SHA1 assembler show off: minor performance updates and new modules for

forgotten CPUs.

5 files changed, 960 insertions, 98 deletions

diff --git a/crypto/sha/asm/sha1-alpha.pl b/crypto/sha/asm/sha1-alpha.pl
new file mode 100644
index 0000000000..dd9b43b697
--- /dev/null
+++ b/crypto/sha/asm/sha1-alpha.pl
@@ -0,0 +1,314 @@
+#!/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]. 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=<<___;
+#include <asm.h>
+#include <regdef.h>
+
+.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
+___
+print $code;
+close STDOUT;
diff --git a/crypto/sha/asm/sha1-armv4-large.pl b/crypto/sha/asm/sha1-armv4-large.pl
index c5c4480f94..f99606a32e 100644
--- a/crypto/sha/asm/sha1-armv4-large.pl
+++ b/crypto/sha/asm/sha1-armv4-large.pl
@@ -86,8 +86,8 @@ $code.=<<___;
 	ldr	$t3,[$Xi,#2*4]
 	add	$e,$K,$e,ror#2			@ E+=K_xx_xx
 	eor	$t0,$t0,$t1
+	eor	$t2,$t2,$t3
 	eor	$t0,$t0,$t2
-	eor	$t0,$t0,$t3
 	add	$e,$e,$a,ror#27			@ E+=ROR(A,27)
 ___
 $code.=<<___ if (!defined($flag));
@@ -131,6 +131,15 @@ ___
 
 sub BODY_40_59 {
 my ($a,$b,$c,$d,$e)=@_;
+if (1) {
+	&Xupdate(@_);
+$code.=<<___;
+	and	$t2,$c,$d
+	and	$t1,$b,$t1,ror#2
+	add	$e,$e,$t2,ror#2
+	add	$e,$e,$t1			@ E+=F_40_59(B,C,D)
+___
+} else {
 	&Xupdate(@_,1);
 $code.=<<___;
 	and	$t1,$b,$c,ror#2
@@ -140,6 +149,7 @@ $code.=<<___;
 	add	$e,$e,$t1			@ E+=F_40_59(B,C,D)
 ___
 }
+}
 
 $code=<<___;
 .text
diff --git a/crypto/sha/asm/sha1-ia64.pl b/crypto/sha/asm/sha1-ia64.pl
index 51c4f47ecb..db28f0805a 100644
--- a/crypto/sha/asm/sha1-ia64.pl
+++ b/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/crypto/sha/asm/sha1-mips.pl b/crypto/sha/asm/sha1-mips.pl
new file mode 100644
index 0000000000..a2f7aaddc1
--- /dev/null
+++ b/crypto/sha/asm/sha1-mips.pl
@@ -0,0 +1,281 @@
+#!/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...
+#
+# The code is somewhat IRIX-centric, i.e. is likely to require minor
+# adaptations for other OSes...
+
+for (@ARGV) {   $big_endian=1 if (/\-DB_ENDIAN/);
+                $big_endian=0 if (/\-DL_ENDIAN/);   }
+if (!defined($big_endian))
+            {   $big_endian=(unpack('L',pack('N',1))==1);   }
+
+# offsets of the Most and Least Significant Bytes
+$MSB=$big_endian?0:3;
+$LSB=3&~$MSB;
+
+@X=(	"\$8",	"\$9",	"\$10",	"\$11",	"\$12",	"\$13",	"\$14",	"\$15",
+	"\$16",	"\$17",	"\$18",	"\$19",	"\$20",	"\$21",	"\$22",	"\$23");
+$ctx="\$4";	# a0
+$inp="\$5";	# a1
+$num="\$6";	# a2
+$A="\$1";
+$B="\$2";
+$C="\$3";
+$D="\$7";
+$E="\$24";	@V=($A,$B,$C,$D,$E);
+$t0="\$25";	# jp,t9
+$t1="\$28";	# gp
+$t2="\$30";	# fp,s8
+$K="\$31";	# ra
+
+$FRAMESIZE=16;
+
+sub BODY_00_14 {
+my ($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	@X[$i],$t1
+	or	@X[$i],$t2
+___
+$code.=<<___;
+	 lwl	@X[$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
+___
+}
+
+$code=<<___;
+#include <asm.h>
+#include <regdef.h>
+
+.text
+
+.set	noat
+.set	noreorder
+.align	5
+.globl	sha1_block_data_order
+.ent	sha1_block_data_order
+sha1_block_data_order:
+	.frame	sp,$FRAMESIZE*SZREG,zero
+	.mask	0xd0ff0000,-$FRAMESIZE*SZREG
+	.set	noreorder
+	PTR_SUB	sp,$FRAMESIZE*SZREG
+	REG_S	\$31,($FRAMESIZE-1)*SZREG(sp)
+	REG_S	\$30,($FRAMESIZE-2)*SZREG(sp)
+	REG_S	\$28,($FRAMESIZE-3)*SZREG(sp)
+	REG_S	\$23,($FRAMESIZE-4)*SZREG(sp)
+	REG_S	\$22,($FRAMESIZE-5)*SZREG(sp)
+	REG_S	\$21,($FRAMESIZE-6)*SZREG(sp)
+	REG_S	\$20,($FRAMESIZE-7)*SZREG(sp)
+	REG_S	\$19,($FRAMESIZE-8)*SZREG(sp)
+	REG_S	\$18,($FRAMESIZE-9)*SZREG(sp)
+	REG_S	\$17,($FRAMESIZE-10)*SZREG(sp)
+	REG_S	\$16,($FRAMESIZE-11)*SZREG(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.=<<___;
+	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]
+	PTR_SUB	$num,1
+	sw	$C,8($ctx)
+	sw	$D,12($ctx)
+	sw	$E,16($ctx)
+	.set	noreorder
+	bnez	$num,.Loop
+	PTR_ADD	$inp,64
+
+	.set	noreorder
+	REG_L	\$31,($FRAMESIZE-1)*SZREG(sp)
+	REG_L	\$30,($FRAMESIZE-2)*SZREG(sp)
+	REG_L	\$28,($FRAMESIZE-3)*SZREG(sp)
+	REG_L	\$23,($FRAMESIZE-4)*SZREG(sp)
+	REG_L	\$22,($FRAMESIZE-5)*SZREG(sp)
+	REG_L	\$21,($FRAMESIZE-6)*SZREG(sp)
+	REG_L	\$20,($FRAMESIZE-7)*SZREG(sp)
+	REG_L	\$19,($FRAMESIZE-8)*SZREG(sp)
+	REG_L	\$18,($FRAMESIZE-9)*SZREG(sp)
+	REG_L	\$17,($FRAMESIZE-10)*SZREG(sp)
+	REG_L	\$16,($FRAMESIZE-11)*SZREG(sp)
+	jr	ra
+	PTR_ADD	sp,$FRAMESIZE*SZREG
+.end	sha1_block_data_order
+___
+print $code;
+close STDOUT;
diff --git a/crypto/sha/asm/sha1-parisc.pl b/crypto/sha/asm/sha1-parisc.pl
new file mode 100644
index 0000000000..6d7bf495b2
--- /dev/null
+++ b/crypto/sha/asm/sha1-parisc.pl
@@ -0,0 +1,259 @@
+#!/usr/bin/env perl
+
+# ====================================================================
+# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
+# project. The module is, however, dual licensed under OpenSSL and
+# CRYPTOGAMS licenses depending on where you obtain it. For further
+# details see http://www.openssl.org/~appro/cryptogams/.
+# ====================================================================
+
+# SHA1 block procedure for PA-RISC.
+
+# June 2009.
+#
+# On PA-7100LC performance is >30% better than gcc 3.2 generated code
+# for aligned input and >50% better for unaligned. Compared to vendor
+# compiler on PA-8600 it's almost 60% faster in 64-bit build and just
+# few percent faster in 32-bit one (this for aligned input, data for
+# unaligned input is not available).
+#
+# Special thanks to polarhome.com for providing HP-UX account.
+
+$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<