-#!/usr/bin/env perl
+#! /usr/bin/env perl
+# Copyright 2010-2016 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
+# 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/.
# ~17%, but it comes for free, because it's same instruction sequence.
# Improvement coefficients are for aligned input.
+# September 2012.
+#
+# Add MIPS[32|64]R2 code (>25% less instructions).
+
######################################################################
# 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
# The return value is placed in $a0. Following coding rules facilitate
# interoperability:
#
-# - never ever touch $tp, "thread pointer", former $gp;
+# - never ever touch $tp, "thread pointer", former $gp [o32 can be
+# excluded from the rule, because it's specified volatile];
# - copy return value to $t0, former $v0 [or to $a0 if you're adapting
# old code];
# - on O32 populate $a4-$a7 with 'lw $aN,4*N($sp)' if necessary;
# ($s0,$s1,$s2,$s3,$s4,$s5,$s6,$s7)=map("\$$_",(16..23));
# ($gp,$sp,$fp,$ra)=map("\$$_",(28..31));
#
-$flavour = shift; # supported flavours are o32,n32,64,nubi32,nubi64
+$flavour = shift || "o32"; # 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
+ $PTR_LA="dla";
+ $PTR_ADD="daddu"; # incidentally works even on n32
+ $PTR_SUB="dsubu"; # 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";
+ $PTR_LA="la";
+ $PTR_ADD="addu";
+ $PTR_SUB="subu";
$REG_S="sw";
$REG_L="lw";
$PTR_SLL="sll";
$SZREG=4;
}
+$pf = ($flavour =~ /nubi/i) ? $t0 : $t2;
#
# <appro@openssl.org>
#
######################################################################
-$output=shift;
-for (@ARGV) { $big_endian=1 if (/\-DB_ENDIAN/);
- $big_endian=0 if (/\-DL_ENDIAN/);
- $output=$_ if (/^\w[\w\-]*\.\w+$/); }
+$big_endian=(`echo MIPSEL | $ENV{CC} -E -`=~/MIPSEL/)?1:0 if ($ENV{CC});
+
+for (@ARGV) { $output=$_ if (/\w[\w\-]*\.\w+$/); }
open STDOUT,">$output";
if (!defined($big_endian)) { $big_endian=(unpack('L',pack('N',1))==1); }
$SLL="dsll"; # shift left logical
$SRL="dsrl"; # shift right logical
$ADDU="daddu";
+ $ROTR="drotr";
@Sigma0=(28,34,39);
@Sigma1=(14,18,41);
@sigma0=( 7, 1, 8); # right shift first
$SLL="sll"; # shift left logical
$SRL="srl"; # shift right logical
$ADDU="addu";
+ $ROTR="rotr";
@Sigma0=( 2,13,22);
@Sigma1=( 6,11,25);
@sigma0=( 3, 7,18); # right shift first
$rounds=64;
}
+$MSB = $big_endian ? 0 : ($SZ-1);
+$LSB = ($SZ-1)&~$MSB;
+
@V=($A,$B,$C,$D,$E,$F,$G,$H)=map("\$$_",(1,2,3,7,24,25,30,31));
@X=map("\$$_",(8..23));
my ($T1,$tmp0,$tmp1,$tmp2)=(@X[4],@X[5],@X[6],@X[7]);
$code.=<<___ if ($i<15);
+#if defined(_MIPS_ARCH_MIPS32R6) || defined(_MIPS_ARCH_MIPS64R6)
+ ${LD} @X[1],`($i+1)*$SZ`($inp)
+#else
${LD}l @X[1],`($i+1)*$SZ+$MSB`($inp)
${LD}r @X[1],`($i+1)*$SZ+$LSB`($inp)
+#endif
___
-$code.=<<___ if (!$big_endian && $i<16); # XXX no 64-bit byte swap yet
- srl $tmp0,@X[0],24 # byte swap($i)
+$code.=<<___ if (!$big_endian && $i<16 && $SZ==4);
+#if defined(_MIPS_ARCH_MIPS32R2) || defined(_MIPS_ARCH_MIPS64R2)
+ wsbh @X[0],@X[0] # byte swap($i)
+ rotr @X[0],@X[0],16
+#else
+ srl $tmp0,@X[0],24 # byte swap($i)
srl $tmp1,@X[0],8
andi $tmp2,@X[0],0xFF00
sll @X[0],@X[0],24
andi $tmp1,0xFF00
sll $tmp2,$tmp2,8
or @X[0],$tmp0
- or $tmp1,$t2
+ or $tmp1,$tmp2
+ or @X[0],$tmp1
+#endif
+___
+$code.=<<___ if (!$big_endian && $i<16 && $SZ==8);
+#if defined(_MIPS_ARCH_MIPS64R2)
+ dsbh @X[0],@X[0] # byte swap($i)
+ dshd @X[0],@X[0]
+#else
+ ori $tmp0,$zero,0xFF
+ dsll $tmp2,$tmp0,32
+ or $tmp0,$tmp2 # 0x000000FF000000FF
+ and $tmp1,@X[0],$tmp0 # byte swap($i)
+ dsrl $tmp2,@X[0],24
+ dsll $tmp1,24
+ and $tmp2,$tmp0
+ dsll $tmp0,8 # 0x0000FF000000FF00
+ or $tmp1,$tmp2
+ and $tmp2,@X[0],$tmp0
+ dsrl @X[0],8
+ dsll $tmp2,8
+ and @X[0],$tmp0
+ or $tmp1,$tmp2
+ or @X[0],$tmp1
+ dsrl $tmp1,@X[0],32
+ dsll @X[0],32
or @X[0],$tmp1
+#endif
___
$code.=<<___;
+#if defined(_MIPS_ARCH_MIPS32R2) || defined(_MIPS_ARCH_MIPS64R2)
+ xor $tmp2,$f,$g # $i
+ $ROTR $tmp0,$e,@Sigma1[0]
+ $ADDU $T1,$X[0],$h
+ $ROTR $tmp1,$e,@Sigma1[1]
+ and $tmp2,$e
+ $ROTR $h,$e,@Sigma1[2]
+ xor $tmp0,$tmp1
+ $ROTR $tmp1,$a,@Sigma0[0]
+ xor $tmp2,$g # Ch(e,f,g)
+ xor $tmp0,$h # Sigma1(e)
+
+ $ROTR $h,$a,@Sigma0[1]
+ $ADDU $T1,$tmp2
+ $LD $tmp2,`$i*$SZ`($Ktbl) # K[$i]
+ xor $h,$tmp1
+ $ROTR $tmp1,$a,@Sigma0[2]
+ $ADDU $T1,$tmp0
+ and $tmp0,$b,$c
+ xor $h,$tmp1 # Sigma0(a)
+ xor $tmp1,$b,$c
+#else
$ADDU $T1,$X[0],$h # $i
$SRL $h,$e,@Sigma1[0]
xor $tmp2,$f,$g
xor $h,$tmp1
$SLL $tmp1,$a,`$SZ*8-@Sigma0[0]`
xor $h,$tmp0
- $ST @X[0],`($i%16)*$SZ`($sp)
+ and $tmp0,$b,$c
xor $h,$tmp1 # Sigma0(a)
-
- or $tmp0,$a,$b
- and $tmp1,$a,$b
- and $tmp0,$c
- or $tmp1,$tmp0 # Maj(a,b,c)
+ xor $tmp1,$b,$c
+#endif
+ $ST @X[0],`($i%16)*$SZ`($sp) # offload to ring buffer
+ $ADDU $h,$tmp0
+ and $tmp1,$a
$ADDU $T1,$tmp2 # +=K[$i]
- $ADDU $h,$tmp1
-
+ $ADDU $h,$tmp1 # +=Maj(a,b,c)
$ADDU $d,$T1
$ADDU $h,$T1
___
$code.=<<___ if ($i>=13);
- $LD @X[3],`(($i+3)%16)*$SZ`($sp) # prefetch
+ $LD @X[3],`(($i+3)%16)*$SZ`($sp) # prefetch from ring buffer
___
}
my ($tmp0,$tmp1,$tmp2,$tmp3)=(@X[4],@X[5],@X[6],@X[7]);
$code.=<<___;
+#if defined(_MIPS_ARCH_MIPS32R2) || defined(_MIPS_ARCH_MIPS64R2)
+ $SRL $tmp2,@X[1],@sigma0[0] # Xupdate($i)
+ $ROTR $tmp0,@X[1],@sigma0[1]
+ $ADDU @X[0],@X[9] # +=X[i+9]
+ xor $tmp2,$tmp0
+ $ROTR $tmp0,@X[1],@sigma0[2]
+
+ $SRL $tmp3,@X[14],@sigma1[0]
+ $ROTR $tmp1,@X[14],@sigma1[1]
+ xor $tmp2,$tmp0 # sigma0(X[i+1])
+ $ROTR $tmp0,@X[14],@sigma1[2]
+ xor $tmp3,$tmp1
+ $ADDU @X[0],$tmp2
+#else
$SRL $tmp2,@X[1],@sigma0[0] # Xupdate($i)
$ADDU @X[0],@X[9] # +=X[i+9]
$SLL $tmp1,@X[1],`$SZ*8-@sigma0[2]`
xor $tmp3,$tmp0
$SRL $tmp0,@X[14],@sigma1[2]
xor $tmp3,$tmp1
-
+#endif
xor $tmp3,$tmp0 # sigma1(X[i+14])
$ADDU @X[0],$tmp3
-
___
&BODY_00_15(@_);
}
$FRAMESIZE=16*$SZ+16*$SZREG;
-$SAVED_REGS_MASK = ($flavour =~ /nubi/i) ? 0xc0fff008 : 0xc0ff0000;
-$pf = ($flavour =~ /nubi/i) ? $t0 : $t2;
-$MSB = 0;
-$LSB = ($SZ-1)&~$MSB;
+$SAVED_REGS_MASK = ($flavour =~ /nubi/i) ? "0xc0fff008" : "0xc0ff0000";
$code.=<<___;
+#include "mips_arch.h"
+
.text
.set noat
+#if !defined(__mips_eabi) && (!defined(__vxworks) || defined(__pic__))
.option pic2
+#endif
.align 5
.globl sha${label}_block_data_order
___
$code.=<<___;
.set reorder
- la $Ktbl,K${label} # PIC-ified 'load address'
+ $PTR_LA $Ktbl,K${label} # PIC-ified 'load address'
$LD $A,0*$SZ($ctx) # load context
$LD $B,1*$SZ($ctx)
.align 5
.Loop:
+#if defined(_MIPS_ARCH_MIPS32R6) || defined(_MIPS_ARCH_MIPS64R6)
+ ${LD} @X[0],($inp)
+#else
${LD}l @X[0],$MSB($inp)
${LD}r @X[0],$LSB($inp)
+#endif
___
for ($i=0;$i<16;$i++)
{ &BODY_00_15($i,@V); unshift(@V,pop(@V)); push(@X,shift(@X)); }
$code.=<<___;
and @X[6],0xfff
li @X[7],$lastK
- $PTR_ADD $Ktbl,16*$SZ # Ktbl+=16
+ .set noreorder
bne @X[6],@X[7],.L16_xx
+ $PTR_ADD $Ktbl,16*$SZ # Ktbl+=16
+ $REG_L @X[15],16*$SZ($sp) # restore pointer to the end of input
$LD @X[0],0*$SZ($ctx)
$LD @X[1],1*$SZ($ctx)
$LD @X[2],2*$SZ($ctx)
+ $PTR_ADD $inp,16*$SZ
$LD @X[3],3*$SZ($ctx)
$ADDU $A,@X[0]
$LD @X[4],4*$SZ($ctx)
$ST $C,2*$SZ($ctx)
$ADDU $H,@X[7]
$ST $D,3*$SZ($ctx)
- $PTR_ADD $inp,16*$SZ
$ST $E,4*$SZ($ctx)
- $REG_L @X[15],16*$SZ($sp) # restore pointer to the end of input
$ST $F,5*$SZ($ctx)
$ST $G,6*$SZ($ctx)
$ST $H,7*$SZ($ctx)
- $PTR_SUB $Ktbl,`($rounds-16)*$SZ` # rewind $Ktbl
bne $inp,@X[15],.Loop
+ $PTR_SUB $Ktbl,`($rounds-16)*$SZ` # rewind $Ktbl
- .set noreorder
$REG_L $ra,$FRAMESIZE-1*$SZREG($sp)
$REG_L $fp,$FRAMESIZE-2*$SZREG($sp)
$REG_L $s11,$FRAMESIZE-3*$SZREG($sp)