	$s1,$s0,#@sigma0[2]
+	vsli.64		$t0,$s0,#`64-@sigma0[0]`
+	vsli.64		$t1,$s0,#`64-@sigma0[1]`
+	vext.8		$s0,@X[($i+4)%8],@X[($i+5)%8],#8	@ X[i+9]
+	veor		$s1,$t0
+	vshr.u64	$d0,$e,#@Sigma1[0]		@ from NEON_00_15
+	vadd.i64	@X[$i%8],$s0
+	vshr.u64	$d1,$e,#@Sigma1[1]		@ from NEON_00_15
+	veor		$s1,$t1				@ sigma0(X[i+1])
+	vshr.u64	$d2,$e,#@Sigma1[2]		@ from NEON_00_15
+	vadd.i64	@X[$i%8],$s1
+___
+	&NEON_00_15(2*$i,@_);
+}
+
+$code.=<<___;
+#if __ARM_ARCH__>=7
+.fpu	neon
+
+.align	4
+.LNEON:
+	dmb				@ errata #451034 on early Cortex A8
+	vstmdb	sp!,{d8-d15}		@ ABI specification says so
+	sub	$Ktbl,r3,#672		@ K512
+	vldmia	$ctx,{$A-$H}		@ load context
+.Loop_neon:
+___
+for($i=0;$i<16;$i++)	{ &NEON_00_15($i,@V); unshift(@V,pop(@V)); }
+$code.=<<___;
+	mov		$cnt,#4
+.L16_79_neon:
+	subs		$cnt,#1
+___
+for(;$i<32;$i++)	{ &NEON_16_79($i,@V); unshift(@V,pop(@V)); }
+$code.=<<___;
+	bne		.L16_79_neon
+
+	vldmia		$ctx,{d24-d31}	@ load context to temp
+	vadd.i64	q8,q12		@ vectorized accumulate
+	vadd.i64	q9,q13
+	vadd.i64	q10,q14
+	vadd.i64	q11,q15
+	vstmia		$ctx,{$A-$H}	@ save context
+	teq		$inp,$len
+	sub		$Ktbl,#640	@ rewind K512
+	bne		.Loop_neon
+
+	vldmia	sp!,{d8-d15}		@ epilogue
+	bx	lr
+#endif
+___
+}
+$code.=<<___;
+.size	sha512_block_data_order,.-sha512_block_data_order
+.asciz	"SHA512 block transform for ARMv4/NEON, CRYPTOGAMS by <appro\@openssl.org>"
 .align	2
+.comm	OPENSSL_armcap_P,4,4
 ___
 
 $code =~ s/\`([^\`]*)\`/eval $1/gem;
diff --git a/openssl/crypto/sha/asm/sha512-mips.pl b/openssl/crypto/sha/asm/sha512-mips.pl
new file mode 100644
index 000000000..ba5b25089
--- /dev/null
+++ b/openssl/crypto/sha/asm/sha512-mips.pl
@@ -0,0 +1,455 @@
+#!/usr/bin/env perl
+
+# ====================================================================
+# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+
+# SHA2 block procedures for MIPS.
+
+# October 2010.
+#
+# SHA256 performance improvement on MIPS R5000 CPU is ~27% over gcc-
+# generated code in o32 build and ~55% in n32/64 build. SHA512 [which
+# for now can only be compiled for MIPS64 ISA] improvement is modest
+# ~17%, but it comes for free, because it's same instruction sequence.
+# Improvement coefficients are for aligned input.
+
+######################################################################
+# There is a number of MIPS ABI in use, O32 and N32/64 are most
+# widely used. Then there is a new contender: NUBI. It appears that if
+# one picks the latter, it's possible to arrange code in ABI neutral
+# manner. Therefore let's stick to NUBI register layout:
+#
+($zero,$at,$t0,$t1,$t2)=map("\$$_",(0..2,24,25));
+($a0,$a1,$a2,$a3,$a4,$a5,$a6,$a7)=map("\$$_",(4..11));
+($s0,$s1,$s2,$s3,$s4,$s5,$s6,$s7,$s8,$s9,$s10,$s11)=map("\$$_",(12..23));
+($gp,$tp,$sp,$fp,$ra)=map("\$$_",(3,28..31));
+#
+# The return value is placed in $a0. Following coding rules facilitate
+# interoperability:
+#
+# - never ever touch $tp, "thread pointer", former $gp [o32 can be
+#   excluded from the rule, because it's specified volatile];
+# - copy return value to $t0, former $v0 [or to $a0 if you're adapting
+#   old code];
+# - on O32 populate $a4-$a7 with 'lw $aN,4*N($sp)' if necessary;
+#
+# For reference here is register layout for N32/64 MIPS ABIs:
+#
+# ($zero,$at,$v0,$v1)=map("\$$_",(0..3));
+# ($a0,$a1,$a2,$a3,$a4,$a5,$a6,$a7)=map("\$$_",(4..11));
+# ($t0,$t1,$t2,$t3,$t8,$t9)=map("\$$_",(12..15,24,25));
+# ($s0,$s1,$s2,$s3,$s4,$s5,$s6,$s7)=map("\$$_",(16..23));
+# ($gp,$sp,$fp,$ra)=map("\$$_",(28..31));
+#
+$flavour = shift; # supported flavours are o32,n32,64,nubi32,nubi64
+
+if ($flavour =~ /64|n32/i) {
+	$PTR_ADD="dadd";	# incidentally works even on n32
+	$PTR_SUB="dsub";	# incidentally works even on n32
+	$REG_S="sd";
+	$REG_L="ld";
+	$PTR_SLL="dsll";	# incidentally works even on n32
+	$SZREG=8;
+} else {
+	$PTR_ADD="add";
+	$PTR_SUB="sub";
+	$REG_S="sw";
+	$REG_L="lw";
+	$PTR_SLL="sll";
+	$SZREG=4;
+}
+$pf = ($flavour =~ /nubi/i) ? $t0 : $t2;
+#
+# <appro@openssl.org>
+#
+######################################################################
+
+$big_endian=(`echo MIPSEL | $ENV{CC} -E -P -`=~/MIPSEL/)?1:0;
+
+for (@ARGV) {	$output=$_ if (/^\w[\w\-]*\.\w+$/);	}
+open STDOUT,">$output";
+
+if (!defined($big_endian)) { $big_endian=(unpack('L',pack('N',1))==1); }
+
+if ($output =~ /512/) {
+	$label="512";
+	$SZ=8;
+	$LD="ld";		# load from memory
+	$ST="sd";		# store to memory
+	$SLL="dsll";		# shift left logical
+	$SRL="dsrl";		# shift right logical
+	$ADDU="daddu";
+	@Sigma0=(28,34,39);
+	@Sigma1=(14,18,41);
+	@sigma0=( 7, 1, 8);	# right shift first
+	@sigma1=( 6,19,61);	# right shift first
+	$lastK=0x817;
+	$rounds=80;
+} else {
+	$label="256";
+	$SZ=4;
+	$LD="lw";		# load from memory
+	$ST="sw";		# store to memory
+	$SLL="sll";		# shift left logical
+	$SRL="srl";		# shift right logical
+	$ADDU="addu";
+	@Sigma0=( 2,13,22);
+	@Sigma1=( 6,11,25);
+	@sigma0=( 3, 7,18);	# right shift first
+	@sigma1=(10,17,19);	# right shift first
+	$lastK=0x8f2;
+	$rounds=64;
+}
+
+$MSB = $big_endian ? 0 : ($SZ-1);
+$LSB = ($SZ-1)&~$MSB;
+
+@V=($A,$B,$C,$D,$E,$F,$G,$H)=map("\$$_",(1,2,3,7,24,25,30,31));
+@X=map("\$$_",(8..23));
+
+$ctx=$a0;
+$inp=$a1;
+$len=$a2;	$Ktbl=$len;
+
+sub BODY_00_15 {
+my ($i,$a,$b,$c,$d,$e,$f,$g,$h)=@_;
+my ($T1,$tmp0,$tmp1,$tmp2)=(@X[4],@X[5],@X[6],@X[7]);
+
+$code.=<<___ if ($i<15);
+	${LD}l	@X[1],`($i+1)*$SZ+$MSB`($inp)
+	${LD}r	@X[1],`($i+1)*$SZ+$LSB`($inp)
+___
+$code.=<<___	if (!$big_endian && $i<16 && $SZ==4);
+	srl	$tmp0,@X[0],24		# byte swap($i)
+	srl	$tmp1,@X[0],8
+	andi	$tmp2,@X[0],0xFF00
+	sll	@X[0],@X[0],24
+	andi	$tmp1,0xFF00
+	sll	$tmp2,$tmp2,8
+	or	@X[0],$tmp0
+	or	$tmp1,$tmp2
+	or	@X[0],$tmp1
+___
+$code.=<<___	if (!$big_endian && $i<16 && $SZ==8);
+	ori	$tmp0,$zero,0xFF
+	dsll	$tmp2,$tmp0,32
+	or	$tmp0,$tmp2		# 0x000000FF000000FF
+	and	$tmp1,@X[0],$tmp0	# byte swap($i)
+	dsrl	$tmp2,@X[0],24
+	dsll	$tmp1,24
+	and	$tmp2,$tmp0
+	dsll	$tmp0,8			# 0x0000FF000000FF00
+	or	$tmp1,$tmp2
+	and	$tmp2,@X[0],$tmp0
+	dsrl	@X[0],8
+	dsll	$tmp2,8
+	and	@X[0],$tmp0
+	or	$tmp1,$tmp2
+	or	@X[0],$tmp1
+	dsrl	$tmp1,@X[0],32
+	dsll	@X[0],32
+	or	@X[0],$tmp1
+___
+$code.=<<___;
+	$ADDU	$T1,$X[0],$h			# $i
+	$SRL	$h,$e,@Sigma1[0]
+	xor	$tmp2,$f,$g
+	$SLL	$tmp1,$e,`$SZ*8-@Sigma1[2]`
+	and	$tmp2,$e
+	$SRL	$tmp0,$e,@Sigma1[1]
+	xor	$h,$tmp1
+	$SLL	$tmp1,$e,`$SZ*8-@Sigma1[1]`
+	xor	$h,$tmp0
+	$SRL	$tmp0,$e,@Sigma1[2]
+	xor	$h,$tmp1
+	$SLL	$tmp1,$e,`$SZ*8-@Sigma1[0]`
+	xor	$h,$tmp0
+	xor	$tmp2,$g			# Ch(e,f,g)
+	xor	$tmp0,$tmp1,$h			# Sigma1(e)
+
+	$SRL	$h,$a,@Sigma0[0]
+	$ADDU	$T1,$tmp2
+	$LD	$tmp2,`$i*$SZ`($Ktbl)		# K[$i]
+	$SLL	$tmp1,$a,`$SZ*8-@Sigma0[2]`
+	$ADDU	$T1,$tmp0
+	$SRL	$tmp0,$a,@Sigma0[1]
+	xor	$h,$tmp1
+	$SLL	$tmp1,$a,`$SZ*8-@Sigma0[1]`
+	xor	$h,$tmp0
+	$SRL	$tmp0,$a,@Sigma0[2]
+	xor	$h,$tmp1
+	$SLL	$tmp1,$a,`$SZ*8-@Sigma0[0]`
+	xor	$h,$tmp0
+	$ST	@X[0],`($i%16)*$SZ`($sp)	# offload to ring buffer
+	xor	$h,$tmp1			# Sigma0(a)
+
+	or	$tmp0,$a,$b
+	and	$tmp1,$a,$b
+	and	$tmp0,$c
+	or	$tmp1,$tmp0			# Maj(a,b,c)
+	$ADDU	$T1,$tmp2			# +=K[$i]
+	$ADDU	$h,$tmp1
+
+	$ADDU	$d,$T1
+	$ADDU	$h,$T1
+___
+$code.=<<___ if ($i>=13);
+	$LD	@X[3],`(($i+3)%16)*$SZ`($sp)	# prefetch from ring buffer
+___
+}
+
+sub BODY_16_XX {
+my $i=@_[0];
+my ($tmp0,$tmp1,$tmp2,$tmp3)=(@X[4],@X[5],@X[6],@X[7]);
+
+$code.=<<___;
+	$SRL	$tmp2,@X[1],@sigma0[0]		# Xupdate($i)
+	$ADDU	@X[0],@X[9]			# +=X[i+9]
+	$SLL	$tmp1,@X[1],`$SZ*8-@sigma0[2]`
+	$SRL	$tmp0,@X[1],@sigma0[1]
+	xor	$tmp2,$tmp1
+	$SLL	$tmp1,`@sigma0[2]-@sigma0[1]`
+	xor	$tmp2,$tmp0
+	$SRL	$tmp0,@X[1],@sigma0[2]
+	xor	$tmp2,$tmp1
+
+	$SRL	$tmp3,@X[14],@sigma1[0]
+	xor	$tmp2,$tmp0			# sigma0(X[i+1])
+	$SLL	$tmp1,@X[14],`$SZ*8-@sigma1[2]`
+	$ADDU	@X[0],$tmp2
+	$SRL	$tmp0,@X[14],@sigma1[1]
+	xor	$tmp3,$tmp1
+	$SLL	$tmp1,`@sigma1[2]-@sigma1[1]`
+	xor	$tmp3,$tmp0
+	$SRL	$tmp0,@X[14],@sigma1[2]
+	xor	$tmp3,$tmp1
+
+	xor	$tmp3,$tmp0			# sigma1(X[i+14])
+	$ADDU	@X[0],$tmp3
+___
+	&BODY_00_15(@_);
+}
+
+$FRAMESIZE=16*$SZ+16*$SZREG;
+$SAVED_REGS_MASK = ($flavour =~ /nubi/i) ? 0xc0fff008 : 0xc0ff0000;
+
+$code.=<<___;
+#ifdef OPENSSL_FIPSCANISTER
+# include <openssl/fipssyms.h>
+#endif
+
+.text
+.set	noat
+#if !defined(__vxworks) || defined(__pic__)
+.option	pic2
+#endif
+
+.align	5
+.globl	sha${label}_block_data_order
+.ent	sha${label}_block_data_order
+sha${label}_block_data_order:
+	.frame	$sp,$FRAMESIZE,$ra
+	.mask	$SAVED_REGS_MASK,-$SZREG
+	.set	noreorder
+___
+$code.=<<___ if ($flavour =~ /o32/i);	# o32 PIC-ification
+	.cpload	$pf
+___
+$code.=<<___;
+	$PTR_SUB $sp,$FRAMESIZE
+	$REG_S	$ra,$FRAMESIZE-1*$SZREG($sp)
+	$REG_S	$fp,$FRAMESIZE-2*$SZREG($sp)
+	$REG_S	$s11,$FRAMESIZE-3*$SZREG($sp)
+	$REG_S	$s10,$FRAMESIZE-4*$SZREG($sp)
+	$REG_S	$s9,$FRAMESIZE-5*$SZREG($sp)
+	$REG_S	$s8,$FRAMESIZE-6*$SZREG($sp)
+	$REG_S	$s7,$FRAMESIZE-7*$SZREG($sp)
+	$REG_S	$s6,$FRAMESIZE-8*$SZREG($sp)
+	$REG_S	$s5,$FRAMESIZE-9*$SZREG($sp)
+	$REG_S	$s4,$FRAMESIZE-10*$SZREG($sp)
+___
+$code.=<<___ if ($flavour =~ /nubi/i);	# optimize non-nubi prologue
+	$REG_S	$s3,$FRAMESIZE-11*$SZREG($sp)
+	$REG_S	$s2,$FRAMESIZE-12*$SZREG($sp)
+	$REG_S	$s1,$FRAMESIZE-13*$SZREG($sp)
+	$REG_S	$s0,$FRAMESIZE-14*$SZREG($sp)
+	$REG_S	$gp,$FRAMESIZE-15*$SZREG($sp)
+___
+$code.=<<___;
+	$PTR_SLL @X[15],$len,`log(16*$SZ)/log(2)`
+___
+$code.=<<___ if ($flavour !~ /o32/i);	# non-o32 PIC-ification
+	.cplocal	$Ktbl
+	.cpsetup	$pf,$zero,sha${label}_block_data_order
+___
+$code.=<<___;
+	.set	reorder
+	la	$Ktbl,K${label}		# PIC-ified 'load address'
+
+	$LD	$A,0*$SZ($ctx)		# load context
+	$LD	$B,1*$SZ($ctx)
+	$LD	$C,2*$SZ($ctx)
+	$LD	$D,3*$SZ($ctx)
+	$LD	$E,4*$SZ($ctx)
+	$LD	$F,5*$SZ($ctx)
+	$LD	$G,6*$SZ($ctx)
+	$LD	$H,7*$SZ($ctx)
+
+	$PTR_ADD @X[15],$inp		# pointer to the end of input
+	$REG_S	@X[15],16*$SZ($sp)
+	b	.Loop
+
+.align	5
+.Loop:
+	${LD}l	@X[0],$MSB($inp)
+	${LD}r	@X[0],$LSB($inp)
+___
+for ($i=0;$i<16;$i++)
+{ &BODY_00_15($i,@V); unshift(@V,pop(@V)); push(@X,shift(@X)); }
+$code.=<<___;
+	b	.L16_xx
+.align	4
+.L16_xx:
+___
+for (;$i<32;$i++)
+{ &BODY_16_XX($i,@V); unshift(@V,pop(@V)); push(@X,shift(@X)); }
+$code.=<<___;
+	and	@X[6],0xfff
+	li	@X[7],$lastK
+	.set	noreorder
+	bne	@X[6],@X[7],.L16_xx
+	$PTR_ADD $Ktbl,16*$SZ		# Ktbl+=16
+
+	$REG_L	@X[15],16*$SZ($sp)	# restore pointer to the end of input
+	$LD	@X[0],0*$SZ($ctx)
+	$LD	@X[1],1*$SZ($ctx)
+	$LD	@X[2],2*$SZ($ctx)
+	$PTR_ADD $inp,16*$SZ
+	$LD	@X[3],3*$SZ($ctx)
+	$ADDU	$A,@X[0]
+	$LD	@X[4],4*$SZ($ctx)
+	$ADDU	$B,@X[1]
+	$LD	@X[5],5*$SZ($ctx)
+	$ADDU	$C,@X[2]
+	$LD	@X[6],6*$SZ($ctx)
+	$ADDU	$D,@X[3]
+	$LD	@X[7],7*$SZ($ctx)
+	$ADDU	$E,@X[4]
+	$ST	$A,0*$SZ($ctx)
+	$ADDU	$F,@X[5]
+	$ST	$B,1*$SZ($ctx)
+	$ADDU	$G,@X[6]
+	$ST	$C,2*$SZ($ctx)
+	$ADDU	$H,@X[7]
+	$ST	$D,3*$SZ($ctx)
+	$ST	$E,4*$SZ($ctx)
+	$ST	$F,5*$SZ($ctx)
+	$ST	$G,6*$SZ($ctx)
+	$ST	$H,7*$SZ($ctx)
+
+	bnel	$inp,@X[15],.Loop
+	$PTR_SUB $Ktbl,`($rounds-16)*$SZ`	# rewind $Ktbl
+
+	$REG_L	$ra,$FRAMESIZE-1*$SZREG($sp)
+	$REG_L	$fp,$FRAMESIZE-2*$SZREG($sp)
+	$REG_L	$s11,$FRAMESIZE-3*$SZREG($sp)
+	$REG_L	$s10,$FRAMESIZE-4*$SZREG($sp)
+	$REG_L	$s9,$FRAMESIZE-5*$SZREG($sp)
+	$REG_L	$s8,$FRAMESIZE-6*$SZREG($sp)
+	$REG_L	$s7,$FRAMESIZE-7*$SZREG($sp)
+	$REG_L	$s6,$FRAMESIZE-8*$SZREG($sp)
+	$REG_L	$s5,$FRAMESIZE-9*$SZREG($sp)
+	$REG_L	$s4,$FRAMESIZE-10*$SZREG($sp)
+___
+$code.=<<___ if ($flavour =~ /nubi/i);
+	$REG_L	$s3,$FRAMESIZE-11*$SZREG($sp)
+	$REG_L	$s2,$FRAMESIZE-12*$SZREG($sp)
+	$REG_L	$s1,$FRAMESIZE-13*$SZREG($sp)
+	$REG_L	$s0,$FRAMESIZE-14*$SZREG($sp)
+	$REG_L	$gp,$FRAMESIZE-15*$SZREG($sp)
+___
+$code.=<<___;
+	jr	$ra
+	$PTR_ADD $sp,$FRAMESIZE
+.end	sha${label}_block_data_order
+
+.rdata
+.align	5
+K${label}:
+___
+if ($SZ==4) {
+$code.=<<___;
+	.word	0x428a2f98, 0x71374491, 0xb5c0fbcf, 0xe9b5dba5
+	.word	0x3956c25b, 0x59f111f1, 0x923f82a4, 0xab1c5ed5
+	.word	0xd807aa98, 0x12835b01, 0x243185be, 0x550c7dc3
+	.word	0x72be5d74, 0x80deb1fe, 0x9bdc06a7, 0xc19bf174
+	.word	0xe49b69c1, 0xefbe4786, 0x0fc19dc6, 0x240ca1cc
+	.word	0x2de92c6f, 0x4a7484aa, 0x5cb0a9dc, 0x76f988da
+	.word	0x983e5152, 0xa831c66d, 0xb00327c8, 0xbf597fc7
+	.word	0xc6e00bf3, 0xd5a79147, 0x06ca6351, 0x14292967
+	.word	0x27b70a85, 0x2e1b2138, 0x4d2c6dfc, 0x53380d13
+	.word	0x650a7354, 0x766a0abb, 0x81c2c92e, 0x92722c85
+	.word	0xa2bfe8a1, 0xa81a664b, 0xc24b8b70, 0xc76c51a3
+	.word	0xd192e819, 0xd6990624, 0xf40e3585, 0x106aa070
+	.word	0x19a4c116, 0x1e376c08, 0x2748774c, 0x34b0bcb5
+	.word	0x391c0cb3, 0x4ed8aa4a, 0x5b9cca4f, 0x682e6ff3
+	.word	0x748f82ee, 0x78a5636f, 0x84c87814, 0x8cc70208
+	.word	0x90befffa, 0xa4506ceb, 0xbef9a3f7, 0xc67178f2
+___
+} else {
+$code.=<<___;
+	.dword	0x428a2f98d728ae22, 0x7137449123ef65cd
+	.dword	0xb5c0fbcfec4d3b2f, 0xe9b5dba58189dbbc
+	.dword	0x3956c25bf348b538, 0x59f111f1b605d019
+	.dword	0x923f82a4af194f9b, 0xab1c5ed5da6d8118
+	.dword	0xd807aa98a3030242, 0x12835b0145706fbe
+	.dword	0x243185be4ee4b28c, 0x550c7dc3d5ffb4e2
+	.dword	0x72be5d74f27b896f, 0x80deb1fe3b1696b1
+	.dword	0x9bdc06a725c71235, 0xc19bf174cf692694
+	.dword	0xe49b69c19ef14ad2, 0xefbe4786384f25e3
+	.dword	0x0fc19dc68b8cd5b5, 0x240ca1cc77ac9c65
+	.dword	0x2de92c6f592b0275, 0x4a7484aa6ea6e483
+	.dword	0x5cb0a9dcbd41fbd4, 0x76f988da831153b5
+	.dword	0x983e5152ee66dfab, 0xa831c66d2db43210
+	.dword	0xb00327c898fb213f, 0xbf597fc7beef0ee4
+	.dword	0xc6e00bf33da88fc2, 0xd5a79147930aa725
+	.dword	0x06ca6351e003826f, 0x142929670a0e6e70
+	.dword	0x27b70a8546d22ffc, 0x2e1b21385c26c926
+	.dword	0x4d2c6dfc5ac42aed, 0x53380d139d95b3df
+	.dword	0x650a73548baf63de, 0x766a0abb3c77b2a8
+	.dword	0x81c2c92e47edaee6, 0x92722c851482353b
+	.dword	0xa2bfe8a14cf10364, 0xa81a664bbc423001
+	.dword	0xc24b8b70d0f89791, 0xc76c51a30654be30
+	.dword	0xd192e819d6ef5218, 0xd69906245565a910
+	.dword	0xf40e35855771202a, 0x106aa07032bbd1b8
+	.dword	0x19a4c116b8d2d0c8, 0x1e376c085141ab53
+	.dword	0x2748774cdf8eeb99, 0x34b0bcb5e19b48a8
+	.dword	0x391c0cb3c5c95a63, 0x4ed8aa4ae3418acb
+	.dword	0x5b9cca4f7763e373, 0x682e6ff3d6b2b8a3
+	.dword	0x748f82ee5defb2fc, 0x78a5636f43172f60
+	.dword	0x84c87814a1f0ab72, 0x8cc702081a6439ec
+	.dword	0x90befffa23631e28, 0xa4506cebde82bde9
+	.dword	0xbef9a3f7b2c67915, 0xc67178f2e372532b
+	.dword	0xca273eceea26619c, 0xd186b8c721c0c207
+	.dword	0xeada7dd6cde0eb1e, 0xf57d4f7fee6ed178
+	.dword	0x06f067aa72176fba, 0x0a637dc5a2c898a6
+	.dword	0x113f9804bef90dae, 0x1b710b35131c471b
+	.dword	0x28db77f523047d84, 0x32caab7b40c72493
+	.dword	0x3c9ebe0a15c9bebc, 0x431d67c49c100d4c
+	.dword	0x4cc5d4becb3e42b6, 0x597f299cfc657e2a
+	.dword	0x5fcb6fab3ad6faec, 0x6c44198c4a475817
+___
+}
+$code.=<<___;
+.asciiz	"SHA${label} for MIPS, CRYPTOGAMS by <appro\@openssl.org>"
+.align	5
+
+___
+
+$code =~ s/\`([^\`]*)\`/eval $1/gem;
+print $code;
+close STDOUT;
diff --git a/openssl/crypto/sha/asm/sha512-parisc.pl b/openssl/crypto/sha/asm/sha512-parisc.pl
new file mode 100644
index 000000000..e24ee58ae
--- /dev/null
+++ b/openssl/crypto/sha/asm/sha512-parisc.pl
@@ -0,0 +1,791 @@
+#!/usr/bin/env perl
+
+# ====================================================================
+# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+
+# SHA256/512 block procedure for PA-RISC.
+
+# June 2009.
+#
+# SHA256 performance is >75% better than gcc 3.2 generated code on
+# PA-7100LC. Compared to code generated by vendor compiler this
+# implementation is almost 70% faster in 64-bit build, but delivers
+# virtually same performance in 32-bit build on PA-8600.
+#
+# SHA512 performance is >2.9x better than gcc 3.2 generated code on
+# PA-7100LC, PA-RISC 1.1 processor. Then implementation detects if the
+# code is executed on PA-RISC 2.0 processor and switches to 64-bit
+# code path delivering adequate peformance even in "blended" 32-bit
+# build. Though 64-bit code is not any faster than code generated by
+# vendor compiler on PA-8600...
+#
+# Special thanks to polarhome.com for providing HP-UX account.
+
+$flavour = shift;
+$output = shift;
+open STDOUT,">$output";
+
+if ($flavour =~ /64/) {
+	$LEVEL		="2.0W";
+	$SIZE_T		=8;
+	$FRAME_MARKER	=80;
+	$SAVED_RP	=16;
+	$PUSH		="std";
+	$PUSHMA		="std,ma";
+	$POP		="ldd";
+	$POPMB		="ldd,mb";
+} else {
+	$LEVEL		="1.0";
+	$SIZE_T		=4;
+	$FRAME_MARKER	=48;
+	$SAVED_RP	=20;
+	$PUSH		="stw";
+	$PUSHMA		="stwm";
+	$POP		="ldw";
+	$POPMB		="ldwm";
+}
+
+if ($output =~ /512/) {
+	$func="sha512_block_data_order";
+	$SZ=8;
+	@Sigma0=(28,34,39);
+	@Sigma1=(14,18,41);
+	@sigma0=(1,  8, 7);
+	@sigma1=(19,61, 6);
+	$rounds=80;
+	$LAST10BITS=0x017;
+	$LD="ldd";
+	$LDM="ldd,ma";
+	$ST="std";
+} else {
+	$func="sha256_block_data_order";
+	$SZ=4;
+	@Sigma0=( 2,13,22);
+	@Sigma1=( 6,11,25);
+	@sigma0=( 7,18, 3);
+	@sigma1=(17,19,10);
+	$rounds=64;
+	$LAST10BITS=0x0f2;
+	$LD="ldw";
+	$LDM="ldwm";
+	$ST="stw";
+}
+
+$FRAME=16*$SIZE_T+$FRAME_MARKER;# 16 saved regs + frame marker
+				#                 [+ argument transfer]
+$XOFF=16*$SZ+32;		# local variables
+$FRAME+=$XOFF;
+$XOFF+=$FRAME_MARKER;		# distance between %sp and local variables
+
+$ctx="%r26";	# zapped by $a0
+$inp="%r25";	# zapped by $a1
+$num="%r24";	# zapped by $t0
+
+$a0 ="%r26";
+$a1 ="%r25";
+$t0 ="%r24";
+$t1 ="%r29";
+$Tbl="%r31";
+
+@V=($A,$B,$C,$D,$E,$F,$G,$H)=("%r17","%r18","%r19","%r20","%r21","%r22","%r23","%r28");
+
+@X=("%r1", "%r2", "%r3", "%r4", "%r5", "%r6", "%r7", "%r8",
+    "%r9", "%r10","%r11","%r12","%r13","%r14","%r15","%r16",$inp);
+
+sub ROUND_00_15 {
+my ($i,$a,$b,$c,$d,$e,$f,$g,$h)=@_;
+$code.=<<___;
+	_ror	$e,$Sigma1[0],$a0
+	and	$f,$e,$t0
+	_ror	$e,$Sigma1[1],$a1
+	addl	$t1,$h,$h
+	andcm	$g,$e,$t1
+	xor	$a1,$a0,$a0
+	_ror	$a1,`$Sigma1[2]-$Sigma1[1]`,$a1
+	or	$t0,$t1,$t1		; Ch(e,f,g)
+	addl	@X[$i%16],$h,$h
+	xor	$a0,$a1,$a1		; Sigma1(e)
+	addl	$t1,$h,$h
+	_ror	$a,$Sigma0[0],$a0
+	addl	$a1,$h,$h
+
+	_ror	$a,$Sigma0[1],$a1
+	and	$a,$b,$t0
+	and	$a,$c,$t1
+	xor	$a1,$a0,$a0
+	_ror	$a1,`$Sigma0[2]-$Sigma0[1]`,$a1
+	xor	$t1,$t0,$t0
+	and	$b,$c,$t1
+	xor	$a0,$a1,$a1		; Sigma0(a)
+	addl	$h,$d,$d
+	xor	$t1,$t0,$t0		; Maj(a,b,c)
+	`"$LDM	$SZ($Tbl),$t1" if ($i<15)`
+	addl	$a1,$h,$h
+	addl	$t0,$h,$h
+
+___
+}
+
+sub ROUND_16_xx {
+my ($i,$a,$b,$c,$d,$e,$f,$g,$h)=@_;
+$i-=16;
+$code.=<<___;
+	_ror	@X[($i+1)%16],$sigma0[0],$a0
+	_ror	@X[($i+1)%16],$sigma0[1],$a1
+	addl	@X[($i+9)%16],@X[$i],@X[$i]
+	_ror	@X[($i+14)%16],$sigma1[0],$t0
+	_ror	@X[($i+14)%16],$sigma1[1],$t1
+	xor	$a1,$a0,$a0
+	_shr	@X[($i+1)%16],$sigma0[2],$a1
+	xor	$t1,$t0,$t0
+	_shr	@X[($i+14)%16],$sigma1[2],$t1
+	xor	$a1,$a0,$a0		; sigma0(X[(i+1)&0x0f])
+	xor	$t1,$t0,$t0		; sigma1(X[(i+14)&0x0f])
+	$LDM	$SZ($Tbl),$t1
+	addl	$a0,@X[$i],@X[$i]
+	addl	$t0,@X[$i],@X[$i]
+___
+$code.=<<___ if ($i==15);
+	extru	$t1,31,10,$a1
+	comiclr,<> $LAST10BITS,$a1,%r0
+	ldo	1($Tbl),$Tbl		; signal end of $Tbl
+___
+&ROUND_00_15($i+16,$a,$b,$c,$d,$e,$f,$g,$h);
+}
+
+$code=<<___;
+	.LEVEL	$LEVEL
+	.SPACE	\$TEXT\$
+	.SUBSPA	\$CODE\$,QUAD=0,ALIGN=8,ACCESS=0x2C,CODE_ONLY
+
+	.ALIGN	64
+L\$table
+___
+$code.=<<___ if ($SZ==8);
+	.WORD	0x428a2f98,0xd728ae22,0x71374491,0x23ef65cd
+	.WORD	0xb5c0fbcf,0xec4d3b2f,0xe9b5dba5,0x8189dbbc
+	.WORD	0x3956c25b,0xf348b538,0x59f111f1,0xb605d019
+	.WORD	0x923f82a4,0xaf194f9b,0xab1c5ed5,0xda6d8118
+	.WORD	0xd807aa98,0xa3030242,0x12835b01,0x45706fbe
+	.WORD	0x243185be,0x4ee4b28c,0x550c7dc3,0xd5ffb4e2
+	.WORD	0x72be5d74,0xf27b896f,0x80deb1fe,0x3b1696b1
+	.WORD	0x9bdc06a7,0x25c71235,0xc19bf174,0xcf692694
+	.WORD	0xe49b69c1,0x9ef14ad2,0xefbe4786,0x384f25e3
+	.WORD	0x0fc19dc6,0x8b8cd5b5,0x240ca1cc,0x77ac9c65
+	.WORD	0x2de92c6f,0x592b0275,0x4a7484aa,0x6ea6e483
+	.WORD	0x5cb0a9dc,0xbd41fbd4,0x76f988da,0x831153b5
+	.WORD	0x983e5152,0xee66dfab,0xa831c66d,0x2db43210
+	.WORD	0xb00327c8,0x98fb213f,0xbf597fc7,0xbeef0ee4
+	.WORD	0xc6e00bf3,0x3da88fc2,0xd5a79147,0x930aa725
+	.WORD	0x06ca6351,0xe003826f,0x14292967,0x0a0e6e70
+	.WORD	0x27b70a85,0x46d22ffc,0x2e1b2138,0x5c26c926
+	.WORD	0x4d2c6dfc,0x5ac42aed,0x53380d13,0x9d95b3df
+	.WORD	0x650a7354,0x8baf63de,0x766a0abb,0x3c77b2a8
+	.WORD	0x81c2c92e,0x47edaee6,0x92722c85,0x1482353b
+	.WORD	0xa2bfe8a1,0x4cf10364,0xa81a664b,0xbc423001
+	.WORD	0xc24b8b70,0xd0f89791,0xc76c51a3,0x0654be30
+	.WORD	0xd192e819,0xd6ef5218,0xd6990624,0x5565a910
+	.WORD	0xf40e3585,0x5771202a,0x106aa070,0x32bbd1b8
+	.WORD	0x19a4c116,0xb8d2d0c8,0x1e376c08,0x5141ab53
+	.WORD	0x2748774c,0xdf8eeb99,0x34b0bcb5,0xe19b48a8
+	.WORD	0x391c0cb3,0xc5c95a63,0x4ed8aa4a,0xe3418acb
+	.WORD	0x5b9cca4f,0x7763e373,0x682e6ff3,0xd6b2b8a3
+	.WORD	0x748f82ee,0x5defb2fc,0x78a5636f,0x43172f60
+	.WORD	0x84c87814,0xa1f0ab72,0x8cc70208,0x1a6439ec
+	.WORD	0x90befffa,0x23631e28,0xa4506ceb,0xde82bde9
+	.WORD	0xbef9a3f7,0xb2c67915,0xc67178f2,0xe372532b
+	.WORD	0xca273ece,0xea26619c,0xd186b8c7,0x21c0c207
+	.WORD	0xeada7dd6,0xcde0eb1e,0xf57d4f7f,0xee6ed178
+	.WORD	0x06f067aa,0x72176fba,0x0a637dc5,0xa2c898a6
+	.WORD	0x113f9804,0xbef90dae,0x1b710b35,0x131c471b
+	.WORD	0x28db77f5,0x23047d84,0x32caab7b,0x40c72493
+	.WORD	0x3c9ebe0a,0x15c9bebc,0x431d67c4,0x9c100d4c
+	.WORD	0x4cc5d4be,0xcb3e42b6,0x597f299c,0xfc657e2a
+	.WORD	0x5fcb6fab,0x3ad6faec,0x6c44198c,0x4a475817
+___
+$code.=<<___ if ($SZ==4);
+	.WORD	0x428a2f98,0x71374491,0xb5c0fbcf,0xe9b5dba5
+	.WORD	0x3956c25b,0x59f111f1,0x923f82a4,0xab1c5ed5
+	.WORD	0xd807aa98,0x12835b01,0x243185be,0x550c7dc3
+	.WORD	0x72be5d74,0x80deb1fe,0x9bdc06a7,0xc19bf174
+	.WORD	0xe49b69c1,0xefbe4786,0x0fc19dc6,0x240ca1cc
+	.WORD	0x2de92c6f,0x4a7484aa,0x5cb0a9dc,0x76f988da
+	.WORD	0x983e5152,0xa831c66d,0xb00327c8,0xbf597fc7
+	.WORD	0xc6e00bf3,0xd5a79147,0x06ca6351,0x14292967
+	.WORD	0x27b70a85,0x2e1b2138,0x4d2c6dfc,0x53380d13
+	.WORD	0x650a7354,0x766a0abb,0x81c2c92e,0x92722c85
+	.WORD	0xa2bfe8a1,0xa81a664b,0xc24b8b70,0xc76c51a3
+	.WORD	0xd192e819,0xd6990624,0xf40e3585,0x106aa070
+	.WORD	0x19a4c116,0x1e376c08,0x2748774c,0x34b0bcb5
+	.WORD	0x391c0cb3,0x4ed8aa4a,0x5b9cca4f,0x682e6ff3
+	.WORD	0x748f82ee,0x78a5636f,0x84c87814,0x8cc70208
+	.WORD	0x90befffa,0xa4506ceb,0xbef9a3f7,0xc67178f2
+___
+$code.=<<___;
+
+	.EXPORT	$func,ENTRY,ARGW0=GR,ARGW1=GR,ARGW2=GR
+	.ALIGN	64
+$func
+	.PROC
+	.CALLINFO	FRAME=`$FRAME-16*$SIZE_T`,NO_CALLS,SAVE_RP,ENTRY_GR=18
+	.ENTRY
+	$PUSH	%r2,-$SAVED_RP(%sp)	; standard prologue
+	$PUSHMA	%r3,$FRAME(%sp)
+	$PUSH	%r4,`-$FRAME+1*$SIZE_T`(%sp)
+	$PUSH	%r5,`-$FRAME+2*$SIZE_T`(%sp)
+	$PUSH	%r6,`-$FRAME+3*$SIZE_T`(%sp)
+	$PUSH	%r7,`-$FRAME+4*$SIZE_T`(%sp)
+	$PUSH	%r8,`-$FRAME+5*$SIZE_T`(%sp)
+	$PUSH	%r9,`-$FRAME+6*$SIZE_T`(%sp)
+	$PUSH	%r10,`-$FRAME+7*$SIZE_T`(%sp)
+	$PUSH	%r11,`-$FRAME+8*$SIZE_T`(%sp)
+	$PUSH	%r12,`-$FRAME+9*$SIZE_T`(%sp)
+	$PUSH	%r13,`-$FRAME+10*$SIZE_T`(%sp)
+	$PUSH	%r14,`-$FRAME+11*$SIZE_T`(%sp)
+	$PUSH	%r15,`-$FRAME+12*$SIZE_T`(%sp)
+	$PUSH	%r16,`-$FRAME+13*$SIZE_T`(%sp)
+	$PUSH	%r17,`-$FRAME+14*$SIZE_T`(%sp)
+	$PUSH	%r18,`-$FRAME+15*$SIZE_T`(%sp)
+
+	_shl	$num,`log(16*$SZ)/log(2)`,$num
+	addl	$inp,$num,$num		; $num to point at the end of $inp
+
+	$PUSH	$num,`-$FRAME_MARKER-4*$SIZE_T`(%sp)	; save arguments
+	$PUSH	$inp,`-$FRAME_MARKER-3*$SIZE_T`(%sp)
+	$PUSH	$ctx,`-$FRAME_MARKER-2*$SIZE_T`(%sp)
+
+	blr	%r0,$Tbl
+	ldi	3,$t1
+L\$pic
+	andcm	$Tbl,$t1,$Tbl		; wipe privilege level
+	ldo	L\$table-L\$pic($Tbl),$Tbl
+___
+$code.=<<___ if ($SZ==8 && $SIZE_T==4);
+	ldi	31,$t1
+	mtctl	$t1,%cr11
+	extrd,u,*= $t1,%sar,1,$t1	; executes on PA-RISC 1.0
+	b	L\$parisc1
+	nop
+___
+$code.=<<___;
+	$LD	`0*$SZ`($ctx),$A	; load context
+	$LD	`1*$SZ`($ctx),$B
+	$LD	`2*$SZ`($ctx),$C
+	$LD	`3*$SZ`($ctx),$D
+	$LD	`4*$SZ`($ctx),$E
+	$LD	`5*$SZ`($ctx),$F
+	$LD	`6*$SZ`($ctx),$G
+	$LD	`7*$SZ`($ctx),$H
+
+	extru	$inp,31,`log($SZ)/log(2)`,$t0
+	sh3addl	$t0,%r0,$t0
+	subi	`8*$SZ`,$t0,$t0
+	mtctl	$t0,%cr11		; load %sar with align factor
+
+L\$oop
+	ldi	`$SZ-1`,$t0
+	$LDM	$SZ($Tbl),$t1
+	andcm	$inp,$t0,$t0		; align $inp
+___
+	for ($i=0;$i<15;$i++) {		# load input block
+	$code.="\t$LD	`$SZ*$i`($t0),@X[$i]\n";		}
+$code.=<<___;
+	cmpb,*=	$inp,$t0,L\$aligned
+	$LD	`$SZ*15`($t0),@X[15]
+	$LD	`$SZ*16`($t0),@X[16]
+___
+	for ($i=0;$i<16;$i++) {		# align data
+	$code.="\t_align	@X[$i],@X[$i+1],@X[$i]\n";	}
+$code.=<<___;
+L\$aligned
+	nop	; otherwise /usr/ccs/bin/as is confused by below .WORD
+___
+
+for($i=0;$i<16;$i++)	{ &ROUND_00_15($i,@V); unshift(@V,pop(@V)); }
+$code.=<<___;
+L\$rounds
+	nop	; otherwise /usr/ccs/bin/as is confused by below .WORD
+___
+for(;$i<32;$i++)	{ &ROUND_16_xx($i,@V); unshift(@V,pop(@V)); }
+$code.=<<___;
+	bb,>=	$Tbl,31,L\$rounds	; end of $Tbl signalled?
+	nop
+
+	$POP	`-$FRAME_MARKER-2*$SIZE_T`(%sp),$ctx	; restore arguments
+	$POP	`-$FRAME_MARKER-3*$SIZE_T`(%sp),$inp
+	$POP	`-$FRAME_MARKER-4*$SIZE_T`(%sp),$num
+	ldo	`-$rounds*$SZ-1`($Tbl),$Tbl		; rewind $Tbl
+
+	$LD	`0*$SZ`($ctx),@X[0]	; load context
+	$LD	`1*$SZ`($ctx),@X[1]
+	$LD	`2*$SZ`($ctx),@X[2]
+	$LD	`3*$SZ`($ctx),@X[3]
+	$LD	`4*$SZ`($ctx),@X[4]
+	$LD	`5*$SZ`($ctx),@X[5]
+	addl	@X[0],$A,$A
+	$LD	`6*$SZ`($ctx),@X[6]
+	addl	@X[1],$B,$B
+	$LD	`7*$SZ`($ctx),@X[7]
+	ldo	`16*$SZ`($inp),$inp	; advance $inp
+
+	$ST	$A,`0*$SZ`($ctx)	; save context
+	addl	@X[2],$C,$C
+	$ST	$B,`1*$SZ`($ctx)
+	addl	@X[3],$D,$D
+	$ST	$C,`2*$SZ`($ctx)
+	addl	@X[4],$E,$E
+	$ST	$D,`3*$SZ`($ctx)
+	addl	@X[5],$F,$F
+	$ST	$E,`4*$SZ`($ctx)
+	addl	@X[6],$G,$G
+	$ST	$F,`5*$SZ`($ctx)
+	addl	@X[7],$H,$H
+	$ST	$G,`6*$SZ`($ctx)
+	$ST	$H,`7*$SZ`($ctx)
+
+	cmpb,*<>,n $inp,$num,L\$oop
+	$PUSH	$inp,`-$FRAME_MARKER-3*$SIZE_T`(%sp)	; save $inp
+___
+if ($SZ==8 && $SIZE_T==4)	# SHA512 for 32-bit PA-RISC 1.0
+{{
+$code.=<<___;
+	b	L\$done
+	nop
+
+	.ALIGN	64
+L\$parisc1
+___
+
+@V=(  $Ahi,  $Alo,  $Bhi,  $Blo,  $Chi,  $Clo,  $Dhi,  $Dlo,
+      $Ehi,  $Elo,  $Fhi,  $Flo,  $Ghi,  $Glo,  $Hhi,  $Hlo) = 
+   ( "%r1", "%r2", "%r3", "%r4", "%r5", "%r6", "%r7", "%r8",
+     "%r9","%r10","%r11","%r12","%r13","%r14","%r15","%r16");
+$a0 ="%r17";
+$a1 ="%r18";
+$a2 ="%r19";
+$a3 ="%r20";
+$t0 ="%r21";
+$t1 ="%r22";
+$t2 ="%r28";
+$t3 ="%r29";
+$Tbl="%r31";
+
+@X=("%r23","%r24","%r25","%r26");	# zaps $num,$inp,$ctx
+
+sub ROUND_00_15_pa1 {
+my ($i,$ahi,$alo,$bhi,$blo,$chi,$clo,$dhi,$dlo,
+       $ehi,$elo,$fhi,$flo,$ghi,$glo,$hhi,$hlo,$flag)=@_;
+my ($Xhi,$Xlo,$Xnhi,$Xnlo) = @X;
+
+$code.=<<___ if (!$flag);
+	ldw	`-$XOFF+8*(($i+1)%16)`(%sp),$Xnhi
+	ldw	`-$XOFF+8*(($i+1)%16)+4`(%sp),$Xnlo	; load X[i+1]
+___
+$code.=<<___;
+	shd	$ehi,$elo,$Sigma1[0],$t0
+	 add	$Xlo,$hlo,$hlo
+	shd	$elo,$ehi,$Sigma1[0],$t1
+	 addc	$Xhi,$hhi,$hhi		; h += X[i]
+	shd	$ehi,$elo,$Sigma1[1],$t2
+	 ldwm	8($Tbl),$Xhi
+	shd	$elo,$ehi,$Sigma1[1],$t3
+	 ldw	-4($Tbl),$Xlo		; load K[i]
+	xor	$t2,$t0,$t0
+	xor	$t3,$t1,$t1
+	 and	$flo,$elo,$a0
+	 and	$fhi,$ehi,$a1
+	shd	$ehi,$elo,$Sigma1[2],$t2
+	 andcm	$glo,$elo,$a2
+	shd	$elo,$ehi,$Sigma1[2],$t3
+	 andcm	$ghi,$ehi,$a3
+	xor	$t2,$t0,$t0
+	xor	$t3,$t1,$t1		; Sigma1(e)
+	add	$Xlo,$hlo,$hlo
+	 xor	$a2,$a0,$a0
+	addc	$Xhi,$hhi,$hhi		; h += K[i]
+	 xor	$a3,$a1,$a1		; Ch(e,f,g)
+
+	 add	$t0,$hlo,$hlo
+	shd	$ahi,$alo,$Sigma0[0],$t0
+	 addc	$t1,$hhi,$hhi		; h += Sigma1(e)
+	shd	$alo,$ahi,$Sigma0[0],$t1	
+	 add	$a0,$hlo,$hlo
+	shd	$ahi,$alo,$Sigma0[1],$t2
+	 addc	$a1,$hhi,$hhi		; h += Ch(e,f,g)
+	shd	$alo,$ahi,$Sigma0[1],$t3
+
+	xor	$t2,$t0,$t0
+	xor	$t3,$t1,$t1
+	shd	$ahi,$alo,$Sigma0[2],$t2
+	and	$alo,$blo,$a0
+	shd	$alo,$ahi,$Sigma0[2],$t3
+	and	$ahi,$bhi,$a1
+	xor	$t2,$t0,$t0
+	xor	$t3,$t1,$t1		; Sigma0(a)
+
+	and	$alo,$clo,$a2
+	and	$ahi,$chi,$a3
+	xor	$a2,$a0,$a0
+	 add	$hlo,$dlo,$dlo
+	xor	$a3,$a1,$a1
+	 addc	$hhi,$dhi,$dhi		; d += h
+	and	$blo,$clo,$a2
+	 add	$t0,$hlo,$hlo
+	and	$bhi,$chi,$a3
+	 addc	$t1,$hhi,$hhi		; h += Sigma0(a)
+	xor	$a2,$a0,$a0
+	 add	$a0,$hlo,$hlo
+	xor	$a3,$a1,$a1		; Maj(a,b,c)
+	 addc	$a1,$hhi,$hhi		; h += Maj(a,b,c)
+
+___
+$code.=<<___ if ($i==15 && $flag);
+	extru	$Xlo,31,10,$Xlo
+	comiclr,= $LAST10BITS,$Xlo,%r0
+	b	L\$rounds_pa1
+	nop
+___
+push(@X,shift(@X)); push(@X,shift(@X));
+}
+
+sub ROUND_16_xx_pa1 {
+my ($Xhi,$Xlo,$Xnhi,$Xnlo) = @X;
+my ($i)=shift;
+$i-=16;
+$code.=<<___;
+	ldw	`-$XOFF+8*(($i+1)%16)`(%sp),$Xnhi
+	ldw	`-$XOFF+8*(($i+1)%16)+4`(%sp),$Xnlo	; load X[i+1]
+	ldw	`-$XOFF+8*(($i+9)%16)`(%sp),$a1
+	ldw	`-$XOFF+8*(($i+9)%16)+4`(%sp),$a0	; load X[i+9]
+	ldw	`-$XOFF+8*(($i+14)%16)`(%sp),$a3
+	ldw	`-$XOFF+8*(($i+14)%16)+4`(%sp),$a2	; load X[i+14]
+	shd	$Xnhi,$Xnlo,$sigma0[0],$t0
+	shd	$Xnlo,$Xnhi,$sigma0[0],$t1
+	 add	$a0,$Xlo,$Xlo
+	shd	$Xnhi,$Xnlo,$sigma0[1],$t2
+	 addc	$a1,$Xhi,$Xhi
+	shd	$Xnlo,$Xnhi,$sigma0[1],$t3
+	xor	$t2,$t0,$t0
+	shd	$Xnhi,$Xnlo,$sigma0[2],$t2
+	xor	$t3,$t1,$t1
+	extru	$Xnhi,`31-$sigma0[2]`,`32-$sigma0[2]`,$t3
+	xor	$t2,$t0,$t0
+	 shd	$a3,$a2,$sigma1[0],$a0
+	xor	$t3,$t1,$t1		; sigma0(X[i+1)&0x0f])
+	 shd	$a2,$a3,$sigma1[0],$a1
+	add	$t0,$Xlo,$Xlo
+	 shd	$a3,$a2,$sigma1[1],$t2
+	addc	$t1,$Xhi,$Xhi
+	 shd	$a2,$a3,$sigma1[1],$t3
+	xor	$t2,$a0,$a0
+	shd	$a3,$a2,$sigma1[2],$t2
+	xor	$t3,$a1,$a1
+	extru	$a3,`31-$sigma1[2]`,`32-$sigma1[2]`,$t3
+	xor	$t2,$a0,$a0
+	xor	$t3,$a1,$a1		; sigma0(X[i+14)&0x0f])
+	add	$a0,$Xlo,$Xlo
+	addc	$a1,$Xhi,$Xhi
+
+	stw	$Xhi,`-$XOFF+8*($i%16)`(%sp)
+	stw	$Xlo,`-$XOFF+8*($i%16)+4`(%sp)
+___
+&ROUND_00_15_pa1($i,@_,1);
+}
+$code.=<<___;
+	ldw	`0*4`($ctx),$Ahi		; load context
+	ldw	`1*4`($ctx),$Alo
+	ldw	`2*4`($ctx),$Bhi
+	ldw	`3*4`($ctx),$Blo
+	ldw	`4*4`($ctx),$Chi
+	ldw	`5*4`($ctx),$Clo
+	ldw	`6*4`($ctx),$Dhi
+	ldw	`7*4`($ctx),$Dlo
+	ldw	`8*4`($ctx),$Ehi
+	ldw	`9*4`($ctx),$Elo
+	ldw	`10*4`($ctx),$Fhi
+	ldw	`11*4`($ctx),$Flo
+	ldw	`12*4`($ctx),$Ghi
+	ldw	`13*4`($ctx),$Glo
+	ldw	`14*4`($ctx),$Hhi
+	ldw	`15*4`($ctx),$Hlo
+
+	extru	$inp,31,2,$t0
+	sh3addl	$t0,%r0,$t0
+	subi	32,$t0,$t0
+	mtctl	$t0,%cr11		; load %sar with align factor
+
+L\$oop_pa1
+	extru	$inp,31,2,$a3
+	comib,=	0,$a3,L\$aligned_pa1
+	sub	$inp,$a3,$inp
+
+	ldw	`0*4`($inp),$X[0]
+	ldw	`1*4`($inp),$X[1]
+	ldw	`2*4`($inp),$t2
+	ldw	`3*4`($inp),$t3
+	ldw	`4*4`($inp),$a0
+	ldw	`5*4`($inp),$a1
+	ldw	`6*4`($inp),$a2
+	ldw	`7*4`($inp),$a3
+	vshd	$X[0],$X[1],$X[0]
+	vshd	$X[1],$t2,$X[1]
+	stw	$X[0],`-$XOFF+0*4`(%sp)
+	ldw	`8*4`($inp),$t0
+	vshd	$t2,$t3,$t2
+	stw	$X[1],`-$XOFF+1*4`(%sp)
+	ldw	`9*4`($inp),$t1
+	vshd	$t3,$a0,$t3
+___
+{
+my @t=($t2,$t3,$a0,$a1,$a2,$a3,$t0,$t1);
+for ($i=2;$i<=(128/4-8);$i++) {
+$code.=<<___;
+	stw	$t[0],`-$XOFF+$i*4`(%sp)
+	ldw	`(8+$i)*4`($inp),$t[0]
+	vshd	$t[1],$t[2],$t[1]
+___
+push(@t,shift(@t));
+}
+for (;$i<(128/4-1);$i++) {
+$code.=<<___;
+	stw	$t[0],`-$XOFF+$i*4`(%sp)
+	vshd	$t[1],$t[2],$t[1]
+___
+push(@t,shift(@t));
+}
+$code.=<<___;
+	b	L\$collected_pa1
+	stw	$t[0],`-$XOFF+$i*4`(%sp)
+
+___
+}
+$code.=<<___;
+L\$aligned_pa1
+	ldw	`0*4`($inp),$X[0]
+	ldw	`1*4`($inp),$X[1]
+	ldw	`2*4`($inp),$t2
+	ldw	`3*4`($inp),$t3
+	ldw	`4*4`($inp),$a0
+	ldw	`5*4`($inp),$a1
+	ldw	`6*4`($inp),$a2
+	ldw	`7*4`($inp),$a3
+	stw	$X[0],`-$XOFF+0*4`(%sp)
+	ldw	`8*4`($inp),$t0
+	stw	$X[1],`-$XOFF+1*4`(%sp)
+	ldw	`9*4`($inp),$t1
+___
+{
+my @t=($t2,$t3,$a0,$a1,$a2,$a3,$t0,$t1);
+for ($i=2;$i<(128/4-8);$i++) {
+$code.=<<___;
+	stw	$t[0],`-$XOFF+$i*4`(%sp)
+	ldw	`(8+$i)*4`($inp),$t[0]
+___
+push(@t,shift(@t));
+}
+for (;$i<128/4;$i++) {
+$code.=<<___;
+	stw	$t[0],`-$XOFF+$i*4`(%sp)
+___
+push(@t,shift(@t));
+}
+$code.="L\$collected_pa1\n";
+}
+
+for($i=0;$i<16;$i++)	{ &ROUND_00_15_pa1($i,@V); unshift(@V,pop(@V)); unshift(@V,pop(@V)); }
+$code.="L\$rounds_pa1\n";
+for(;$i<32;$i++)	{ &ROUND_16_xx_pa1($i,@V); unshift(@V,pop(@V)); unshift(@V,pop(@V)); }
+
+$code.=<<___;
+	$POP	`-$FRAME_MARKER-2*$SIZE_T`(%sp),$ctx	; restore arguments
+	$POP	`-$FRAME_MARKER-3*$SIZE_T`(%sp),$inp
+	$POP	`-$FRAME_MARKER-4*$SIZE_T`(%sp),$num
+	ldo	`-$rounds*$SZ`($Tbl),$Tbl		; rewind $Tbl
+
+	ldw	`0*4`($ctx),$t1		; update context
+	ldw	`1*4`($ctx),$t0
+	ldw	`2*4`($ctx),$t3
+	ldw	`3*4`($ctx),$t2
+	ldw	`4*4`($ctx),$a1
+	ldw	`5*4`($ctx),$a0
+	ldw	`6*4`($ctx),$a3
+	add	$t0,$Alo,$Alo
+	ldw	`7*4`($ctx),$a2
+	addc	$t1,$Ahi,$Ahi
+	ldw	`8*4`($ctx),$t1
+	add	$t2,$Blo,$Blo
+	ldw	`9*4`($ctx),$t0
+	addc	$t3,$Bhi,$Bhi
+	ldw	`10*4`($ctx),$t3
+	add	$a0,$Clo,$Clo
+	ldw	`11*4`($ctx),$t2
+	addc	$a1,$Chi,$Chi
+	ldw	`12*4`($ctx),$a1
+	add	$a2,$Dlo,$Dlo
+	ldw	`13*4`($ctx),$a0
+	addc	$a3,$Dhi,$Dhi
+	ldw	`14*4`($ctx),$a3
+	add	$t0,$Elo,$Elo
+	ldw	`15*4`($ctx),$a2
+	addc	$t1,$Ehi,$Ehi
+	stw	$Ahi,`0*4`($ctx)
+	add	$t2,$Flo,$Flo
+	stw	$Alo,`1*4`($ctx)
+	addc	$t3,$Fhi,$Fhi
+	stw	$Bhi,`2*4`($ctx)
+	add	$a0,$Glo,$Glo
+	stw	$Blo,`3*4`($ctx)
+	addc	$a1,$Ghi,$Ghi
+	stw	$Chi,`4*4`($ctx)
+	add	$a2,$Hlo,$Hlo
+	stw	$Clo,`5*4`($ctx)
+	addc	$a3,$Hhi,$Hhi
+	stw	$Dhi,`6*4`($ctx)
+	ldo	`16*$SZ`($inp),$inp	; advance $inp
+	stw	$Dlo,`7*4`($ctx)
+	stw	$Ehi,`8*4`($ctx)
+	stw	$Elo,`9*4`($ctx)
+	stw	$Fhi,`10*4`($ctx)
+	stw	$Flo,`11*4`($ctx)
+	stw	$Ghi,`12*4`($ctx)
+	stw	$Glo,`13*4`($ctx)
+	stw	$Hhi,`14*4`($ctx)
+	comb,=	$inp,$num,L\$done
+	stw	$Hlo,`15*4`($ctx)
+	b	L\$oop_pa1
+	$PUSH	$inp,`-$FRAME_MARKER-3*$SIZE_T`(%sp)	; save $inp
+L\$done
+___
+}}
+$code.=<<___;
+	$POP	`-$FRAME-$SAVED_RP`(%sp),%r2		; standard epilogue
+	$POP	`-$FRAME+1*$SIZE_T`(%sp),%r4
+	$POP	`-$FRAME+2*$SIZE_T`(%sp),%r5
+	$POP	`-$FRAME+3*$SIZE_T`(%sp),%r6
+	$POP	`-$FRAME+4*$SIZE_T`(%sp),%r7
+	$POP	`-$FRAME+5*$SIZE_T`(%sp),%r8
+	$POP	`-$FRAME+6*$SIZE_T`(%sp),%r9
+	$POP	`-$FRAME+7*$SIZE_T`(%sp),%r10
+	$POP	`-$FRAME+8*$SIZE_T`(%sp),%r11
+	$POP	`-$FRAME+9*$SIZE_T`(%sp),%r12
+	$POP	`-$FRAME+10*$SIZE_T`(%sp),%r13
+	$POP	`-$FRAME+11*$SIZE_T`(%sp),%r14
+	$POP	`-$FRAME+12*$SIZE_T`(%sp),%r15
+	$POP	`-$FRAME+13*$SIZE_T`(%sp),%r16
+	$POP	`-$FRAME+14*$SIZE_T`(%sp),%r17
+	$POP	`-$FRAME+15*$SIZE_T`(%sp),%r18
+	bv	(%r2)
+	.EXIT
+	$POPMB	-$FRAME(%sp),%r3
+	.PROCEND
+	.STRINGZ "SHA`64*$SZ` block transform for PA-RISC, CRYPTOGAMS by <appro\@openssl.org>"
+___
+
+# Explicitly encode PA-RISC 2.0 instructions used in this module, so
+# that it can be compiled with .LEVEL 1.0. It should be noted that I
+# wouldn't have to do this, if GNU assembler understood .ALLOW 2.0
+# directive...
+
+my $ldd = sub {
+  my ($mod,$args) = @_;
+  my $orig = "ldd$mod\t$args";
+
+    if ($args =~ /(\-?[0-9]+)\(%r([0-9]+)\),%r([0-9]+)/) # format 3 suffices
+    {	my $opcode=(0x14<<26)|($2<<21)|($3<<16)|(($1&0x1FF8)<<1)|(($1>>13)&1);
+	$opcode|=(1<<3) if ($mod =~ /^,m/);
+	$opcode|=(1<<2) if ($mod =~ /^,mb/);
+	sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig;
+    }
+    else { "\t".$orig; }
+};
+
+my $std = sub {
+  my ($mod,$args) = @_;
+  my $orig = "std$mod\t$args";
+
+    if ($args =~ /%r([0-9]+),(\-?[0-9]+)\(%r([0-9]+)\)/) # format 3 suffices
+    {	my $opcode=(0x1c<<26)|($3<<21)|($1<<16)|(($2&0x1FF8)<<1)|(($2>>13)&1);
+	sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig;
+    }
+    else { "\t".$orig; }
+};
+
+my $extrd = sub {
+  my ($mod,$args) = @_;
+  my $orig = "extrd$mod\t$args";
+
+    # I only have ",u" completer, it's implicitly encoded...
+    if ($args =~ /%r([0-9]+),([0-9]+),([0-9]+),%r([0-9]+)/)	# format 15
+    {	my $opcode=(0x36<<26)|($1<<21)|($4<<16);
+	my $len=32-$3;
+	$opcode |= (($2&0x20)<<6)|(($2&0x1f)<<5);		# encode pos
+	$opcode |= (($len&0x20)<<7)|($len&0x1f);		# encode len
+	sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig;
+    }
+    elsif ($args =~ /%r([0-9]+),%sar,([0-9]+),%r([0-9]+)/)	# format 12
+    {	my $opcode=(0x34<<26)|($1<<21)|($3<<16)|(2<<11)|(1<<9);
+	my $len=32-$2;
+	$opcode |= (($len&0x20)<<3)|($len&0x1f);		# encode len
+	$opcode |= (1<<13) if ($mod =~ /,\**=/);
+	sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig;
+    }
+    else { "\t".$orig; }
+};
+
+my $shrpd = sub {
+  my ($mod,$args) = @_;
+  my $orig = "shrpd$mod\t$args";
+
+    if ($args =~ /%r([0-9]+),%r([0-9]+),([0-9]+),%r([0-9]+)/)	# format 14
+    {	my $opcode=(0x34<<26)|($2<<21)|($1<<16)|(1<<10)|$4;
+	my $cpos=63-$3;
+	$opcode |= (($cpos&0x20)<<6)|(($cpos&0x1f)<<5);		# encode sa
+	sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig;
+    }
+    elsif ($args =~ /%r([0-9]+),%r([0-9]+),%sar,%r([0-9]+)/)	# format 11
+    {	sprintf "\t.WORD\t0x%08x\t; %s",
+		(0x34<<26)|($2<<21)|($1<<16)|(1<<9)|$3,$orig;
+    }
+    else { "\t".$orig; }
+};
+
+sub assemble {
+  my ($mnemonic,$mod,$args)=@_;
+  my $opcode = eval("\$$mnemonic");
+
+    ref($opcode) eq 'CODE' ? &$opcode($mod,$args) : "\t$mnemonic$mod\t$args";
+}
+
+foreach (split("\n",$code)) {
+	s/\`([^\`]*)\`/eval $1/ge;
+
+	s/shd\s+(%r[0-9]+),(%r[0-9]+),([0-9]+)/
+		$3>31 ? sprintf("shd\t%$2,%$1,%d",$3-32)	# rotation for >=32
+		:       sprintf("shd\t%$1,%$2,%d",$3)/e			or
+	# translate made up instructons: _ror, _shr, _align, _shl
+	s/_ror(\s+)(%r[0-9]+),/
+		($SZ==4 ? "shd" : "shrpd")."$1$2,$2,"/e			or
+
+	s/_shr(\s+%r[0-9]+),([0-9]+),/
+		$SZ==4 ? sprintf("extru%s,%d,%d,",$1,31-$2,32-$2)
+		:        sprintf("extrd,u%s,%d,%d,",$1,63-$2,64-$2)/e	or
+
+	s/_align(\s+%r[0-9]+,%r[0-9]+),/
+		($SZ==4 ? "vshd$1," : "shrpd$1,%sar,")/e		or
+
+	s/_shl(\s+%r[0-9]+),([0-9]+),/
+		$SIZE_T==4 ? sprintf("zdep%s,%d,%d,",$1,31-$2,32-$2)
+		:            sprintf("depd,z%s,%d,%d,",$1,63-$2,64-$2)/e;
+
+	s/^\s+([a-z]+)([\S]*)\s+([\S]*)/&assemble($1,$2,$3)/e if ($SIZE_T==4);
+
+	s/cmpb,\*/comb,/ if ($SIZE_T==4);
+
+	print $_,"\n";
+}
+
+close STDOUT;
diff --git a/openssl/crypto/sha/asm/sha512-ppc.pl b/openssl/crypto/sha/asm/sha512-ppc.pl
index 768a6a6fa..6b44a68e5 100644
--- a/openssl/crypto/sha/asm/sha512-ppc.pl
+++ b/openssl/crypto/sha/asm/sha512-ppc.pl
@@ -40,6 +40,7 @@ $output =shift;
 
 if ($flavour =~ /64/) {
 	$SIZE_T=8;
+	$LRSAVE=2*$SIZE_T;
 	$STU="stdu";
 	$UCMP="cmpld";
 	$SHL="sldi";
@@ -47,6 +48,7 @@ if ($flavour =~ /64/) {
 	$PUSH="std";
 } elsif ($flavour =~ /32/) {
 	$SIZE_T=4;
+	$LRSAVE=$SIZE_T;
 	$STU="stwu";
 	$UCMP="cmplw";
 	$SHL="slwi";
@@ -87,7 +89,8 @@ if ($output =~ /512/) {
 	$SHR="srwi";
 }
 
-$FRAME=32*$SIZE_T;
+$FRAME=32*$SIZE_T+16*$SZ;
+$LOCALS=6*$SIZE_T;
 
 $sp ="r1";
 $toc="r2";
@@ -179,13 +182,12 @@ $code=<<___;
 .globl	$func
 .align	6
 $func:
+	$STU	$sp,-$FRAME($sp)
 	mflr	r0
-	$STU	$sp,`-($FRAME+16*$SZ)`($sp)
 	$SHL	$num,$num,`log(16*$SZ)/log(2)`
 
 	$PUSH	$ctx,`$FRAME-$SIZE_T*22`($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)
@@ -206,6 +208,7 @@ $func:
 	$PUSH	r29,`$FRAME-$SIZE_T*3`($sp)
 	$PUSH	r30,`$FRAME-$SIZE_T*2`($sp)
 	$PUSH	r31,`$FRAME-$SIZE_T*1`($sp)
+	$PUSH	r0,`$FRAME+$LRSAVE`($sp)
 
 	$LD	$A,`0*$SZ`($ctx)
 	mr	$inp,r4				; incarnate $inp
@@ -217,7 +220,7 @@ $func:
 	$LD	$G,`6*$SZ`($ctx)
 	$LD	$H,`7*$SZ`($ctx)
 
-	b	LPICmeup
+	bl	LPICmeup
 LPICedup:
 	andi.	r0,$inp,3
 	bne	Lunaligned
@@ -226,40 +229,14 @@ Laligned:
 	$PUSH	$num,`$FRAME-$SIZE_T*24`($sp)	; end pointer
 	$PUSH	$inp,`$FRAME-$SIZE_T*23`($sp)	; inp pointer
 	bl	Lsha2_block_private
-Ldone:
-	$POP	r0,`$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,`$FRAME+16*$SZ`
-	blr
-___
+	b	Ldone
 
-# PowerPC specification allows an implementation to be ill-behaved
-# upon unaligned access which crosses page boundary. "Better safe
-# than sorry" principle makes me treat it specially. But I don't
-# look for particular offending word, but rather for the input
-# block which crosses the boundary. Once found that block is aligned
-# and hashed separately...
-$code.=<<___;
+; PowerPC specification allows an implementation to be ill-behaved
+; upon unaligned access which crosses page boundary. "Better safe
+; than sorry" principle makes me treat it specially. But I don't
+; look for particular offending word, but rather for the input
+; block which crosses the boundary. Once found that block is aligned
+; and hashed separately...
 .align	4
 Lunaligned:
 	subfic	$t1,$inp,4096
@@ -278,7 +255,7 @@ Lunaligned:
 Lcross_page:
 	li	$t1,`16*$SZ/4`
 	mtctr	$t1
-	addi	r20,$sp,$FRAME			; aligned spot below the frame
+	addi	r20,$sp,$LOCALS			; aligned spot below the frame
 Lmemcpy:
 	lbz	r16,0($inp)
 	lbz	r17,1($inp)
@@ -293,8 +270,8 @@ Lmemcpy:
 	bdnz	Lmemcpy
 
 	$PUSH	$inp,`$FRAME-$SIZE_T*26`($sp)	; save real inp
-	addi	$t1,$sp,`$FRAME+16*$SZ`		; fictitious end pointer
-	addi	$inp,$sp,$FRAME			; fictitious inp pointer
+	addi	$t1,$sp,`$LOCALS+16*$SZ`	; fictitious end pointer
+	addi	$inp,$sp,$LOCALS		; fictitious inp pointer
 	$PUSH	$num,`$FRAME-$SIZE_T*25`($sp)	; save real num
 	$PUSH	$t1,`$FRAME-$SIZE_T*24`($sp)	; end pointer
 	$PUSH	$inp,`$FRAME-$SIZE_T*23`($sp)	; inp pointer
@@ -303,10 +280,36 @@ Lmemcpy:
 	$POP	$num,`$FRAME-$SIZE_T*25`($sp)	; restore real num
 	addic.	$num,$num,`-16*$SZ`		; num--
 	bne-	Lunaligned
-	b	Ldone
-___
 
-$code.=<<___;
+Ldone:
+	$POP	r0,`$FRAME+$LRSAVE`($sp)
+	$POP	$toc,`$FRAME-$SIZE_T*20`($sp)
+	$POP	r13,`$FRAME-$SIZE_T*19`($sp)
+	$POP	r14,`$FRAME-$SIZE_T*18`($sp)
+	$POP	r15,`$FRAME-$SIZE_T*17`($sp)
+	$POP	r16,`$FRAME-$SIZE_T*16`($sp)
+	$POP	r17,`$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
+	.long	0
+	.byte	0,12,4,1,0x80,18,3,0
+	.long	0
+
 .align	4
 Lsha2_block_private:
 ___
@@ -372,6 +375,8 @@ $code.=<<___;
 	$ST	$H,`7*$SZ`($ctx)
 	bne	Lsha2_block_private
 	blr
+	.long	0
+	.byte	0,12,0x14,0,0,0,0,0
 ___
 
 # Ugly hack here, because PPC assembler syntax seem to vary too
@@ -379,22 +384,15 @@ ___
 $code.=<<___;
 .align	6
 LPICmeup:
-	bl	LPIC
-	addi	$Tbl,$Tbl,`64-4`	; "distance" between . and last nop
-	b	LPICedup
-	nop
-	nop
-	nop
-	nop
-	nop
-LPIC:	mflr	$Tbl
+	mflr	r0
+	bcl	20,31,\$+4
+	mflr	$Tbl	; vvvvvv "distance" between . and 1st data entry
+	addi	$Tbl,$Tbl,`64-8`
+	mtlr	r0
 	blr
-	nop
-	nop
-	nop
-	nop
-	nop
-	nop
+	.long	0
+	.byte	0,12,0x14,0,0,0,0,0
+	.space	`64-9*4`
 ___
 $code.=<<___ if ($SZ==8);
 	.long	0x428a2f98,0xd728ae22,0x71374491,0x23ef65cd
diff --git a/openssl/crypto/sha/asm/sha512-s390x.pl b/openssl/crypto/sha/asm/sha512-s390x.pl
index e7ef2d5a9..079a3fc78 100644
--- a/openssl/crypto/sha/asm/sha512-s390x.pl
+++ b/openssl/crypto/sha/asm/sha512-s390x.pl
@@ -26,6 +26,26 @@
 # favour dual-issue z10 pipeline. Hardware SHA256/512 is ~4.7x faster
 # than software.
 
+# November 2010.
+#
+# Adapt for -m31 build. If kernel supports what's called "highgprs"
+# feature on Linux [see /proc/cpuinfo], it's possible to use 64-bit
+# instructions and achieve "64-bit" performance even in 31-bit legacy
+# application context. The feature is not specific to any particular
+# processor, as long as it's "z-CPU". Latter implies that the code
+# remains z/Architecture specific. On z900 SHA256 was measured to
+# perform 2.4x and SHA512 - 13x better than code generated by gcc 4.3.
+
+$flavour = shift;
+
+if ($flavour =~ /3[12]/) {
+	$SIZE_T=4;
+	$g="";
+} else {
+	$SIZE_T=8;
+	$g="g";
+}
+
 $t0="%r0";
 $t1="%r1";
 $ctx="%r2";	$t2="%r2";
@@ -44,7 +64,7 @@ $tbl="%r13";
 $T1="%r14";
 $sp="%r15";
 
-$output=shift;
+while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {}
 open STDOUT,">$output";
 
 if ($output =~ /512/) {
@@ -78,7 +98,8 @@ if ($output =~ /512/) {
 }
 $Func="sha${label}_block_data_order";
 $Table="K${label}";
-$frame=160+16*$SZ;
+$stdframe=16*$SIZE_T+4*8;
+$frame=$stdframe+16*$SZ;
 
 sub BODY_00_15 {
 my ($i,$a,$b,$c,$d,$e,$f,$g,$h) = @_;
@@ -93,9 +114,9 @@ $code.=<<___;
 	xgr	$t0,$t1
 	$ROT	$t1,$t1,`$Sigma1[2]-$Sigma1[1]`
 	 xgr	$t2,$g
-	$ST	$T1,`160+$SZ*($i%16)`($sp)
+	$ST	$T1,`$stdframe+$SZ*($i%16)`($sp)
 	xgr	$t0,$t1			# Sigma1(e)
-	la	$T1,0($T1,$h)		# T1+=h
+	algr	$T1,$h			# T1+=h
 	 ngr	$t2,$e
 	 lgr	$t1,$a
 	algr	$T1,$t0			# T1+=Sigma1(e)
@@ -113,7 +134,7 @@ $code.=<<___;
 	 ngr	$t2,$b
 	algr	$h,$T1			# h+=T1
 	 ogr	$t2,$t1			# Maj(a,b,c)
-	la	$d,0($d,$T1)		# d+=T1
+	algr	$d,$T1			# d+=T1
 	algr	$h,$t2			# h+=Maj(a,b,c)
 ___
 }
@@ -122,19 +143,19 @@ sub BODY_16_XX {
 my ($i,$a,$b,$c,$d,$e,$f,$g,$h) = @_;
 
 $code.=<<___;
-	$LD	$T1,`160+$SZ*(($i+1)%16)`($sp)	### $i
-	$LD	$t1,`160+$SZ*(($i+14)%16)`($sp)
+	$LD	$T1,`$stdframe+$SZ*(($i+1)%16)`($sp)	### $i
+	$LD	$t1,`$stdframe+$SZ*(($i+14)%16)`($sp)
 	$ROT	$t0,$T1,$sigma0[0]
 	$SHR	$T1,$sigma0[2]
 	$ROT	$t2,$t0,`$sigma0[1]-$sigma0[0]`
 	xgr	$T1,$t0
 	$ROT	$t0,$t1,$sigma1[0]
-	xgr	$T1,$t2				# sigma0(X[i+1])
+	xgr	$T1,$t2					# sigma0(X[i+1])
 	$SHR	$t1,$sigma1[2]
-	$ADD	$T1,`160+$SZ*($i%16)`($sp)	# +=X[i]
+	$ADD	$T1,`$stdframe+$SZ*($i%16)`($sp)	# +=X[i]
 	xgr	$t1,$t0
 	$ROT	$t0,$t0,`$sigma1[1]-$sigma1[0]`
-	$ADD	$T1,`160+$SZ*(($i+9)%16)`($sp)	# +=X[i+9]
+	$ADD	$T1,`$stdframe+$SZ*(($i+9)%16)`($sp)	# +=X[i+9]
 	xgr	$t1,$t0				# sigma1(X[i+14])
 	algr	$T1,$t1				# +=sigma1(X[i+14])
 ___
@@ -212,6 +233,7 @@ $code.=<<___;
 .globl	$Func
 .type	$Func,\@function
 $Func:
+	sllg	$len,$len,`log(16*$SZ)/log(2)`
 ___
 $code.=<<___ if ($kimdfunc);
 	larl	%r1,OPENSSL_s390xcap_P
@@ -219,15 +241,15 @@ $code.=<<___ if ($kimdfunc);
 	tmhl	%r0,0x4000	# check for message-security assist
 	jz	.Lsoftware
 	lghi	%r0,0
-	la	%r1,16($sp)
+	la	%r1,`2*$SIZE_T`($sp)
 	.long	0xb93e0002	# kimd %r0,%r2
-	lg	%r0,16($sp)
+	lg	%r0,`2*$SIZE_T`($sp)
 	tmhh	%r0,`0x8000>>$kimdfunc`
 	jz	.Lsoftware
 	lghi	%r0,$kimdfunc
 	lgr	%r1,$ctx
 	lgr	%r2,$inp
-	sllg	%r3,$len,`log(16*$SZ)/log(2)`
+	lgr	%r3,$len
 	.long	0xb93e0002	# kimd %r0,%r2
 	brc	1,.-4		# pay attention to "partial completion"
 	br	%r14
@@ -235,13 +257,12 @@ $code.=<<___ if ($kimdfunc);
 .Lsoftware:
 ___
 $code.=<<___;
-	sllg	$len,$len,`log(16*$SZ)/log(2)`
 	lghi	%r1,-$frame
-	agr	$len,$inp
-	stmg	$ctx,%r15,16($sp)
+	la	$len,0($len,$inp)
+	stm${g}	$ctx,%r15,`2*$SIZE_T`($sp)
 	lgr	%r0,$sp
 	la	$sp,0(%r1,$sp)
-	stg	%r0,0($sp)
+	st${g}	%r0,0($sp)
 
 	larl	$tbl,$Table
 	$LD	$A,`0*$SZ`($ctx)
@@ -265,7 +286,7 @@ $code.=<<___;
 	clgr	$len,$t0
 	jne	.Lrounds_16_xx
 
-	lg	$ctx,`$frame+16`($sp)
+	l${g}	$ctx,`$frame+2*$SIZE_T`($sp)
 	la	$inp,`16*$SZ`($inp)
 	$ADD	$A,`0*$SZ`($ctx)
 	$ADD	$B,`1*$SZ`($ctx)
@@ -283,14 +304,14 @@ $code.=<<___;
 	$ST	$F,`5*$SZ`($ctx)
 	$ST	$G,`6*$SZ`($ctx)
 	$ST	$H,`7*$SZ`($ctx)
-	clg	$inp,`$frame+32`($sp)
+	cl${g}	$inp,`$frame+4*$SIZE_T`($sp)
 	jne	.Lloop
 
-	lmg	%r6,%r15,`$frame+48`($sp)	
+	lm${g}	%r6,%r15,`$frame+6*$SIZE_T`($sp)	
 	br	%r14
 .size	$Func,.-$Func
 .string	"SHA${label} block transform for s390x, CRYPTOGAMS by <appro\@openssl.org>"
-.comm	OPENSSL_s390xcap_P,8,8
+.comm	OPENSSL_s390xcap_P,16,8
 ___
 
 $code =~ s/\`([^\`]*)\`/eval $1/gem;
diff --git a/openssl/crypto/sha/asm/sha512-sparcv9.pl b/openssl/crypto/sha/asm/sha512-sparcv9.pl
index ec5d78135..585740789 100644
--- a/openssl/crypto/sha/asm/sha512-sparcv9.pl
+++ b/openssl/crypto/sha/asm/sha512-sparcv9.pl
@@ -305,9 +305,9 @@ $code.=<<___;
 	srlx	@X[(($i+9)/2)%8],32,$tmp1	! X[i+9]
 	xor	$tmp0,$tmp2,$tmp2		! sigma1(X[i+14])
 	srl	@X[($i/2)%8],0,$tmp0
+	add	$tmp2,$tmp1,$tmp1
 	add	$xi,$T1,$T1			! +=X[i]
 	xor	$tmp0,@X[($i/2)%8],@X[($i/2)%8]
-	add	$tmp2,$T1,$T1
 	add	$tmp1,$T1,$T1
 
 	srl	$T1,0,$T1
@@ -318,9 +318,9 @@ ___
 $code.=<<___;
 	srlx	@X[($i/2)%8],32,$tmp1		! X[i]
 	xor	$tmp0,$tmp2,$tmp2		! sigma1(X[i+14])
-	srl	@X[($i/2)%8],0,@X[($i/2)%8]
 	add	$xi,$T1,$T1			! +=X[i+9]
-	add	$tmp2,$T1,$T1
+	add	$tmp2,$tmp1,$tmp1
+	srl	@X[($i/2)%8],0,@X[($i/2)%8]
 	add	$tmp1,$T1,$T1
 
 	sllx	$T1,32,$tmp0
diff --git a/openssl/crypto/sha/asm/sha512-x86_64.pl b/openssl/crypto/sha/asm/sha512-x86_64.pl
index e6643f8cf..f611a2d89 100644
--- a/openssl/crypto/sha/asm/sha512-x86_64.pl
+++ b/openssl/crypto/sha/asm/sha512-x86_64.pl
@@ -95,50 +95,44 @@ sub ROUND_00_15()
 { my ($i,$a,$b,$c,$d,$e,$f,$g,$h) = @_;
 
 $code.=<<___;
-	mov	$e,$a0
-	mov	$e,$a1
+	ror	\$`$Sigma1[2]-$Sigma1[1]`,$a0
 	mov	$f,$a2
+	mov	$T1,`$SZ*($i&0xf)`(%rsp)
 
-	ror	\$$Sigma1[0],$a0
-	ror	\$$Sigma1[1],$a1
+	ror	\$`$Sigma0[2]-$Sigma0[1]`,$a1
+	xor	$e,$a0
 	xor	$g,$a2			# f^g
 
-	xor	$a1,$a0
-	ror	\$`$Sigma1[2]-$Sigma1[1]`,$a1
+	ror	\$`$Sigma1[1]-$Sigma1[0]`,$a0
+	add	$h,$T1			# T1+=h
+	xor	$a,$a1
+
+	add	($Tbl,$round,$SZ),$T1	# T1+=K[round]
 	and	$e,$a2			# (f^g)&e
-	mov	$T1,`$SZ*($i&0xf)`(%rsp)
+	mov	$b,$h
 
-	xor	$a1,$a0			# Sigma1(e)
+	ror	\$`$Sigma0[1]-$Sigma0[0]`,$a1
+	xor	$e,$a0
 	xor	$g,$a2			# Ch(e,f,g)=((f^g)&e)^g
-	add	$h,$T1			# T1+=h
-
-	mov	$a,$h
-	add	$a0,$T1			# T1+=Sigma1(e)
 
+	xor	$c,$h			# b^c
+	xor	$a,$a1
 	add	$a2,$T1			# T1+=Ch(e,f,g)
-	mov	$a,$a0
-	mov	$a,$a1
+	mov	$b,$a2
 
-	ror	\$$Sigma0[0],$h
-	ror	\$$Sigma0[1],$a0
-	mov	$a,$a2
-	add	($Tbl,$round,$SZ),$T1	# T1+=K[round]
+	ror	\$$Sigma1[0],$a0	# Sigma1(e)
+	and	$a,$h			# h=(b^c)&a
+	and	$c,$a2			# b&c
 
-	xor	$a0,$h
-	ror	\$`$Sigma0[2]-$Sigma0[1]`,$a0
-	or	$c,$a1			# a|c
+	ror	\$$Sigma0[0],$a1	# Sigma0(a)
+	add	$a0,$T1			# T1+=Sigma1(e)
+	add	$a2,$h			# h+=b&c (completes +=Maj(a,b,c)
 
-	xor	$a0,$h			# h=Sigma0(a)
-	and	$c,$a2			# a&c
 	add	$T1,$d			# d+=T1
-
-	and	$b,$a1			# (a|c)&b
 	add	$T1,$h			# h+=T1
-
-	or	$a2,$a1			# Maj(a,b,c)=((a|c)&b)|(a&c)
 	lea	1($round),$round	# round++
+	add	$a1,$h			# h+=Sigma0(a)
 
-	add	$a1,$h			# h+=Maj(a,b,c)
 ___
 }
 
@@ -147,32 +141,30 @@ sub ROUND_16_XX()
 
 $code.=<<___;
 	mov	`$SZ*(($i+1)&0xf)`(%rsp),$a0
-	mov	`$SZ*(($i+14)&0xf)`(%rsp),$T1
-
-	mov	$a0,$a2
+	mov	`$SZ*(($i+14)&0xf)`(%rsp),$a1
+	mov	$a0,$T1
+	mov	$a1,$a2
 
+	ror	\$`$sigma0[1]-$sigma0[0]`,$T1
+	xor	$a0,$T1
 	shr	\$$sigma0[2],$a0
-	ror	\$$sigma0[0],$a2
-
-	xor	$a2,$a0
-	ror	\$`$sigma0[1]-$sigma0[0]`,$a2
 
-	xor	$a2,$a0			# sigma0(X[(i+1)&0xf])
-	mov	$T1,$a1
+	ror	\$$sigma0[0],$T1
+	xor	$T1,$a0			# sigma0(X[(i+1)&0xf])
+	mov	`$SZ*(($i+9)&0xf)`(%rsp),$T1
 
-	shr	\$$sigma1[2],$T1
-	ror	\$$sigma1[0],$a1
-
-	xor	$a1,$T1
-	ror	\$`$sigma1[1]-$sigma1[0]`,$a1
-
-	xor	$a1,$T1			# sigma1(X[(i+14)&0xf])
+	ror	\$`$sigma1[1]-$sigma1[0]`,$a2
+	xor	$a1,$a2
+	shr	\$$sigma1[2],$a1
 
+	ror	\$$sigma1[0],$a2
 	add	$a0,$T1
-
-	add	`$SZ*(($i+9)&0xf)`(%rsp),$T1
+	xor	$a2,$a1			# sigma1(X[(i+14)&0xf])
 
 	add	`$SZ*($i&0xf)`(%rsp),$T1
+	mov	$e,$a0
+	add	$a1,$T1
+	mov	$a,$a1
 ___
 	&ROUND_00_15(@_);
 }
@@ -219,6 +211,8 @@ $func:
 ___
 	for($i=0;$i<16;$i++) {
 		$code.="	mov	$SZ*$i($inp),$T1\n";
+		$code.="	mov	@ROT[4],$a0\n";
+		$code.="	mov	@ROT[0],$a1\n";
 		$code.="	bswap	$T1\n";
 		&ROUND_00_15($i,@ROT);
 		unshift(@ROT,pop(@ROT));