Less S390.

ok deraadt@

3 files changed, 0 insertions, 1124 deletions

diff --git a/lib/libcrypto/bn/asm/s390x-gf2m.pl b/lib/libcrypto/bn/asm/s390x-gf2m.pl
deleted file mode 100644
index cd9f13eca29..00000000000
--- a/lib/libcrypto/bn/asm/s390x-gf2m.pl
+++ /dev/null
@@ -1,221 +0,0 @@
-#!/usr/bin/env perl
-#
-# ====================================================================
-# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
-# project. The module is, however, dual licensed under OpenSSL and
-# CRYPTOGAMS licenses depending on where you obtain it. For further
-# details see http://www.openssl.org/~appro/cryptogams/.
-# ====================================================================
-#
-# May 2011
-#
-# The module implements bn_GF2m_mul_2x2 polynomial multiplication used
-# in bn_gf2m.c. It's kind of low-hanging mechanical port from C for
-# the time being... gcc 4.3 appeared to generate poor code, therefore
-# the effort. And indeed, the module delivers 55%-90%(*) improvement
-# on haviest ECDSA verify and ECDH benchmarks for 163- and 571-bit
-# key lengths on z990, 30%-55%(*) - on z10, and 70%-110%(*) - on z196.
-# This is for 64-bit build. In 32-bit "highgprs" case improvement is
-# even higher, for example on z990 it was measured 80%-150%. ECDSA
-# sign is modest 9%-12% faster. Keep in mind that these coefficients
-# are not ones for bn_GF2m_mul_2x2 itself, as not all CPU time is
-# burnt in it...
-#
-# (*)	gcc 4.1 was observed to deliver better results than gcc 4.3,
-#	so that improvement coefficients can vary from one specific
-#	setup to another.
-
-$flavour = shift;
-
-if ($flavour =~ /3[12]/) {
-        $SIZE_T=4;
-        $g="";
-} else {
-        $SIZE_T=8;
-        $g="g";
-}
-
-while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {}
-open STDOUT,">$output";
-
-$stdframe=16*$SIZE_T+4*8;
-
-$rp="%r2";
-$a1="%r3";
-$a0="%r4";
-$b1="%r5";
-$b0="%r6";
-
-$ra="%r14";
-$sp="%r15";
-
-@T=("%r0","%r1");
-@i=("%r12","%r13");
-
-($a1,$a2,$a4,$a8,$a12,$a48)=map("%r$_",(6..11));
-($lo,$hi,$b)=map("%r$_",(3..5)); $a=$lo; $mask=$a8;
-
-$code.=<<___;
-.text
-
-.type	_mul_1x1,\@function
-.align	16
-_mul_1x1:
-	lgr	$a1,$a
-	sllg	$a2,$a,1
-	sllg	$a4,$a,2
-	sllg	$a8,$a,3
-
-	srag	$lo,$a1,63			# broadcast 63rd bit
-	nihh	$a1,0x1fff
-	srag	@i[0],$a2,63			# broadcast 62nd bit
-	nihh	$a2,0x3fff
-	srag	@i[1],$a4,63			# broadcast 61st bit
-	nihh	$a4,0x7fff
-	ngr	$lo,$b
-	ngr	@i[0],$b
-	ngr	@i[1],$b
-
-	lghi	@T[0],0
-	lgr	$a12,$a1
-	stg	@T[0],`$stdframe+0*8`($sp)	# tab[0]=0
-	xgr	$a12,$a2
-	stg	$a1,`$stdframe+1*8`($sp)	# tab[1]=a1
-	 lgr	$a48,$a4
-	stg	$a2,`$stdframe+2*8`($sp)	# tab[2]=a2
-	 xgr	$a48,$a8
-	stg	$a12,`$stdframe+3*8`($sp)	# tab[3]=a1^a2
-	 xgr	$a1,$a4
-
-	stg	$a4,`$stdframe+4*8`($sp)	# tab[4]=a4
-	xgr	$a2,$a4
-	stg	$a1,`$stdframe+5*8`($sp)	# tab[5]=a1^a4
-	xgr	$a12,$a4
-	stg	$a2,`$stdframe+6*8`($sp)	# tab[6]=a2^a4
-	 xgr	$a1,$a48
-	stg	$a12,`$stdframe+7*8`($sp)	# tab[7]=a1^a2^a4
-	 xgr	$a2,$a48
-
-	stg	$a8,`$stdframe+8*8`($sp)	# tab[8]=a8
-	xgr	$a12,$a48
-	stg	$a1,`$stdframe+9*8`($sp)	# tab[9]=a1^a8
-	 xgr	$a1,$a4
-	stg	$a2,`$stdframe+10*8`($sp)	# tab[10]=a2^a8
-	 xgr	$a2,$a4
-	stg	$a12,`$stdframe+11*8`($sp)	# tab[11]=a1^a2^a8
-
-	xgr	$a12,$a4
-	stg	$a48,`$stdframe+12*8`($sp)	# tab[12]=a4^a8
-	 srlg	$hi,$lo,1
-	stg	$a1,`$stdframe+13*8`($sp)	# tab[13]=a1^a4^a8
-	 sllg	$lo,$lo,63
-	stg	$a2,`$stdframe+14*8`($sp)	# tab[14]=a2^a4^a8
-	 srlg	@T[0],@i[0],2
-	stg	$a12,`$stdframe+15*8`($sp)	# tab[15]=a1^a2^a4^a8
-
-	lghi	$mask,`0xf<<3`
-	sllg	$a1,@i[0],62
-	 sllg	@i[0],$b,3
-	srlg	@T[1],@i[1],3
-	 ngr	@i[0],$mask
-	sllg	$a2,@i[1],61
-	 srlg	@i[1],$b,4-3
-	xgr	$hi,@T[0]
-	 ngr	@i[1],$mask
-	xgr	$lo,$a1
-	xgr	$hi,@T[1]
-	xgr	$lo,$a2
-
-	xg	$lo,$stdframe(@i[0],$sp)
-	srlg	@i[0],$b,8-3
-	ngr	@i[0],$mask
-___
-for($n=1;$n<14;$n++) {
-$code.=<<___;
-	lg	@T[1],$stdframe(@i[1],$sp)
-	srlg	@i[1],$b,`($n+2)*4`-3
-	sllg	@T[0],@T[1],`$n*4`
-	ngr	@i[1],$mask
-	srlg	@T[1],@T[1],`64-$n*4`
-	xgr	$lo,@T[0]
-	xgr	$hi,@T[1]
-___
-	push(@i,shift(@i)); push(@T,shift(@T));
-}
-$code.=<<___;
-	lg	@T[1],$stdframe(@i[1],$sp)
-	sllg	@T[0],@T[1],`$n*4`
-	srlg	@T[1],@T[1],`64-$n*4`
-	xgr	$lo,@T[0]
-	xgr	$hi,@T[1]
-
-	lg	@T[0],$stdframe(@i[0],$sp)
-	sllg	@T[1],@T[0],`($n+1)*4`
-	srlg	@T[0],@T[0],`64-($n+1)*4`
-	xgr	$lo,@T[1]
-	xgr	$hi,@T[0]
-
-	br	$ra
-.size	_mul_1x1,.-_mul_1x1
-
-.globl	bn_GF2m_mul_2x2
-.type	bn_GF2m_mul_2x2,\@function
-.align	16
-bn_GF2m_mul_2x2:
-	stm${g}	%r3,%r15,3*$SIZE_T($sp)
-
-	lghi	%r1,-$stdframe-128
-	la	%r0,0($sp)
-	la	$sp,0(%r1,$sp)			# alloca
-	st${g}	%r0,0($sp)			# back chain
-___
-if ($SIZE_T==8) {
-my @r=map("%r$_",(6..9));
-$code.=<<___;
-	bras	$ra,_mul_1x1			# a1·b1
-	stmg	$lo,$hi,16($rp)
-
-	lg	$a,`$stdframe+128+4*$SIZE_T`($sp)
-	lg	$b,`$stdframe+128+6*$SIZE_T`($sp)
-	bras	$ra,_mul_1x1			# a0·b0
-	stmg	$lo,$hi,0($rp)
-
-	lg	$a,`$stdframe+128+3*$SIZE_T`($sp)
-	lg	$b,`$stdframe+128+5*$SIZE_T`($sp)
-	xg	$a,`$stdframe+128+4*$SIZE_T`($sp)
-	xg	$b,`$stdframe+128+6*$SIZE_T`($sp)
-	bras	$ra,_mul_1x1			# (a0+a1)·(b0+b1)
-	lmg	@r[0],@r[3],0($rp)
-
-	xgr	$lo,$hi
-	xgr	$hi,@r[1]
-	xgr	$lo,@r[0]
-	xgr	$hi,@r[2]
-	xgr	$lo,@r[3]	
-	xgr	$hi,@r[3]
-	xgr	$lo,$hi
-	stg	$hi,16($rp)
-	stg	$lo,8($rp)
-___
-} else {
-$code.=<<___;
-	sllg	%r3,%r3,32
-	sllg	%r5,%r5,32
-	or	%r3,%r4
-	or	%r5,%r6
-	bras	$ra,_mul_1x1
-	rllg	$lo,$lo,32
-	rllg	$hi,$hi,32
-	stmg	$lo,$hi,0($rp)
-___
-}
-$code.=<<___;
-	lm${g}	%r6,%r15,`$stdframe+128+6*$SIZE_T`($sp)
-	br	$ra
-.size	bn_GF2m_mul_2x2,.-bn_GF2m_mul_2x2
-.string	"GF(2^m) Multiplication for s390x, CRYPTOGAMS by <appro\@openssl.org>"
-___
-
-$code =~ s/\`([^\`]*)\`/eval($1)/gem;
-print $code;
-close STDOUT;
diff --git a/lib/libcrypto/bn/asm/s390x-mont.pl b/lib/libcrypto/bn/asm/s390x-mont.pl
deleted file mode 100644
index d23251033b0..00000000000
--- a/lib/libcrypto/bn/asm/s390x-mont.pl
+++ /dev/null
@@ -1,225 +0,0 @@
-#!/usr/bin/env perl
-
-# ====================================================================
-# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
-# project. The module is, however, dual licensed under OpenSSL and
-# CRYPTOGAMS licenses depending on where you obtain it. For further
-# details see http://www.openssl.org/~appro/cryptogams/.
-# ====================================================================
-
-# April 2007.
-#
-# Performance improvement over vanilla C code varies from 85% to 45%
-# depending on key length and benchmark. Unfortunately in this context
-# these are not very impressive results [for code that utilizes "wide"
-# 64x64=128-bit multiplication, which is not commonly available to C
-# programmers], at least hand-coded bn_asm.c replacement is known to
-# provide 30-40% better results for longest keys. Well, on a second
-# thought it's not very surprising, because z-CPUs are single-issue
-# and _strictly_ in-order execution, while bn_mul_mont is more or less
-# dependent on CPU ability to pipe-line instructions and have several
-# of them "in-flight" at the same time. I mean while other methods,
-# for example Karatsuba, aim to minimize amount of multiplications at
-# the cost of other operations increase, bn_mul_mont aim to neatly
-# "overlap" multiplications and the other operations [and on most
-# platforms even minimize the amount of the other operations, in
-# particular references to memory]. But it's possible to improve this
-# module performance by implementing dedicated squaring code-path and
-# possibly by unrolling loops...
-
-# January 2009.
-#
-# Reschedule to minimize/avoid Address Generation Interlock hazard,
-# make inner loops counter-based.
-
-$mn0="%r0";
-$num="%r1";
-
-# int bn_mul_mont(
-$rp="%r2";		# BN_ULONG *rp,
-$ap="%r3";		# const BN_ULONG *ap,
-$bp="%r4";		# const BN_ULONG *bp,
-$np="%r5";		# const BN_ULONG *np,
-$n0="%r6";		# const BN_ULONG *n0,
-#$num="160(%r15)"	# int num);
-
-$bi="%r2";	# zaps rp
-$j="%r7";
-
-$ahi="%r8";
-$alo="%r9";
-$nhi="%r10";
-$nlo="%r11";
-$AHI="%r12";
-$NHI="%r13";
-$count="%r14";
-$sp="%r15";
-
-$code.=<<___;
-.text
-.globl	bn_mul_mont
-.type	bn_mul_mont,\@function
-bn_mul_mont:
-	lgf	$num,164($sp)	# pull $num
-	sla	$num,3		# $num to enumerate bytes
-	la	$bp,0($num,$bp)
-
-	stg	%r2,16($sp)
-
-	cghi	$num,16		#
-	lghi	%r2,0		#
-	blr	%r14		# if($num<16) return 0;
-	cghi	$num,128	#
-	bhr	%r14		# if($num>128) return 0;
-
-	stmg	%r3,%r15,24($sp)
-
-	lghi	$rp,-160-8	# leave room for carry bit
-	lcgr	$j,$num		# -$num
-	lgr	%r0,$sp
-	la	$rp,0($rp,$sp)
-	la	$sp,0($j,$rp)	# alloca
-	stg	%r0,0($sp)	# back chain
-
-	sra	$num,3		# restore $num
-	la	$bp,0($j,$bp)	# restore $bp
-	ahi	$num,-1		# adjust $num for inner loop
-	lg	$n0,0($n0)	# pull n0
-
-	lg	$bi,0($bp)
-	lg	$alo,0($ap)
-	mlgr	$ahi,$bi	# ap[0]*bp[0]
-	lgr	$AHI,$ahi
-
-	lgr	$mn0,$alo	# "tp[0]"*n0
-	msgr	$mn0,$n0
-
-	lg	$nlo,0($np)	#
-	mlgr	$nhi,$mn0	# np[0]*m1
-	algr	$nlo,$alo	# +="tp[0]"
-	lghi	$NHI,0
-	alcgr	$NHI,$nhi
-
-	la	$j,8(%r0)	# j=1
-	lr	$count,$num
-
-.align	16
-.L1st:
-	lg	$alo,0($j,$ap)
-	mlgr	$ahi,$bi	# ap[j]*bp[0]
-	algr	$alo,$AHI
-	lghi	$AHI,0
-	alcgr	$AHI,$ahi
-
-	lg	$nlo,0($j,$np)
-	mlgr	$nhi,$mn0	# np[j]*m1
-	algr	$nlo,$NHI
-	lghi	$NHI,0
-	alcgr	$nhi,$NHI	# +="tp[j]"
-	algr	$nlo,$alo
-	alcgr	$NHI,$nhi
-
-	stg	$nlo,160-8($j,$sp)	# tp[j-1]=
-	la	$j,8($j)	# j++
-	brct	$count,.L1st
-
-	algr	$NHI,$AHI
-	lghi	$AHI,0
-	alcgr	$AHI,$AHI	# upmost overflow bit
-	stg	$NHI,160-8($j,$sp)
-	stg	$AHI,160($j,$sp)
-	la	$bp,8($bp)	# bp++
-
-.Louter:
-	lg	$bi,0($bp)	# bp[i]
-	lg	$alo,0($ap)
-	mlgr	$ahi,$bi	# ap[0]*bp[i]
-	alg	$alo,160($sp)	# +=tp[0]
-	lghi	$AHI,0
-	alcgr	$AHI,$ahi
-
-	lgr	$mn0,$alo
-	msgr	$mn0,$n0	# tp[0]*n0
-
-	lg	$nlo,0($np)	# np[0]
-	mlgr	$nhi,$mn0	# np[0]*m1
-	algr	$nlo,$alo	# +="tp[0]"
-	lghi	$NHI,0
-	alcgr	$NHI,$nhi
-
-	la	$j,8(%r0)	# j=1
-	lr	$count,$num
-
-.align	16
-.Linner:
-	lg	$alo,0($j,$ap)
-	mlgr	$ahi,$bi	# ap[j]*bp[i]
-	algr	$alo,$AHI
-	lghi	$AHI,0
-	alcgr	$ahi,$AHI
-	alg	$alo,160($j,$sp)# +=tp[j]
-	alcgr	$AHI,$ahi
-
-	lg	$nlo,0($j,$np)
-	mlgr	$nhi,$mn0	# np[j]*m1
-	algr	$nlo,$NHI
-	lghi	$NHI,0
-	alcgr	$nhi,$NHI
-	algr	$nlo,$alo	# +="tp[j]"
-	alcgr	$NHI,$nhi
-
-	stg	$nlo,160-8($j,$sp)	# tp[j-1]=
-	la	$j,8($j)	# j++
-	brct	$count,.Linner
-
-	algr	$NHI,$AHI
-	lghi	$AHI,0
-	alcgr	$AHI,$AHI
-	alg	$NHI,160($j,$sp)# accumulate previous upmost overflow bit
-	lghi	$ahi,0
-	alcgr	$AHI,$ahi	# new upmost overflow bit
-	stg	$NHI,160-8($j,$sp)
-	stg	$AHI,160($j,$sp)
-
-	la	$bp,8($bp)	# bp++
-	clg	$bp,160+8+32($j,$sp)	# compare to &bp[num]
-	jne	.Louter
-
-	lg	$rp,160+8+16($j,$sp)	# reincarnate rp
-	la	$ap,160($sp)
-	ahi	$num,1		# restore $num, incidentally clears "borrow"
-
-	la	$j,0(%r0)
-	lr	$count,$num
-.Lsub:	lg	$alo,0($j,$ap)
-	slbg	$alo,0($j,$np)
-	stg	$alo,0($j,$rp)
-	la	$j,8($j)
-	brct	$count,.Lsub
-	lghi	$ahi,0
-	slbgr	$AHI,$ahi	# handle upmost carry
-
-	ngr	$ap,$AHI
-	lghi	$np,-1
-	xgr	$np,$AHI
-	ngr	$np,$rp
-	ogr	$ap,$np		# ap=borrow?tp:rp
-
-	la	$j,0(%r0)
-	lgr	$count,$num
-.Lcopy:	lg	$alo,0($j,$ap)	# copy or in-place refresh
-	stg	$j,160($j,$sp)	# zap tp
-	stg	$alo,0($j,$rp)
-	la	$j,8($j)
-	brct	$count,.Lcopy
-
-	la	%r1,160+8+48($j,$sp)
-	lmg	%r6,%r15,0(%r1)
-	lghi	%r2,1		# signal "processed"
-	br	%r14
-.size	bn_mul_mont,.-bn_mul_mont
-.string	"Montgomery Multiplication for s390x, CRYPTOGAMS by <appro\@openssl.org>"
-___
-
-print $code;
-close STDOUT;
diff --git a/lib/libcrypto/bn/asm/s390x.S b/lib/libcrypto/bn/asm/s390x.S
deleted file mode 100755
index 8f45f5d513c..00000000000
--- a/lib/libcrypto/bn/asm/s390x.S
+++ /dev/null
@@ -1,678 +0,0 @@
-.ident "s390x.S, version 1.0"
-// ====================================================================
-// Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
-// project.
-//
-// Rights for redistribution and usage in source and binary forms are
-// granted according to the OpenSSL license. Warranty of any kind is
-// disclaimed.
-// ====================================================================
-
-.text
-
-#define zero	%r0
-
-// BN_ULONG bn_mul_add_words(BN_ULONG *r2,BN_ULONG *r3,int r4,BN_ULONG r5);
-.globl	bn_mul_add_words
-.type	bn_mul_add_words,@function
-.align	4
-bn_mul_add_words:
-	lghi	zero,0		// zero = 0
-	la	%r1,0(%r2)	// put rp aside
-	lghi	%r2,0		// i=0;
-	ltgfr	%r4,%r4
-	bler	%r14		// if (len<=0) return 0;
-
-	stmg	%r6,%r10,48(%r15)
-	lghi	%r8,0		// carry = 0
-	srag	%r10,%r4,2	// cnt=len/4
-	jz	.Loop1_madd
-
-.Loop4_madd:
-	lg	%r7,0(%r2,%r3)	// ap[i]
-	mlgr	%r6,%r5		// *=w
-	algr	%r7,%r8		// +=carry
-	alcgr	%r6,zero
-	alg	%r7,0(%r2,%r1)	// +=rp[i]
-	alcgr	%r6,zero
-	stg	%r7,0(%r2,%r1)	// rp[i]=
-
-	lg	%r9,8(%r2,%r3)
-	mlgr	%r8,%r5
-	algr	%r9,%r6
-	alcgr	%r8,zero
-	alg	%r9,8(%r2,%r1)
-	alcgr	%r8,zero
-	stg	%r9,8(%r2,%r1)
-
-	lg	%r7,16(%r2,%r3)
-	mlgr	%r6,%r5
-	algr	%r7,%r8
-	alcgr	%r6,zero
-	alg	%r7,16(%r2,%r1)
-	alcgr	%r6,zero
-	stg	%r7,16(%r2,%r1)
-
-	lg	%r9,24(%r2,%r3)
-	mlgr	%r8,%r5
-	algr	%r9,%r6
-	alcgr	%r8,zero
-	alg	%r9,24(%r2,%r1)
-	alcgr	%r8,zero
-	stg	%r9,24(%r2,%r1)
-
-	la	%r2,32(%r2)	// i+=4
-	brct	%r10,.Loop4_madd
-
-	lghi	%r10,3
-	nr	%r4,%r10	// cnt=len%4
-	jz	.Lend_madd
-
-.Loop1_madd:
-	lg	%r7,0(%r2,%r3)	// ap[i]
-	mlgr	%r6,%r5		// *=w
-	algr	%r7,%r8		// +=carry
-	alcgr	%r6,zero
-	alg	%r7,0(%r2,%r1)	// +=rp[i]
-	alcgr	%r6,zero
-	stg	%r7,0(%r2,%r1)	// rp[i]=
-
-	lgr	%r8,%r6
-	la	%r2,8(%r2)	// i++
-	brct	%r4,.Loop1_madd
-
-.Lend_madd:
-	lgr	%r2,%r8
-	lmg	%r6,%r10,48(%r15)
-	br	%r14
-.size	bn_mul_add_words,.-bn_mul_add_words
-
-// BN_ULONG bn_mul_words(BN_ULONG *r2,BN_ULONG *r3,int r4,BN_ULONG r5);
-.globl	bn_mul_words
-.type	bn_mul_words,@function
-.align	4
-bn_mul_words:
-	lghi	zero,0		// zero = 0
-	la	%r1,0(%r2)	// put rp aside
-	lghi	%r2,0		// i=0;
-	ltgfr	%r4,%r4
-	bler	%r14		// if (len<=0) return 0;
-
-	stmg	%r6,%r10,48(%r15)
-	lghi	%r8,0		// carry = 0
-	srag	%r10,%r4,2	// cnt=len/4
-	jz	.Loop1_mul
-
-.Loop4_mul:
-	lg	%r7,0(%r2,%r3)	// ap[i]
-	mlgr	%r6,%r5		// *=w
-	algr	%r7,%r8		// +=carry
-	alcgr	%r6,zero
-	stg	%r7,0(%r2,%r1)	// rp[i]=
-
-	lg	%r9,8(%r2,%r3)
-	mlgr	%r8,%r5
-	algr	%r9,%r6
-	alcgr	%r8,zero
-	stg	%r9,8(%r2,%r1)
-
-	lg	%r7,16(%r2,%r3)
-	mlgr	%r6,%r5
-	algr	%r7,%r8
-	alcgr	%r6,zero
-	stg	%r7,16(%r2,%r1)
-
-	lg	%r9,24(%r2,%r3)
-	mlgr	%r8,%r5
-	algr	%r9,%r6
-	alcgr	%r8,zero
-	stg	%r9,24(%r2,%r1)
-
-	la	%r2,32(%r2)	// i+=4
-	brct	%r10,.Loop4_mul
-
-	lghi	%r10,3
-	nr	%r4,%r10	// cnt=len%4
-	jz	.Lend_mul
-
-.Loop1_mul:
-	lg	%r7,0(%r2,%r3)	// ap[i]
-	mlgr	%r6,%r5		// *=w
-	algr	%r7,%r8		// +=carry
-	alcgr	%r6,zero
-	stg	%r7,0(%r2,%r1)	// rp[i]=
-
-	lgr	%r8,%r6
-	la	%r2,8(%r2)	// i++
-	brct	%r4,.Loop1_mul
-
-.Lend_mul:
-	lgr	%r2,%r8
-	lmg	%r6,%r10,48(%r15)
-	br	%r14
-.size	bn_mul_words,.-bn_mul_words
-
-// void bn_sqr_words(BN_ULONG *r2,BN_ULONG *r2,int r4)
-.globl	bn_sqr_words
-.type	bn_sqr_words,@function
-.align	4
-bn_sqr_words:
-	ltgfr	%r4,%r4
-	bler	%r14
-
-	stmg	%r6,%r7,48(%r15)
-	srag	%r1,%r4,2	// cnt=len/4
-	jz	.Loop1_sqr
-
-.Loop4_sqr:
-	lg	%r7,0(%r3)
-	mlgr	%r6,%r7
-	stg	%r7,0(%r2)
-	stg	%r6,8(%r2)
-
-	lg	%r7,8(%r3)
-	mlgr	%r6,%r7
-	stg	%r7,16(%r2)
-	stg	%r6,24(%r2)
-
-	lg	%r7,16(%r3)
-	mlgr	%r6,%r7
-	stg	%r7,32(%r2)
-	stg	%r6,40(%r2)
-
-	lg	%r7,24(%r3)
-	mlgr	%r6,%r7
-	stg	%r7,48(%r2)
-	stg	%r6,56(%r2)
-
-	la	%r3,32(%r3)
-	la	%r2,64(%r2)
-	brct	%r1,.Loop4_sqr
-
-	lghi	%r1,3
-	nr	%r4,%r1		// cnt=len%4
-	jz	.Lend_sqr
-
-.Loop1_sqr:
-	lg	%r7,0(%r3)
-	mlgr	%r6,%r7
-	stg	%r7,0(%r2)
-	stg	%r6,8(%r2)
-
-	la	%r3,8(%r3)
-	la	%r2,16(%r2)
-	brct	%r4,.Loop1_sqr
-
-.Lend_sqr:
-	lmg	%r6,%r7,48(%r15)
-	br	%r14
-.size	bn_sqr_words,.-bn_sqr_words
-
-// BN_ULONG bn_div_words(BN_ULONG h,BN_ULONG l,BN_ULONG d);
-.globl	bn_div_words
-.type	bn_div_words,@function
-.align	4
-bn_div_words:
-	dlgr	%r2,%r4
-	lgr	%r2,%r3
-	br	%r14
-.size	bn_div_words,.-bn_div_words
-
-// BN_ULONG bn_add_words(BN_ULONG *r2,BN_ULONG *r3,BN_ULONG *r4,int r5);
-.globl	bn_add_words
-.type	bn_add_words,@function
-.align	4
-bn_add_words:
-	la	%r1,0(%r2)	// put rp aside
-	lghi	%r2,0		// i=0
-	ltgfr	%r5,%r5
-	bler	%r14		// if (len<=0) return 0;
-
-	stg	%r6,48(%r15)
-	lghi	%r6,3
-	nr	%r6,%r5		// len%4
-	sra	%r5,2		// len/4, use sra because it sets condition code
-	jz	.Loop1_add	// carry is incidentally cleared if branch taken
-	algr	%r2,%r2		// clear carry
-
-.Loop4_add:
-	lg	%r0,0(%r2,%r3)
-	alcg	%r0,0(%r2,%r4)
-	stg	%r0,0(%r2,%r1)
-	lg	%r0,8(%r2,%r3)
-	alcg	%r0,8(%r2,%r4)
-	stg	%r0,8(%r2,%r1)
-	lg	%r0,16(%r2,%r3)
-	alcg	%r0,16(%r2,%r4)
-	stg	%r0,16(%r2,%r1)
-	lg	%r0,24(%r2,%r3)
-	alcg	%r0,24(%r2,%r4)
-	stg	%r0,24(%r2,%r1)
-
-	la	%r2,32(%r2)	// i+=4
-	brct	%r5,.Loop4_add
-
-	la	%r6,1(%r6)	// see if len%4 is zero ...
-	brct	%r6,.Loop1_add	// without touching condition code:-)
-
-.Lexit_add:
-	lghi	%r2,0
-	alcgr	%r2,%r2
-	lg	%r6,48(%r15)
-	br	%r14
-
-.Loop1_add:
-	lg	%r0,0(%r2,%r3)
-	alcg	%r0,0(%r2,%r4)
-	stg	%r0,0(%r2,%r1)
-
-	la	%r2,8(%r2)	// i++
-	brct	%r6,.Loop1_add
-
-	j	.Lexit_add
-.size	bn_add_words,.-bn_add_words
-
-// BN_ULONG bn_sub_words(BN_ULONG *r2,BN_ULONG *r3,BN_ULONG *r4,int r5);
-.globl	bn_sub_words
-.type	bn_sub_words,@function
-.align	4
-bn_sub_words:
-	la	%r1,0(%r2)	// put rp aside
-	lghi	%r2,0		// i=0
-	ltgfr	%r5,%r5
-	bler	%r14		// if (len<=0) return 0;
-
-	stg	%r6,48(%r15)
-	lghi	%r6,3
-	nr	%r6,%r5		// len%4
-	sra	%r5,2		// len/4, use sra because it sets condition code
-	jnz	.Loop4_sub	// borrow is incidentally cleared if branch taken
-	slgr	%r2,%r2		// clear borrow
-
-.Loop1_sub:
-	lg	%r0,0(%r2,%r3)
-	slbg	%r0,0(%r2,%r4)
-	stg	%r0,0(%r2,%r1)
-
-	la	%r2,8(%r2)	// i++
-	brct	%r6,.Loop1_sub
-	j	.Lexit_sub
-
-.Loop4_sub:
-	lg	%r0,0(%r2,%r3)
-	slbg	%r0,0(%r2,%r4)
-	stg	%r0,0(%r2,%r1)
-	lg	%r0,8(%r2,%r3)
-	slbg	%r0,8(%r2,%r4)
-	stg	%r0,8(%r2,%r1)
-	lg	%r0,16(%r2,%r3)
-	slbg	%r0,16(%r2,%r4)
-	stg	%r0,16(%r2,%r1)
-	lg	%r0,24(%r2,%r3)
-	slbg	%r0,24(%r2,%r4)
-	stg	%r0,24(%r2,%r1)
-
-	la	%r2,32(%r2)	// i+=4
-	brct	%r5,.Loop4_sub
-
-	la	%r6,1(%r6)	// see if len%4 is zero ...
-	brct	%r6,.Loop1_sub	// without touching condition code:-)
-
-.Lexit_sub:
-	lghi	%r2,0
-	slbgr	%r2,%r2
-	lcgr	%r2,%r2
-	lg	%r6,48(%r15)
-	br	%r14
-.size	bn_sub_words,.-bn_sub_words
-
-#define c1	%r1
-#define c2	%r5
-#define c3	%r8
-
-#define mul_add_c(ai,bi,c1,c2,c3)	\
-	lg	%r7,ai*8(%r3);		\
-	mlg	%r6,bi*8(%r4);		\
-	algr	c1,%r7;			\
-	alcgr	c2,%r6;			\
-	alcgr	c3,zero
-
-// void bn_mul_comba8(BN_ULONG *r2,BN_ULONG *r3,BN_ULONG *r4);
-.globl	bn_mul_comba8
-.type	bn_mul_comba8,@function
-.align	4
-bn_mul_comba8:
-	stmg	%r6,%r8,48(%r15)
-
-	lghi	c1,0
-	lghi	c2,0
-	lghi	c3,0
-	lghi	zero,0
-
-	mul_add_c(0,0,c1,c2,c3);
-	stg	c1,0*8(%r2)
-	lghi	c1,0
-
-	mul_add_c(0,1,c2,c3,c1);
-	mul_add_c(1,0,c2,c3,c1);
-	stg	c2,1*8(%r2)
-	lghi	c2,0
-
-	mul_add_c(2,0,c3,c1,c2);
-	mul_add_c(1,1,c3,c1,c2);
-	mul_add_c(0,2,c3,c1,c2);
-	stg	c3,2*8(%r2)
-	lghi	c3,0
-
-	mul_add_c(0,3,c1,c2,c3);
-	mul_add_c(1,2,c1,c2,c3);
-	mul_add_c(2,1,c1,c2,c3);
-	mul_add_c(3,0,c1,c2,c3);
-	stg	c1,3*8(%r2)
-	lghi	c1,0
-
-	mul_add_c(4,0,c2,c3,c1);
-	mul_add_c(3,1,c2,c3,c1);
-	mul_add_c(2,2,c2,c3,c1);
-	mul_add_c(1,3,c2,c3,c1);
-	mul_add_c(0,4,c2,c3,c1);
-	stg	c2,4*8(%r2)
-	lghi	c2,0
-
-	mul_add_c(0,5,c3,c1,c2);
-	mul_add_c(1,4,c3,c1,c2);
-	mul_add_c(2,3,c3,c1,c2);
-	mul_add_c(3,2,c3,c1,c2);
-	mul_add_c(4,1,c3,c1,c2);
-	mul_add_c(5,0,c3,c1,c2);
-	stg	c3,5*8(%r2)
-	lghi	c3,0
-
-	mul_add_c(6,0,c1,c2,c3);
-	mul_add_c(5,1,c1,c2,c3);
-	mul_add_c(4,2,c1,c2,c3);
-	mul_add_c(3,3,c1,c2,c3);
-	mul_add_c(2,4,c1,c2,c3);
-	mul_add_c(1,5,c1,c2,c3);
-	mul_add_c(0,6,c1,c2,c3);
-	stg	c1,6*8(%r2)
-	lghi	c1,0
-
-	mul_add_c(0,7,c2,c3,c1);
-	mul_add_c(1,6,c2,c3,c1);
-	mul_add_c(2,5,c2,c3,c1);
-	mul_add_c(3,4,c2,c3,c1);
-	mul_add_c(4,3,c2,c3,c1);
-	mul_add_c(5,2,c2,c3,c1);
-	mul_add_c(6,1,c2,c3,c1);
-	mul_add_c(7,0,c2,c3,c1);
-	stg	c2,7*8(%r2)
-	lghi	c2,0
-
-	mul_add_c(7,1,c3,c1,c2);
-	mul_add_c(6,2,c3,c1,c2);
-	mul_add_c(5,3,c3,c1,c2);
-	mul_add_c(4,4,c3,c1,c2);
-	mul_add_c(3,5,c3,c1,c2);
-	mul_add_c(2,6,c3,c1,c2);
-	mul_add_c(1,7,c3,c1,c2);
-	stg	c3,8*8(%r2)
-	lghi	c3,0
-
-	mul_add_c(2,7,c1,c2,c3);
-	mul_add_c(3,6,c1,c2,c3);
-	mul_add_c(4,5,c1,c2,c3);
-	mul_add_c(5,4,c1,c2,c3);
-	mul_add_c(6,3,c1,c2,c3);
-	mul_add_c(7,2,c1,c2,c3);
-	stg	c1,9*8(%r2)
-	lghi	c1,0
-
-	mul_add_c(7,3,c2,c3,c1);
-	mul_add_c(6,4,c2,c3,c1);
-	mul_add_c(5,5,c2,c3,c1);
-	mul_add_c(4,6,c2,c3,c1);
-	mul_add_c(3,7,c2,c3,c1);
-	stg	c2,10*8(%r2)
-	lghi	c2,0
-
-	mul_add_c(4,7,c3,c1,c2);
-	mul_add_c(5,6,c3,c1,c2);
-	mul_add_c(6,5,c3,c1,c2);
-	mul_add_c(7,4,c3,c1,c2);
-	stg	c3,11*8(%r2)
-	lghi	c3,0
-
-	mul_add_c(7,5,c1,c2,c3);
-	mul_add_c(6,6,c1,c2,c3);
-	mul_add_c(5,7,c1,c2,c3);
-	stg	c1,12*8(%r2)
-	lghi	c1,0
-
-
-	mul_add_c(6,7,c2,c3,c1);
-	mul_add_c(7,6,c2,c3,c1);
-	stg	c2,13*8(%r2)
-	lghi	c2,0
-
-	mul_add_c(7,7,c3,c1,c2);
-	stg	c3,14*8(%r2)
-	stg	c1,15*8(%r2)
-
-	lmg	%r6,%r8,48(%r15)
-	br	%r14
-.size	bn_mul_comba8,.-bn_mul_comba8
-
-// void bn_mul_comba4(BN_ULONG *r2,BN_ULONG *r3,BN_ULONG *r4);
-.globl	bn_mul_comba4
-.type	bn_mul_comba4,@function
-.align	4
-bn_mul_comba4:
-	stmg	%r6,%r8,48(%r15)
-
-	lghi	c1,0
-	lghi	c2,0
-	lghi	c3,0
-	lghi	zero,0
-
-	mul_add_c(0,0,c1,c2,c3);
-	stg	c1,0*8(%r3)
-	lghi	c1,0
-
-	mul_add_c(0,1,c2,c3,c1);
-	mul_add_c(1,0,c2,c3,c1);
-	stg	c2,1*8(%r2)
-	lghi	c2,0
-
-	mul_add_c(2,0,c3,c1,c2);
-	mul_add_c(1,1,c3,c1,c2);
-	mul_add_c(0,2,c3,c1,c2);
-	stg	c3,2*8(%r2)
-	lghi	c3,0
-
-	mul_add_c(0,3,c1,c2,c3);
-	mul_add_c(1,2,c1,c2,c3);
-	mul_add_c(2,1,c1,c2,c3);
-	mul_add_c(3,0,c1,c2,c3);
-	stg	c1,3*8(%r2)
-	lghi	c1,0
-
-	mul_add_c(3,1,c2,c3,c1);
-	mul_add_c(2,2,c2,c3,c1);
-	mul_add_c(1,3,c2,c3,c1);
-	stg	c2,4*8(%r2)
-	lghi	c2,0
-
-	mul_add_c(2,3,c3,c1,c2);
-	mul_add_c(3,2,c3,c1,c2);
-	stg	c3,5*8(%r2)
-	lghi	c3,0
-
-	mul_add_c(3,3,c1,c2,c3);
-	stg	c1,6*8(%r2)
-	stg	c2,7*8(%r2)
-
-	stmg	%r6,%r8,48(%r15)
-	br	%r14
-.size	bn_mul_comba4,.-bn_mul_comba4
-
-#define sqr_add_c(ai,c1,c2,c3)		\
-	lg	%r7,ai*8(%r3);		\
-	mlgr	%r6,%r7;		\
-	algr	c1,%r7;			\
-	alcgr	c2,%r6;			\
-	alcgr	c3,zero
-
-#define sqr_add_c2(ai,aj,c1,c2,c3)	\
-	lg	%r7,ai*8(%r3);		\
-	mlg	%r6,aj*8(%r3);		\
-	algr	c1,%r7;			\
-	alcgr	c2,%r6;			\
-	alcgr	c3,zero;		\
-	algr	c1,%r7;			\
-	alcgr	c2,%r6;			\
-	alcgr	c3,zero
-
-// void bn_sqr_comba8(BN_ULONG *r2,BN_ULONG *r3);
-.globl	bn_sqr_comba8
-.type	bn_sqr_comba8,@function
-.align	4
-bn_sqr_comba8:
-	stmg	%r6,%r8,48(%r15)
-
-	lghi	c1,0
-	lghi	c2,0
-	lghi	c3,0
-	lghi	zero,0
-
-	sqr_add_c(0,c1,c2,c3);
-	stg	c1,0*8(%r2)
-	lghi	c1,0
-
-	sqr_add_c2(1,0,c2,c3,c1);
-	stg	c2,1*8(%r2)
-	lghi	c2,0
-
-	sqr_add_c(1,c3,c1,c2);
-	sqr_add_c2(2,0,c3,c1,c2);
-	stg	c3,2*8(%r2)
-	lghi	c3,0
-
-	sqr_add_c2(3,0,c1,c2,c3);
-	sqr_add_c2(2,1,c1,c2,c3);
-	stg	c1,3*8(%r2)
-	lghi	c1,0
-
-	sqr_add_c(2,c2,c3,c1);
-	sqr_add_c2(3,1,c2,c3,c1);
-	sqr_add_c2(4,0,c2,c3,c1);
-	stg	c2,4*8(%r2)
-	lghi	c2,0
-
-	sqr_add_c2(5,0,c3,c1,c2);
-	sqr_add_c2(4,1,c3,c1,c2);
-	sqr_add_c2(3,2,c3,c1,c2);
-	stg	c3,5*8(%r2)
-	lghi	c3,0
-
-	sqr_add_c(3,c1,c2,c3);
-	sqr_add_c2(4,2,c1,c2,c3);
-	sqr_add_c2(5,1,c1,c2,c3);
-	sqr_add_c2(6,0,c1,c2,c3);
-	stg	c1,6*8(%r2)
-	lghi	c1,0
-
-	sqr_add_c2(7,0,c2,c3,c1);
-	sqr_add_c2(6,1,c2,c3,c1);
-	sqr_add_c2(5,2,c2,c3,c1);
-	sqr_add_c2(4,3,c2,c3,c1);
-	stg	c2,7*8(%r2)
-	lghi	c2,0
-
-	sqr_add_c(4,c3,c1,c2);
-	sqr_add_c2(5,3,c3,c1,c2);
-	sqr_add_c2(6,2,c3,c1,c2);
-	sqr_add_c2(7,1,c3,c1,c2);
-	stg	c3,8*8(%r2)
-	lghi	c3,0
-
-	sqr_add_c2(7,2,c1,c2,c3);
-	sqr_add_c2(6,3,c1,c2,c3);
-	sqr_add_c2(5,4,c1,c2,c3);
-	stg	c1,9*8(%r2)
-	lghi	c1,0
-
-	sqr_add_c(5,c2,c3,c1);
-	sqr_add_c2(6,4,c2,c3,c1);
-	sqr_add_c2(7,3,c2,c3,c1);
-	stg	c2,10*8(%r2)
-	lghi	c2,0
-
-	sqr_add_c2(7,4,c3,c1,c2);
-	sqr_add_c2(6,5,c3,c1,c2);
-	stg	c3,11*8(%r2)
-	lghi	c3,0
-
-	sqr_add_c(6,c1,c2,c3);
-	sqr_add_c2(7,5,c1,c2,c3);
-	stg	c1,12*8(%r2)
-	lghi	c1,0
-
-	sqr_add_c2(7,6,c2,c3,c1);
-	stg	c2,13*8(%r2)
-	lghi	c2,0
-
-	sqr_add_c(7,c3,c1,c2);
-	stg	c3,14*8(%r2)
-	stg	c1,15*8(%r2)
-
-	lmg	%r6,%r8,48(%r15)
-	br	%r14
-.size	bn_sqr_comba8,.-bn_sqr_comba8
-
-// void bn_sqr_comba4(BN_ULONG *r2,BN_ULONG *r3);
-.globl bn_sqr_comba4
-.type	bn_sqr_comba4,@function
-.align	4
-bn_sqr_comba4:
-	stmg	%r6,%r8,48(%r15)
-
-	lghi	c1,0
-	lghi	c2,0
-	lghi	c3,0
-	lghi	zero,0
-
-	sqr_add_c(0,c1,c2,c3);
-	stg	c1,0*8(%r2)
-	lghi	c1,0
-
-	sqr_add_c2(1,0,c2,c3,c1);
-	stg	c2,1*8(%r2)
-	lghi	c2,0
-
-	sqr_add_c(1,c3,c1,c2);
-	sqr_add_c2(2,0,c3,c1,c2);
-	stg	c3,2*8(%r2)
-	lghi	c3,0
-
-	sqr_add_c2(3,0,c1,c2,c3);
-	sqr_add_c2(2,1,c1,c2,c3);
-	stg	c1,3*8(%r2)
-	lghi	c1,0
-
-	sqr_add_c(2,c2,c3,c1);
-	sqr_add_c2(3,1,c2,c3,c1);
-	stg	c2,4*8(%r2)
-	lghi	c2,0
-
-	sqr_add_c2(3,2,c3,c1,c2);
-	stg	c3,5*8(%r2)
-	lghi	c3,0
-
-	sqr_add_c(3,c1,c2,c3);
-	stg	c1,6*8(%r2)
-	stg	c2,7*8(%r2)
-
-	lmg	%r6,%r8,48(%r15)
-	br	%r14
-.size	bn_sqr_comba4,.-bn_sqr_comba4