[thirdparty/openssl.git] / crypto / sha / asm / sha1-ia64.pl

#! /usr/bin/env perl
# Copyright 2004-2016 The OpenSSL Project Authors. All Rights Reserved.
#
# Licensed under the OpenSSL license (the "License").  You may not use
# this file except in compliance with the License.  You can obtain a copy
# in the file LICENSE in the source distribution or at
# https://www.openssl.org/source/license.html

#
# ====================================================================
# 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/.
# ====================================================================
#
# Eternal question is what's wrong with compiler generated code? The
# trick is that it's possible to reduce the number of shifts required
# to perform rotations by maintaining copy of 32-bit value in upper
# bits of 64-bit register. Just follow mux2 and shrp instructions...
# Performance under big-endian OS such as HP-UX is 179MBps*1GHz, which
# is >50% better than HP C and >2x better than gcc.

$output = pop;

$code=<<___;
.ident  \"sha1-ia64.s, version 1.3\"
.ident  \"IA-64 ISA artwork by Andy Polyakov <appro\@fy.chalmers.se>\"
.explicit

___


if ($^O eq "hpux") {
    $ADDP="addp4";
    for (@ARGV) { $ADDP="add" if (/[\+DD|\-mlp]64/); }
} else { $ADDP="add"; }

#$human=1;
if ($human) {	# useful for visual code auditing...
	($A,$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, $K_60_79) =
	    (	"K_00_19","K_20_39","K_40_59","K_60_79"	);
	@X= (	"X0", "X1", "X2", "X3", "X4", "X5", "X6", "X7",
		"X8", "X9","X10","X11","X12","X13","X14","X15"	);
}
else {
	($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", "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;
my	($i,$a,$b,$c,$d,$e)=@_;
my	$j=$i+1;
my	$Xn=@X[$j%16];

$code.=<<___ if ($i==0);
{ .mmi;	ld1	$X[$i]=[inp],2		    // MSB
	ld1	tmp2=[tmp3],2		};;
{ .mmi;	ld1	tmp0=[inp],2
	ld1	tmp4=[tmp3],2		    // LSB
	dep	$X[$i]=$X[$i],tmp2,8,8	};;
___
if ($i<15) {
	$code.=<<___;
{ .mmi;	ld1	$Xn=[inp],2		    // forward Xload
	nop.m	0x0
	dep	tmp1=tmp0,tmp4,8,8	};;
{ .mmi;	ld1	tmp2=[tmp3],2		    // forward Xload
	and	tmp4=$c,$b
	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;	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		    // forward Xload
	add	$e=$e,tmp4		    // e+=F_00_19(b,c,d)
	shrp	$b=tmp6,tmp6,2		}   // b=ROTATE(b,30)
{ .mmi;	ld1	tmp4=[tmp3],2		    // forward Xload
	or	tmp5=tmp1,tmp5		    // ROTATE(a,5)
	mux2	tmp6=$a,0x44		};; // see b in next iteration
{ .mii;	add	$e=$e,tmp5		    // e+=ROTATE(a,5)
	dep	$Xn=$Xn,tmp2,8,8	    // forward Xload
	mux2	$X[$i]=$X[$i],0x44	} //;;

___
	}
else	{
	$code.=<<___;
{ .mii;	and	tmp3=$c,$b
	dep	tmp1=tmp0,tmp4,8,8;;
	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;	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	$Xn=$Xn,$X[($j+2)%16]	    // forward Xupdate
	xor	tmp3=$X[($j+8)%16],$X[($j+13)%16] // forward Xupdate
	nop.i	0			};;
{ .mmi;	add	$e=$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	$e=$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	};;

___
	}
}

sub BODY_16_19 {
local	*code=shift;
my	($i,$a,$b,$c,$d,$e)=@_;
my	$j=$i+1;
my	$Xn=@X[$j%16];

$code.=<<___;
{ .mib;	add	$e=$e,$K_00_19		    // e+=K_00_19
	dep.z	tmp5=$a,5,27		}   // a<<5
{ .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	$Xn=$Xn,$X[($j+2)%16]	    // forward Xupdate
	xor	tmp3=$X[($j+8)%16],$X[($j+13)%16]	// forward Xupdate
	nop.i	0			};;
{ .mmi;	add	$e=$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	$e=$e,tmp1		    // e+=ROTATE(a,5)
	shrp	$Xn=$Xn,$Xn,31		    // ROTATE(x[0]^x[2]^x[8]^x[13],1)
	nop.i	0			};;

___
}

sub BODY_20_39 {
local	*code=shift;
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;	add	$e=$e,$Konst		    // e+=K_XX_XX
	dep.z	tmp5=$a,5,27		}   // a<<5
{ .mib;	xor	tmp0=$c,$b
	xor	$Xn=$Xn,$X[($j+2)%16]	};; // forward Xupdate
{ .mib;	add	$e=$e,$X[$i%16]		    // e+=Xupdate
	extr.u	tmp1=$a,27,5		}   // a>>27
{ .mib;	xor	tmp0=tmp0,$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	$e=$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;	add	$e=$e,$Konst		    // e+=K_60_79
	dep.z	tmp5=$a,5,27		}   // a<<5
{ .mib;	xor	tmp0=$c,$b
	add	$h1=$h1,$a		};; // wrap up
{ .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
{ .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

___
}
}

sub BODY_40_59 {
local	*code=shift;
my	($i,$a,$b,$c,$d,$e)=@_;
my	$j=$i+1;
my	$Xn=@X[$j%16];

$code.=<<___;
{ .mib;	add	$e=$e,$K_40_59		    // e+=K_40_59
	dep.z	tmp5=$a,5,27		}   // a<<5
{ .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=$a,27,5		}   // a>>27
{ .mmi;	and	tmp0=tmp0,$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;	xor	$Xn=$Xn,tmp3
	mux2	tmp6=$a,0x44		};; // see b in next iteration
{ .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			};;

___
}
sub BODY_60_79	{ &BODY_20_39(@_,$K_60_79); }

$code.=<<___;
.text

tmp0=r8;
tmp1=r9;
tmp2=r10;
tmp3=r11;
ctx=r32;	// in0
inp=r33;	// in1

// void sha1_block_data_order(SHA_CTX *c,const void *p,size_t num);
.global	sha1_block_data_order#
.proc	sha1_block_data_order#
.align	32
sha1_block_data_order:
	.prologue
{ .mmi;	alloc	tmp1=ar.pfs,3,14,0,0
	$ADDP	tmp0=4,ctx
	.save	ar.lc,r3
	mov	r3=ar.lc		}
{ .mmi;	$ADDP	ctx=0,ctx
	$ADDP	inp=0,inp
	mov	r2=pr			};;
tmp4=in2;
tmp5=loc12;
tmp6=loc13;
	.body
{ .mlx;	ld4	$h0=[ctx],8
	movl	$K_00_19=0x5a827999	}
{ .mlx;	ld4	$h1=[tmp0],8
	movl	$K_20_39=0x6ed9eba1	};;
{ .mlx;	ld4	$h2=[ctx],8
	movl	$K_40_59=0x8f1bbcdc	}
{ .mlx;	ld4	$h3=[tmp0]
	movl	$K_60_79=0xca62c1d6	};;
{ .mmi;	ld4	$h4=[ctx],-16
	add	in2=-1,in2		    // adjust num for ar.lc
	mov	ar.ec=1			};;
{ .mmi;	nop.m	0
	add	tmp3=1,inp
	mov	ar.lc=in2		};; // brp.loop.imp: too far

.Ldtop:
{ .mmi;	mov	$A=$h0
	mov	$B=$h1
	mux2	tmp6=$h1,0x44		}
{ .mmi;	mov	$C=$h2
	mov	$D=$h3
	mov	$E=$h4			};;

___

{ my $i;
  my @V=($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,$i,@V); unshift(@V,pop(@V)); }
	for(;$i<40;$i++)	{ &BODY_20_39(\$code,$i,@V); unshift(@V,pop(@V)); }
	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[0] eq $A) and ($V[4] eq $E)) or die;	# double-check
}

$code.=<<___;
{ .mmb;	add	$h0=$h0,$A
	add	$h2=$h2,$C
	br.ctop.dptk.many	.Ldtop	};;
.Ldend:
{ .mmi;	add	tmp0=4,ctx
	mov	ar.lc=r3		};;
{ .mmi;	st4	[ctx]=$h0,8
	st4	[tmp0]=$h1,8		};;
{ .mmi;	st4	[ctx]=$h2,8
	st4	[tmp0]=$h3		};;
{ .mib;	st4	[ctx]=$h4,-16
	mov	pr=r2,0x1ffff
	br.ret.sptk.many	b0	};;
.endp	sha1_block_data_order#
stringz	"SHA1 block transform for IA64, CRYPTOGAMS by <appro\@openssl.org>"
___

open STDOUT,">$output" if $output;
print $code;
Commit	Line	Data
6aa36e8e RS	1	#! /usr/bin/env perl
	2	# Copyright 2004-2016 The OpenSSL Project Authors. All Rights Reserved.
	3	#
	4	# Licensed under the OpenSSL license (the "License"). You may not use
	5	# this file except in compliance with the License. You can obtain a copy
	6	# in the file LICENSE in the source distribution or at
	7	# https://www.openssl.org/source/license.html
	8
b4e0ce51 AP	9	#
	10	# ====================================================================
	11	# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
f0f61f6d AP	12	# project. The module is, however, dual licensed under OpenSSL and
	13	# CRYPTOGAMS licenses depending on where you obtain it. For further
	14	# details see http://www.openssl.org/~appro/cryptogams/.
b4e0ce51 AP	15	# ====================================================================
	16	#
	17	# Eternal question is what's wrong with compiler generated code? The
	18	# trick is that it's possible to reduce the number of shifts required
	19	# to perform rotations by maintaining copy of 32-bit value in upper
	20	# bits of 64-bit register. Just follow mux2 and shrp instructions...
	21	# Performance under big-endian OS such as HP-UX is 179MBps*1GHz, which
f0f61f6d	22	# is >50% better than HP C and >2x better than gcc.
b4e0ce51	23
e87e380a RL	24	$output = pop;
e87e380a RL	25
b4e0ce51	26	$code=<<___;
5727f1f7	27	.ident \"sha1-ia64.s, version 1.3\"
b4e0ce51 AP	28	.ident \"IA-64 ISA artwork by Andy Polyakov <appro\@fy.chalmers.se>\"
	29	.explicit
	30
	31	___
	32
	33
	34	if ($^O eq "hpux") {
	35	$ADDP="addp4";
	36	for (@ARGV) { $ADDP="add" if (/[\+DD\|\-mlp]64/); }
	37	} else { $ADDP="add"; }
b4e0ce51 AP	38
	39	#$human=1;
	40	if ($human) { # useful for visual code auditing...
5727f1f7	41	($A,$B,$C,$D,$E) = ("A","B","C","D","E");
b4e0ce51 AP	42	($h0,$h1,$h2,$h3,$h4) = ("h0","h1","h2","h3","h4");
	43	($K_00_19, $K_20_39, $K_40_59, $K_60_79) =
	44	( "K_00_19","K_20_39","K_40_59","K_60_79" );
	45	@X= ( "X0", "X1", "X2", "X3", "X4", "X5", "X6", "X7",
	46	"X8", "X9","X10","X11","X12","X13","X14","X15" );
	47	}
	48	else {
5727f1f7 AP	49	($A,$B,$C,$D,$E) = ("loc0","loc1","loc2","loc3","loc4");
5727f1f7 AP	50	($h0,$h1,$h2,$h3,$h4) = ("loc5","loc6","loc7","loc8","loc9");
b4e0ce51	51	($K_00_19, $K_20_39, $K_40_59, $K_60_79) =
5727f1f7	52	( "r14", "r15", "loc10", "loc11" );
b4e0ce51 AP	53	@X= ( "r16", "r17", "r18", "r19", "r20", "r21", "r22", "r23",
	54	"r24", "r25", "r26", "r27", "r28", "r29", "r30", "r31" );
	55	}
	56
	57	sub BODY_00_15 {
	58	local *code=shift;
5727f1f7 AP	59	my ($i,$a,$b,$c,$d,$e)=@_;
	60	my $j=$i+1;
	61	my $Xn=@X[$j%16];
b4e0ce51	62
232a938c	63	$code.=<<___ if ($i==0);
5727f1f7	64	{ .mmi; ld1 $X[$i]=[inp],2 // MSB
232a938c AP	65	ld1 tmp2=[tmp3],2 };;
	66	{ .mmi; ld1 tmp0=[inp],2
	67	ld1 tmp4=[tmp3],2 // LSB
5727f1f7	68	dep $X[$i]=$X[$i],tmp2,8,8 };;
b4e0ce51	69	___
b4e0ce51 AP	70	if ($i<15) {
b4e0ce51 AP	71	$code.=<<___;
5727f1f7 AP	72	{ .mmi; ld1 $Xn=[inp],2 // forward Xload
5727f1f7 AP	73	nop.m 0x0
232a938c	74	dep tmp1=tmp0,tmp4,8,8 };;
5727f1f7	75	{ .mmi; ld1 tmp2=[tmp3],2 // forward Xload
232a938c	76	and tmp4=$c,$b
5727f1f7 AP	77	dep $X[$i]=$X[$i],tmp1,16,16} //;;
	78	{ .mmi; add $e=$e,$K_00_19 // e+=K_00_19
	79	andcm tmp1=$d,$b
232a938c	80	dep.z tmp5=$a,5,27 };; // a<<5
5727f1f7 AP	81	{ .mmi; add $e=$e,$X[$i] // e+=Xload
5727f1f7 AP	82	or tmp4=tmp4,tmp1 // F_00_19(b,c,d)=(b&c)\|(~b&d)
b4e0ce51	83	extr.u tmp1=$a,27,5 };; // a>>27
5727f1f7 AP	84	{ .mmi; ld1 tmp0=[inp],2 // forward Xload
5727f1f7 AP	85	add $e=$e,tmp4 // e+=F_00_19(b,c,d)
b4e0ce51	86	shrp $b=tmp6,tmp6,2 } // b=ROTATE(b,30)
5727f1f7	87	{ .mmi; ld1 tmp4=[tmp3],2 // forward Xload
232a938c	88	or tmp5=tmp1,tmp5 // ROTATE(a,5)
b4e0ce51	89	mux2 tmp6=$a,0x44 };; // see b in next iteration
5727f1f7 AP	90	{ .mii; add $e=$e,tmp5 // e+=ROTATE(a,5)
	91	dep $Xn=$Xn,tmp2,8,8 // forward Xload
	92	mux2 $X[$i]=$X[$i],0x44 } //;;
b4e0ce51 AP	93
	94	___
	95	}
	96	else {
	97	$code.=<<___;
232a938c AP	98	{ .mii; and tmp3=$c,$b
232a938c AP	99	dep tmp1=tmp0,tmp4,8,8;;
5727f1f7 AP	100	dep $X[$i]=$X[$i],tmp1,16,16} //;;
	101	{ .mmi; add $e=$e,$K_00_19 // e+=K_00_19
	102	andcm tmp1=$d,$b
232a938c	103	dep.z tmp5=$a,5,27 };; // a<<5
5727f1f7 AP	104	{ .mmi; add $e=$e,$X[$i] // e+=Xupdate
5727f1f7 AP	105	or tmp4=tmp3,tmp1 // F_00_19(b,c,d)=(b&c)\|(~b&d)
b4e0ce51	106	extr.u tmp1=$a,27,5 } // a>>27
5727f1f7 AP	107	{ .mmi; xor $Xn=$Xn,$X[($j+2)%16] // forward Xupdate
5727f1f7 AP	108	xor tmp3=$X[($j+8)%16],$X[($j+13)%16] // forward Xupdate
b4e0ce51	109	nop.i 0 };;
5727f1f7 AP	110	{ .mmi; add $e=$e,tmp4 // e+=F_00_19(b,c,d)
5727f1f7 AP	111	xor $Xn=$Xn,tmp3 // forward Xupdate
b4e0ce51 AP	112	shrp $b=tmp6,tmp6,2 } // b=ROTATE(b,30)
	113	{ .mmi; or tmp1=tmp1,tmp5 // ROTATE(a,5)
	114	mux2 tmp6=$a,0x44 };; // see b in next iteration
5727f1f7 AP	115	{ .mii; add $e=$e,tmp1 // e+=ROTATE(a,5)
	116	shrp $Xn=$Xn,$Xn,31 // ROTATE(x[0]^x[2]^x[8]^x[13],1)
	117	mux2 $X[$i]=$X[$i],0x44 };;
b4e0ce51 AP	118
	119	___
	120	}
	121	}
	122
	123	sub BODY_16_19 {
	124	local *code=shift;
5727f1f7 AP	125	my ($i,$a,$b,$c,$d,$e)=@_;
	126	my $j=$i+1;
	127	my $Xn=@X[$j%16];
b4e0ce51 AP	128
b4e0ce51 AP	129	$code.=<<___;
5727f1f7	130	{ .mib; add $e=$e,$K_00_19 // e+=K_00_19
b4e0ce51	131	dep.z tmp5=$a,5,27 } // a<<5
5727f1f7 AP	132	{ .mib; andcm tmp1=$d,$b
	133	and tmp0=$c,$b };;
	134	{ .mmi; add $e=$e,$X[$i%16] // e+=Xupdate
	135	or tmp0=tmp0,tmp1 // F_00_19(b,c,d)=(b&c)\|(~b&d)
b4e0ce51	136	extr.u tmp1=$a,27,5 } // a>>27
5727f1f7 AP	137	{ .mmi; xor $Xn=$Xn,$X[($j+2)%16] // forward Xupdate
5727f1f7 AP	138	xor tmp3=$X[($j+8)%16],$X[($j+13)%16] // forward Xupdate
b4e0ce51	139	nop.i 0 };;
5727f1f7 AP	140	{ .mmi; add $e=$e,tmp0 // f+=F_00_19(b,c,d)
5727f1f7 AP	141	xor $Xn=$Xn,tmp3 // forward Xupdate
b4e0ce51 AP	142	shrp $b=tmp6,tmp6,2 } // b=ROTATE(b,30)
	143	{ .mmi; or tmp1=tmp1,tmp5 // ROTATE(a,5)
	144	mux2 tmp6=$a,0x44 };; // see b in next iteration
5727f1f7 AP	145	{ .mii; add $e=$e,tmp1 // e+=ROTATE(a,5)
5727f1f7 AP	146	shrp $Xn=$Xn,$Xn,31 // ROTATE(x[0]^x[2]^x[8]^x[13],1)
b4e0ce51 AP	147	nop.i 0 };;
	148
	149	___
	150	}
	151
	152	sub BODY_20_39 {
	153	local *code=shift;
5727f1f7	154	my ($i,$a,$b,$c,$d,$e,$Konst)=@_;
b4e0ce51	155	$Konst = $K_20_39 if (!defined($Konst));
5727f1f7 AP	156	my $j=$i+1;
5727f1f7 AP	157	my $Xn=@X[$j%16];
b4e0ce51 AP	158
	159	if ($i<79) {
	160	$code.=<<___;
5727f1f7	161	{ .mib; add $e=$e,$Konst // e+=K_XX_XX
b4e0ce51 AP	162	dep.z tmp5=$a,5,27 } // a<<5
b4e0ce51 AP	163	{ .mib; xor tmp0=$c,$b
5727f1f7 AP	164	xor $Xn=$Xn,$X[($j+2)%16] };; // forward Xupdate
5727f1f7 AP	165	{ .mib; add $e=$e,$X[$i%16] // e+=Xupdate
b4e0ce51	166	extr.u tmp1=$a,27,5 } // a>>27
5727f1f7 AP	167	{ .mib; xor tmp0=tmp0,$d // F_20_39(b,c,d)=b^c^d
	168	xor $Xn=$Xn,$X[($j+8)%16] };; // forward Xupdate
	169	{ .mmi; add $e=$e,tmp0 // e+=F_20_39(b,c,d)
	170	xor $Xn=$Xn,$X[($j+13)%16] // forward Xupdate
b4e0ce51 AP	171	shrp $b=tmp6,tmp6,2 } // b=ROTATE(b,30)
	172	{ .mmi; or tmp1=tmp1,tmp5 // ROTATE(a,5)
	173	mux2 tmp6=$a,0x44 };; // see b in next iteration
5727f1f7 AP	174	{ .mii; add $e=$e,tmp1 // e+=ROTATE(a,5)
5727f1f7 AP	175	shrp $Xn=$Xn,$Xn,31 // ROTATE(x[0]^x[2]^x[8]^x[13],1)
b4e0ce51 AP	176	nop.i 0 };;
	177
	178	___
	179	}
	180	else {
	181	$code.=<<___;
5727f1f7	182	{ .mib; add $e=$e,$Konst // e+=K_60_79
b4e0ce51 AP	183	dep.z tmp5=$a,5,27 } // a<<5
b4e0ce51 AP	184	{ .mib; xor tmp0=$c,$b
b4e0ce51	185	add $h1=$h1,$a };; // wrap up
5727f1f7 AP	186	{ .mib; add $e=$e,$X[$i%16] // e+=Xupdate
	187	extr.u tmp1=$a,27,5 } // a>>27
	188	{ .mib; xor tmp0=tmp0,$d // F_20_39(b,c,d)=b^c^d
b4e0ce51	189	add $h3=$h3,$c };; // wrap up
5727f1f7 AP	190	{ .mmi; add $e=$e,tmp0 // e+=F_20_39(b,c,d)
	191	or tmp1=tmp1,tmp5 // ROTATE(a,5)
	192	shrp $b=tmp6,tmp6,2 };; // b=ROTATE(b,30) ;;?
	193	{ .mmi; add $e=$e,tmp1 // e+=ROTATE(a,5)
	194	add tmp3=1,inp // used in unaligned codepath
b4e0ce51 AP	195	add $h4=$h4,$d };; // wrap up
	196
	197	___
	198	}
	199	}
	200
	201	sub BODY_40_59 {
	202	local *code=shift;
5727f1f7 AP	203	my ($i,$a,$b,$c,$d,$e)=@_;
	204	my $j=$i+1;
	205	my $Xn=@X[$j%16];
b4e0ce51 AP	206
b4e0ce51 AP	207	$code.=<<___;
5727f1f7	208	{ .mib; add $e=$e,$K_40_59 // e+=K_40_59
b4e0ce51	209	dep.z tmp5=$a,5,27 } // a<<5
5727f1f7 AP	210	{ .mib; and tmp1=$c,$d
	211	xor tmp0=$c,$d };;
	212	{ .mmi; add $e=$e,$X[$i%16] // e+=Xupdate
	213	add tmp5=tmp5,tmp1 // a<<5+(c&d)
b4e0ce51	214	extr.u tmp1=$a,27,5 } // a>>27
5727f1f7 AP	215	{ .mmi; and tmp0=tmp0,$b
	216	xor $Xn=$Xn,$X[($j+2)%16] // forward Xupdate
	217	xor tmp3=$X[($j+8)%16],$X[($j+13)%16] };; // forward Xupdate
	218	{ .mmi; add $e=$e,tmp0 // e+=b&(c^d)
	219	add tmp5=tmp5,tmp1 // ROTATE(a,5)+(c&d)
b4e0ce51	220	shrp $b=tmp6,tmp6,2 } // b=ROTATE(b,30)
5727f1f7	221	{ .mmi; xor $Xn=$Xn,tmp3
b4e0ce51	222	mux2 tmp6=$a,0x44 };; // see b in next iteration
5727f1f7 AP	223	{ .mii; add $e=$e,tmp5 // e+=ROTATE(a,5)+(c&d)
	224	shrp $Xn=$Xn,$Xn,31 // ROTATE(x[0]^x[2]^x[8]^x[13],1)
	225	nop.i 0x0 };;
b4e0ce51 AP	226
	227	___
	228	}
	229	sub BODY_60_79 { &BODY_20_39(@_,$K_60_79); }
	230
	231	$code.=<<___;
	232	.text
	233
	234	tmp0=r8;
	235	tmp1=r9;
	236	tmp2=r10;
	237	tmp3=r11;
	238	ctx=r32; // in0
	239	inp=r33; // in1
	240
c5f17d45 AP	241	// void sha1_block_data_order(SHA_CTX c,const void p,size_t num);
	242	.global sha1_block_data_order#
	243	.proc sha1_block_data_order#
b4e0ce51	244	.align 32
c5f17d45	245	sha1_block_data_order:
b4e0ce51	246	.prologue
5727f1f7	247	{ .mmi; alloc tmp1=ar.pfs,3,14,0,0
b4e0ce51	248	$ADDP tmp0=4,ctx
ef428d56	249	.save ar.lc,r3
b4e0ce51 AP	250	mov r3=ar.lc }
	251	{ .mmi; $ADDP ctx=0,ctx
	252	$ADDP inp=0,inp
	253	mov r2=pr };;
	254	tmp4=in2;
5727f1f7 AP	255	tmp5=loc12;
5727f1f7 AP	256	tmp6=loc13;
b4e0ce51 AP	257	.body
	258	{ .mlx; ld4 $h0=[ctx],8
	259	movl $K_00_19=0x5a827999 }
	260	{ .mlx; ld4 $h1=[tmp0],8
	261	movl $K_20_39=0x6ed9eba1 };;
	262	{ .mlx; ld4 $h2=[ctx],8
	263	movl $K_40_59=0x8f1bbcdc }
	264	{ .mlx; ld4 $h3=[tmp0]
	265	movl $K_60_79=0xca62c1d6 };;
	266	{ .mmi; ld4 $h4=[ctx],-16
	267	add in2=-1,in2 // adjust num for ar.lc
	268	mov ar.ec=1 };;
	269	{ .mmi; nop.m 0
	270	add tmp3=1,inp
	271	mov ar.lc=in2 };; // brp.loop.imp: too far
	272
	273	.Ldtop:
	274	{ .mmi; mov $A=$h0
	275	mov $B=$h1
	276	mux2 tmp6=$h1,0x44 }
	277	{ .mmi; mov $C=$h2
	278	mov $D=$h3
	279	mov $E=$h4 };;
	280
	281	___
	282
f9c5e5d9 AP	283	{ my $i;
f9c5e5d9 AP	284	my @V=($A,$B,$C,$D,$E);
b4e0ce51	285
232a938c	286	for($i=0;$i<16;$i++) { &BODY_00_15(\$code,$i,@V); unshift(@V,pop(@V)); }
f0f61f6d AP	287	for(;$i<20;$i++) { &BODY_16_19(\$code,$i,@V); unshift(@V,pop(@V)); }
	288	for(;$i<40;$i++) { &BODY_20_39(\$code,$i,@V); unshift(@V,pop(@V)); }
	289	for(;$i<60;$i++) { &BODY_40_59(\$code,$i,@V); unshift(@V,pop(@V)); }
	290	for(;$i<80;$i++) { &BODY_60_79(\$code,$i,@V); unshift(@V,pop(@V)); }
b4e0ce51	291
5727f1f7	292	(($V[0] eq $A) and ($V[4] eq $E)) or die; # double-check
f0f61f6d	293	}
b4e0ce51 AP	294
b4e0ce51 AP	295	$code.=<<___;
5727f1f7 AP	296	{ .mmb; add $h0=$h0,$A
5727f1f7 AP	297	add $h2=$h2,$C
b4e0ce51 AP	298	br.ctop.dptk.many .Ldtop };;
	299	.Ldend:
	300	{ .mmi; add tmp0=4,ctx
	301	mov ar.lc=r3 };;
	302	{ .mmi; st4 [ctx]=$h0,8
	303	st4 [tmp0]=$h1,8 };;
	304	{ .mmi; st4 [ctx]=$h2,8
	305	st4 [tmp0]=$h3 };;
	306	{ .mib; st4 [ctx]=$h4,-16
	307	mov pr=r2,0x1ffff
	308	br.ret.sptk.many b0 };;
c5f17d45	309	.endp sha1_block_data_order#
232a938c	310	stringz "SHA1 block transform for IA64, CRYPTOGAMS by <appro\@openssl.org>"
b4e0ce51 AP	311	___
b4e0ce51 AP	312
e87e380a	313	open STDOUT,">$output" if $output;
b4e0ce51	314	print $code;