Commit b66723b2 authored by Andy Polyakov's avatar Andy Polyakov
Browse files

MIPS assembler pack update from HEAD.

parent cf96d71c

crypto/aes/asm/aes-mips.pl

0 → 100644
+1611 −0

File added.

Preview size limit exceeded, changes collapsed.

crypto/bn/asm/mips-mont.pl

0 → 100644
+426 −0
Original line number Diff line number Diff line 
#!/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/.

# ====================================================================



# This module doesn't present direct interest for OpenSSL, because it

# doesn't provide better performance for longer keys, at least not on

# in-order-execution cores. While 512-bit RSA sign operations can be

# 65% faster in 64-bit mode, 1024-bit ones are only 15% faster, and

# 4096-bit ones are up to 15% slower. In 32-bit mode it varies from

# 16% improvement for 512-bit RSA sign to -33% for 4096-bit RSA

# verify:-( All comparisons are against bn_mul_mont-free assembler.

# The module might be of interest to embedded system developers, as

# the code is smaller than 1KB, yet offers >3x improvement on MIPS64

# and 75-30% [less for longer keys] on MIPS32 over compiler-generated

# code.



######################################################################

# 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";

	$SZREG=8;

} else {

	$PTR_ADD="add";

	$PTR_SUB="sub";

	$REG_S="sw";

	$REG_L="lw";

	$SZREG=4;

}

$SAVED_REGS_MASK = ($flavour =~ /nubi/i) ? 0x00fff000 : 0x00ff0000;

#

# <appro@openssl.org>

#

######################################################################



while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {}

open STDOUT,">$output";



if ($flavour =~ /64|n32/i) {

	$LD="ld";

	$ST="sd";

	$MULTU="dmultu";

	$ADDU="daddu";

	$SUBU="dsubu";

	$BNSZ=8;

} else {

	$LD="lw";

	$ST="sw";

	$MULTU="multu";

	$ADDU="addu";

	$SUBU="subu";

	$BNSZ=4;

}



# int bn_mul_mont(

$rp=$a0;	# BN_ULONG *rp,

$ap=$a1;	# const BN_ULONG *ap,

$bp=$a2;	# const BN_ULONG *bp,

$np=$a3;	# const BN_ULONG *np,

$n0=$a4;	# const BN_ULONG *n0,

$num=$a5;	# int num);



$lo0=$a6;

$hi0=$a7;

$lo1=$t1;

$hi1=$t2;

$aj=$s0;

$bi=$s1;

$nj=$s2;

$tp=$s3;

$alo=$s4;

$ahi=$s5;

$nlo=$s6;

$nhi=$s7;

$tj=$s8;

$i=$s9;

$j=$s10;

$m1=$s11;



$FRAMESIZE=14;



$code=<<___;

.text



.set	noat

.set	noreorder



.align	5

.globl	bn_mul_mont

.ent	bn_mul_mont

bn_mul_mont:

___

$code.=<<___ if ($flavour =~ /o32/i);

	lw	$n0,16($sp)

	lw	$num,20($sp)

___

$code.=<<___;

	slt	$at,$num,4

	bnez	$at,1f

	li	$t0,0

	slt	$at,$num,17	# on in-order CPU

	bnezl	$at,bn_mul_mont_internal

	nop

1:	jr	$ra

	li	$a0,0

.end	bn_mul_mont



.align	5

.ent	bn_mul_mont_internal

bn_mul_mont_internal:

	.frame	$fp,$FRAMESIZE*$SZREG,$ra

	.mask	0x40000000|$SAVED_REGS_MASK,-$SZREG

	$PTR_SUB $sp,$FRAMESIZE*$SZREG

	$REG_S	$fp,($FRAMESIZE-1)*$SZREG($sp)

	$REG_S	$s11,($FRAMESIZE-2)*$SZREG($sp)

	$REG_S	$s10,($FRAMESIZE-3)*$SZREG($sp)

	$REG_S	$s9,($FRAMESIZE-4)*$SZREG($sp)

	$REG_S	$s8,($FRAMESIZE-5)*$SZREG($sp)

	$REG_S	$s7,($FRAMESIZE-6)*$SZREG($sp)

	$REG_S	$s6,($FRAMESIZE-7)*$SZREG($sp)

	$REG_S	$s5,($FRAMESIZE-8)*$SZREG($sp)

	$REG_S	$s4,($FRAMESIZE-9)*$SZREG($sp)

___

$code.=<<___ if ($flavour =~ /nubi/i);

	$REG_S	$s3,($FRAMESIZE-10)*$SZREG($sp)

	$REG_S	$s2,($FRAMESIZE-11)*$SZREG($sp)

	$REG_S	$s1,($FRAMESIZE-12)*$SZREG($sp)

	$REG_S	$s0,($FRAMESIZE-13)*$SZREG($sp)

___

$code.=<<___;

	move	$fp,$sp



	.set	reorder

	$LD	$n0,0($n0)

	$LD	$bi,0($bp)	# bp[0]

	$LD	$aj,0($ap)	# ap[0]

	$LD	$nj,0($np)	# np[0]



	$PTR_SUB $sp,2*$BNSZ	# place for two extra words

	sll	$num,`log($BNSZ)/log(2)`

	li	$at,-4096

	$PTR_SUB $sp,$num

	and	$sp,$at



	$MULTU	$aj,$bi

	$LD	$alo,$BNSZ($ap)

	$LD	$nlo,$BNSZ($np)

	mflo	$lo0

	mfhi	$hi0

	$MULTU	$lo0,$n0

	mflo	$m1



	$MULTU	$alo,$bi

	mflo	$alo

	mfhi	$ahi



	$MULTU	$nj,$m1

	mflo	$lo1

	mfhi	$hi1

	$MULTU	$nlo,$m1

	$ADDU	$lo1,$lo0

	sltu	$at,$lo1,$lo0

	$ADDU	$hi1,$at

	mflo	$nlo

	mfhi	$nhi



	move	$tp,$sp

	li	$j,2*$BNSZ

.align	4

.L1st:

	.set	noreorder

	$PTR_ADD $aj,$ap,$j

	$PTR_ADD $nj,$np,$j

	$LD	$aj,($aj)

	$LD	$nj,($nj)



	$MULTU	$aj,$bi

	$ADDU	$lo0,$alo,$hi0

	$ADDU	$lo1,$nlo,$hi1

	sltu	$at,$lo0,$hi0

	sltu	$t0,$lo1,$hi1

	$ADDU	$hi0,$ahi,$at

	$ADDU	$hi1,$nhi,$t0

	mflo	$alo

	mfhi	$ahi



	$ADDU	$lo1,$lo0

	sltu	$at,$lo1,$lo0

	$MULTU	$nj,$m1

	$ADDU	$hi1,$at

	addu	$j,$BNSZ

	$ST	$lo1,($tp)

	sltu	$t0,$j,$num

	mflo	$nlo

	mfhi	$nhi



	bnez	$t0,.L1st

	$PTR_ADD $tp,$BNSZ

	.set	reorder



	$ADDU	$lo0,$alo,$hi0

	sltu	$at,$lo0,$hi0

	$ADDU	$hi0,$ahi,$at



	$ADDU	$lo1,$nlo,$hi1

	sltu	$t0,$lo1,$hi1

	$ADDU	$hi1,$nhi,$t0

	$ADDU	$lo1,$lo0

	sltu	$at,$lo1,$lo0

	$ADDU	$hi1,$at



	$ST	$lo1,($tp)



	$ADDU	$hi1,$hi0

	sltu	$at,$hi1,$hi0

	$ST	$hi1,$BNSZ($tp)

	$ST	$at,2*$BNSZ($tp)



	li	$i,$BNSZ

.align	4

.Louter:

	$PTR_ADD $bi,$bp,$i

	$LD	$bi,($bi)

	$LD	$aj,($ap)

	$LD	$alo,$BNSZ($ap)

	$LD	$tj,($sp)



	$MULTU	$aj,$bi

	$LD	$nj,($np)

	$LD	$nlo,$BNSZ($np)

	mflo	$lo0

	mfhi	$hi0

	$ADDU	$lo0,$tj

	$MULTU	$lo0,$n0

	sltu	$at,$lo0,$tj

	$ADDU	$hi0,$at

	mflo	$m1



	$MULTU	$alo,$bi

	mflo	$alo

	mfhi	$ahi



	$MULTU	$nj,$m1

	mflo	$lo1

	mfhi	$hi1



	$MULTU	$nlo,$m1

	$ADDU	$lo1,$lo0

	sltu	$at,$lo1,$lo0

	$ADDU	$hi1,$at

	mflo	$nlo

	mfhi	$nhi



	move	$tp,$sp

	li	$j,2*$BNSZ

	$LD	$tj,$BNSZ($tp)

.align	4

.Linner:

	.set	noreorder

	$PTR_ADD $aj,$ap,$j

	$PTR_ADD $nj,$np,$j

	$LD	$aj,($aj)

	$LD	$nj,($nj)



	$MULTU	$aj,$bi

	$ADDU	$lo0,$alo,$hi0

	$ADDU	$lo1,$nlo,$hi1

	sltu	$at,$lo0,$hi0

	sltu	$t0,$lo1,$hi1

	$ADDU	$hi0,$ahi,$at

	$ADDU	$hi1,$nhi,$t0

	mflo	$alo

	mfhi	$ahi



	$ADDU	$lo0,$tj

	addu	$j,$BNSZ

	$MULTU	$nj,$m1

	sltu	$at,$lo0,$tj

	$ADDU	$lo1,$lo0

	$ADDU	$hi0,$at

	sltu	$t0,$lo1,$lo0

	$LD	$tj,2*$BNSZ($tp)

	$ADDU	$hi1,$t0

	sltu	$at,$j,$num

	mflo	$nlo

	mfhi	$nhi

	$ST	$lo1,($tp)

	bnez	$at,.Linner

	$PTR_ADD $tp,$BNSZ

	.set	reorder



	$ADDU	$lo0,$alo,$hi0

	sltu	$at,$lo0,$hi0

	$ADDU	$hi0,$ahi,$at

	$ADDU	$lo0,$tj

	sltu	$t0,$lo0,$tj

	$ADDU	$hi0,$t0



	$LD	$tj,2*$BNSZ($tp)

	$ADDU	$lo1,$nlo,$hi1

	sltu	$at,$lo1,$hi1

	$ADDU	$hi1,$nhi,$at

	$ADDU	$lo1,$lo0

	sltu	$t0,$lo1,$lo0

	$ADDU	$hi1,$t0

	$ST	$lo1,($tp)



	$ADDU	$lo1,$hi1,$hi0

	sltu	$hi1,$lo1,$hi0

	$ADDU	$lo1,$tj

	sltu	$at,$lo1,$tj

	$ADDU	$hi1,$at

	$ST	$lo1,$BNSZ($tp)

	$ST	$hi1,2*$BNSZ($tp)



	addu	$i,$BNSZ

	sltu	$t0,$i,$num

	bnez	$t0,.Louter



	.set	noreorder

	$PTR_ADD $tj,$sp,$num	# &tp[num]

	move	$tp,$sp

	move	$ap,$sp

	li	$hi0,0		# clear borrow bit



.align	4

.Lsub:	$LD	$lo0,($tp)

	$LD	$lo1,($np)

	$PTR_ADD $tp,$BNSZ

	$PTR_ADD $np,$BNSZ

	$SUBU	$lo1,$lo0,$lo1	# tp[i]-np[i]

	sgtu	$at,$lo1,$lo0

	$SUBU	$lo0,$lo1,$hi0

	sgtu	$hi0,$lo0,$lo1

	$ST	$lo0,($rp)

	or	$hi0,$at

	sltu	$at,$tp,$tj

	bnez	$at,.Lsub

	$PTR_ADD $rp,$BNSZ



	$SUBU	$hi0,$hi1,$hi0	# handle upmost overflow bit

	move	$tp,$sp

	$PTR_SUB $rp,$num	# restore rp

	not	$hi1,$hi0



	and	$ap,$hi0,$sp

	and	$bp,$hi1,$rp

	or	$ap,$ap,$bp	# ap=borrow?tp:rp



.align	4

.Lcopy:	$LD	$aj,($ap)

	$PTR_ADD $ap,$BNSZ

	$ST	$zero,($tp)

	$PTR_ADD $tp,$BNSZ

	sltu	$at,$tp,$tj

	$ST	$aj,($rp)

	bnez	$at,.Lcopy

	$PTR_ADD $rp,$BNSZ



	li	$a0,1

	li	$t0,1



	.set	noreorder

	move	$sp,$fp

	$REG_L	$fp,($FRAMESIZE-1)*$SZREG($sp)

	$REG_L	$s11,($FRAMESIZE-2)*$SZREG($sp)

	$REG_L	$s10,($FRAMESIZE-3)*$SZREG($sp)

	$REG_L	$s9,($FRAMESIZE-4)*$SZREG($sp)

	$REG_L	$s8,($FRAMESIZE-5)*$SZREG($sp)

	$REG_L	$s7,($FRAMESIZE-6)*$SZREG($sp)

	$REG_L	$s6,($FRAMESIZE-7)*$SZREG($sp)

	$REG_L	$s5,($FRAMESIZE-8)*$SZREG($sp)

	$REG_L	$s4,($FRAMESIZE-9)*$SZREG($sp)

___

$code.=<<___ if ($flavour =~ /nubi/i);

	$REG_L	$s3,($FRAMESIZE-10)*$SZREG($sp)

	$REG_L	$s2,($FRAMESIZE-11)*$SZREG($sp)

	$REG_L	$s1,($FRAMESIZE-12)*$SZREG($sp)

	$REG_L	$s0,($FRAMESIZE-13)*$SZREG($sp)

___

$code.=<<___;

	jr	$ra

	$PTR_ADD $sp,$FRAMESIZE*$SZREG

.end	bn_mul_mont_internal

.rdata

.asciiz	"Montgomery Multiplication for MIPS, CRYPTOGAMS by <appro\@openssl.org>"

___



$code =~ s/\`([^\`]*)\`/eval $1/gem;



print $code;

close STDOUT;

crypto/bn/asm/mips.pl

0 → 100644
+2585 −0

File added.

Preview size limit exceeded, changes collapsed.

crypto/sha/asm/sha1-mips.pl

0 → 100644
+354 −0
Original line number Diff line number Diff line 
#!/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;

crypto/sha/asm/sha512-mips.pl

0 → 100644
+455 −0

File added.

Preview size limit exceeded, changes collapsed.