crypto: vmx - Add support for VMS instructions by ASM
OpenSSL implements optimized ASM algorithms which support VMX instructions on Power 8 CPU. These scripts generate an endian-agnostic ASM implementation in order to support both big and little-endian. - aesp8-ppc.pl: implements suport for AES instructions implemented by POWER8 processor. - ghashp8-ppc.pl: implements support for GHASH for Power8. - ppc-xlate.pl: ppc assembler distiller. These code has been adopted from OpenSSL project in collaboration with the original author (Andy Polyakov <appro@openssl.org>). Signed-off-by: Leonidas S. Barbosa <leosilva@linux.vnet.ibm.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>hifive-unleashed-5.1
Leonidas S. Barbosa
Herbert Xu
parent
cc333cd68d
commit
5c380d623e
File diff suppressed because it is too large
Load Diff
|
|
@ -0,0 +1,234 @@
|
|||
#!/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/.
|
||||
# ====================================================================
|
||||
#
|
||||
# GHASH for for PowerISA v2.07.
|
||||
#
|
||||
# July 2014
|
||||
#
|
||||
# Accurate performance measurements are problematic, because it's
|
||||
# always virtualized setup with possibly throttled processor.
|
||||
# Relative comparison is therefore more informative. This initial
|
||||
# version is ~2.1x slower than hardware-assisted AES-128-CTR, ~12x
|
||||
# faster than "4-bit" integer-only compiler-generated 64-bit code.
|
||||
# "Initial version" means that there is room for futher improvement.
|
||||
|
||||
$flavour=shift;
|
||||
$output =shift;
|
||||
|
||||
if ($flavour =~ /64/) {
|
||||
$SIZE_T=8;
|
||||
$LRSAVE=2*$SIZE_T;
|
||||
$STU="stdu";
|
||||
$POP="ld";
|
||||
$PUSH="std";
|
||||
} elsif ($flavour =~ /32/) {
|
||||
$SIZE_T=4;
|
||||
$LRSAVE=$SIZE_T;
|
||||
$STU="stwu";
|
||||
$POP="lwz";
|
||||
$PUSH="stw";
|
||||
} else { die "nonsense $flavour"; }
|
||||
|
||||
$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
|
||||
( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or
|
||||
( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or
|
||||
die "can't locate ppc-xlate.pl";
|
||||
|
||||
open STDOUT,"| $^X $xlate $flavour $output" || die "can't call $xlate: $!";
|
||||
|
||||
my ($Xip,$Htbl,$inp,$len)=map("r$_",(3..6)); # argument block
|
||||
|
||||
my ($Xl,$Xm,$Xh,$IN)=map("v$_",(0..3));
|
||||
my ($zero,$t0,$t1,$t2,$xC2,$H,$Hh,$Hl,$lemask)=map("v$_",(4..12));
|
||||
my $vrsave="r12";
|
||||
|
||||
$code=<<___;
|
||||
.machine "any"
|
||||
|
||||
.text
|
||||
|
||||
.globl .gcm_init_p8
|
||||
.align 5
|
||||
.gcm_init_p8:
|
||||
lis r0,0xfff0
|
||||
li r8,0x10
|
||||
mfspr $vrsave,256
|
||||
li r9,0x20
|
||||
mtspr 256,r0
|
||||
li r10,0x30
|
||||
lvx_u $H,0,r4 # load H
|
||||
|
||||
vspltisb $xC2,-16 # 0xf0
|
||||
vspltisb $t0,1 # one
|
||||
vaddubm $xC2,$xC2,$xC2 # 0xe0
|
||||
vxor $zero,$zero,$zero
|
||||
vor $xC2,$xC2,$t0 # 0xe1
|
||||
vsldoi $xC2,$xC2,$zero,15 # 0xe1...
|
||||
vsldoi $t1,$zero,$t0,1 # ...1
|
||||
vaddubm $xC2,$xC2,$xC2 # 0xc2...
|
||||
vspltisb $t2,7
|
||||
vor $xC2,$xC2,$t1 # 0xc2....01
|
||||
vspltb $t1,$H,0 # most significant byte
|
||||
vsl $H,$H,$t0 # H<<=1
|
||||
vsrab $t1,$t1,$t2 # broadcast carry bit
|
||||
vand $t1,$t1,$xC2
|
||||
vxor $H,$H,$t1 # twisted H
|
||||
|
||||
vsldoi $H,$H,$H,8 # twist even more ...
|
||||
vsldoi $xC2,$zero,$xC2,8 # 0xc2.0
|
||||
vsldoi $Hl,$zero,$H,8 # ... and split
|
||||
vsldoi $Hh,$H,$zero,8
|
||||
|
||||
stvx_u $xC2,0,r3 # save pre-computed table
|
||||
stvx_u $Hl,r8,r3
|
||||
stvx_u $H, r9,r3
|
||||
stvx_u $Hh,r10,r3
|
||||
|
||||
mtspr 256,$vrsave
|
||||
blr
|
||||
.long 0
|
||||
.byte 0,12,0x14,0,0,0,2,0
|
||||
.long 0
|
||||
.size .gcm_init_p8,.-.gcm_init_p8
|
||||
|
||||
.globl .gcm_gmult_p8
|
||||
.align 5
|
||||
.gcm_gmult_p8:
|
||||
lis r0,0xfff8
|
||||
li r8,0x10
|
||||
mfspr $vrsave,256
|
||||
li r9,0x20
|
||||
mtspr 256,r0
|
||||
li r10,0x30
|
||||
lvx_u $IN,0,$Xip # load Xi
|
||||
|
||||
lvx_u $Hl,r8,$Htbl # load pre-computed table
|
||||
le?lvsl $lemask,r0,r0
|
||||
lvx_u $H, r9,$Htbl
|
||||
le?vspltisb $t0,0x07
|
||||
lvx_u $Hh,r10,$Htbl
|
||||
le?vxor $lemask,$lemask,$t0
|
||||
lvx_u $xC2,0,$Htbl
|
||||
le?vperm $IN,$IN,$IN,$lemask
|
||||
vxor $zero,$zero,$zero
|
||||
|
||||
vpmsumd $Xl,$IN,$Hl # H.lo·Xi.lo
|
||||
vpmsumd $Xm,$IN,$H # H.hi·Xi.lo+H.lo·Xi.hi
|
||||
vpmsumd $Xh,$IN,$Hh # H.hi·Xi.hi
|
||||
|
||||
vpmsumd $t2,$Xl,$xC2 # 1st phase
|
||||
|
||||
vsldoi $t0,$Xm,$zero,8
|
||||
vsldoi $t1,$zero,$Xm,8
|
||||
vxor $Xl,$Xl,$t0
|
||||
vxor $Xh,$Xh,$t1
|
||||
|
||||
vsldoi $Xl,$Xl,$Xl,8
|
||||
vxor $Xl,$Xl,$t2
|
||||
|
||||
vsldoi $t1,$Xl,$Xl,8 # 2nd phase
|
||||
vpmsumd $Xl,$Xl,$xC2
|
||||
vxor $t1,$t1,$Xh
|
||||
vxor $Xl,$Xl,$t1
|
||||
|
||||
le?vperm $Xl,$Xl,$Xl,$lemask
|
||||
stvx_u $Xl,0,$Xip # write out Xi
|
||||
|
||||
mtspr 256,$vrsave
|
||||
blr
|
||||
.long 0
|
||||
.byte 0,12,0x14,0,0,0,2,0
|
||||
.long 0
|
||||
.size .gcm_gmult_p8,.-.gcm_gmult_p8
|
||||
|
||||
.globl .gcm_ghash_p8
|
||||
.align 5
|
||||
.gcm_ghash_p8:
|
||||
lis r0,0xfff8
|
||||
li r8,0x10
|
||||
mfspr $vrsave,256
|
||||
li r9,0x20
|
||||
mtspr 256,r0
|
||||
li r10,0x30
|
||||
lvx_u $Xl,0,$Xip # load Xi
|
||||
|
||||
lvx_u $Hl,r8,$Htbl # load pre-computed table
|
||||
le?lvsl $lemask,r0,r0
|
||||
lvx_u $H, r9,$Htbl
|
||||
le?vspltisb $t0,0x07
|
||||
lvx_u $Hh,r10,$Htbl
|
||||
le?vxor $lemask,$lemask,$t0
|
||||
lvx_u $xC2,0,$Htbl
|
||||
le?vperm $Xl,$Xl,$Xl,$lemask
|
||||
vxor $zero,$zero,$zero
|
||||
|
||||
lvx_u $IN,0,$inp
|
||||
addi $inp,$inp,16
|
||||
subi $len,$len,16
|
||||
le?vperm $IN,$IN,$IN,$lemask
|
||||
vxor $IN,$IN,$Xl
|
||||
b Loop
|
||||
|
||||
.align 5
|
||||
Loop:
|
||||
subic $len,$len,16
|
||||
vpmsumd $Xl,$IN,$Hl # H.lo·Xi.lo
|
||||
subfe. r0,r0,r0 # borrow?-1:0
|
||||
vpmsumd $Xm,$IN,$H # H.hi·Xi.lo+H.lo·Xi.hi
|
||||
and r0,r0,$len
|
||||
vpmsumd $Xh,$IN,$Hh # H.hi·Xi.hi
|
||||
add $inp,$inp,r0
|
||||
|
||||
vpmsumd $t2,$Xl,$xC2 # 1st phase
|
||||
|
||||
vsldoi $t0,$Xm,$zero,8
|
||||
vsldoi $t1,$zero,$Xm,8
|
||||
vxor $Xl,$Xl,$t0
|
||||
vxor $Xh,$Xh,$t1
|
||||
|
||||
vsldoi $Xl,$Xl,$Xl,8
|
||||
vxor $Xl,$Xl,$t2
|
||||
lvx_u $IN,0,$inp
|
||||
addi $inp,$inp,16
|
||||
|
||||
vsldoi $t1,$Xl,$Xl,8 # 2nd phase
|
||||
vpmsumd $Xl,$Xl,$xC2
|
||||
le?vperm $IN,$IN,$IN,$lemask
|
||||
vxor $t1,$t1,$Xh
|
||||
vxor $IN,$IN,$t1
|
||||
vxor $IN,$IN,$Xl
|
||||
beq Loop # did $len-=16 borrow?
|
||||
|
||||
vxor $Xl,$Xl,$t1
|
||||
le?vperm $Xl,$Xl,$Xl,$lemask
|
||||
stvx_u $Xl,0,$Xip # write out Xi
|
||||
|
||||
mtspr 256,$vrsave
|
||||
blr
|
||||
.long 0
|
||||
.byte 0,12,0x14,0,0,0,4,0
|
||||
.long 0
|
||||
.size .gcm_ghash_p8,.-.gcm_ghash_p8
|
||||
|
||||
.asciz "GHASH for PowerISA 2.07, CRYPTOGAMS by <appro\@openssl.org>"
|
||||
.align 2
|
||||
___
|
||||
|
||||
foreach (split("\n",$code)) {
|
||||
if ($flavour =~ /le$/o) { # little-endian
|
||||
s/le\?//o or
|
||||
s/be\?/#be#/o;
|
||||
} else {
|
||||
s/le\?/#le#/o or
|
||||
s/be\?//o;
|
||||
}
|
||||
print $_,"\n";
|
||||
}
|
||||
|
||||
close STDOUT; # enforce flush
|
||||
|
|
@ -0,0 +1,226 @@
|
|||
#!/usr/bin/env perl
|
||||
|
||||
# PowerPC assembler distiller by <appro>.
|
||||
|
||||
my $flavour = shift;
|
||||
my $output = shift;
|
||||
open STDOUT,">$output" || die "can't open $output: $!";
|
||||
|
||||
my %GLOBALS;
|
||||
my $dotinlocallabels=($flavour=~/linux/)?1:0;
|
||||
|
||||
################################################################
|
||||
# directives which need special treatment on different platforms
|
||||
################################################################
|
||||
my $globl = sub {
|
||||
my $junk = shift;
|
||||
my $name = shift;
|
||||
my $global = \$GLOBALS{$name};
|
||||
my $ret;
|
||||
|
||||
$name =~ s|^[\.\_]||;
|
||||
|
||||
SWITCH: for ($flavour) {
|
||||
/aix/ && do { $name = ".$name";
|
||||
last;
|
||||
};
|
||||
/osx/ && do { $name = "_$name";
|
||||
last;
|
||||
};
|
||||
/linux.*(32|64le)/
|
||||
&& do { $ret .= ".globl $name\n";
|
||||
$ret .= ".type $name,\@function";
|
||||
last;
|
||||
};
|
||||
/linux.*64/ && do { $ret .= ".globl $name\n";
|
||||
$ret .= ".type $name,\@function\n";
|
||||
$ret .= ".section \".opd\",\"aw\"\n";
|
||||
$ret .= ".align 3\n";
|
||||
$ret .= "$name:\n";
|
||||
$ret .= ".quad .$name,.TOC.\@tocbase,0\n";
|
||||
$ret .= ".previous\n";
|
||||
|
||||
$name = ".$name";
|
||||
last;
|
||||
};
|
||||
}
|
||||
|
||||
$ret = ".globl $name" if (!$ret);
|
||||
$$global = $name;
|
||||
$ret;
|
||||
};
|
||||
my $text = sub {
|
||||
my $ret = ($flavour =~ /aix/) ? ".csect\t.text[PR],7" : ".text";
|
||||
$ret = ".abiversion 2\n".$ret if ($flavour =~ /linux.*64le/);
|
||||
$ret;
|
||||
};
|
||||
my $machine = sub {
|
||||
my $junk = shift;
|
||||
my $arch = shift;
|
||||
if ($flavour =~ /osx/)
|
||||
{ $arch =~ s/\"//g;
|
||||
$arch = ($flavour=~/64/) ? "ppc970-64" : "ppc970" if ($arch eq "any");
|
||||
}
|
||||
".machine $arch";
|
||||
};
|
||||
my $size = sub {
|
||||
if ($flavour =~ /linux/)
|
||||
{ shift;
|
||||
my $name = shift; $name =~ s|^[\.\_]||;
|
||||
my $ret = ".size $name,.-".($flavour=~/64$/?".":"").$name;
|
||||
$ret .= "\n.size .$name,.-.$name" if ($flavour=~/64$/);
|
||||
$ret;
|
||||
}
|
||||
else
|
||||
{ ""; }
|
||||
};
|
||||
my $asciz = sub {
|
||||
shift;
|
||||
my $line = join(",",@_);
|
||||
if ($line =~ /^"(.*)"$/)
|
||||
{ ".byte " . join(",",unpack("C*",$1),0) . "\n.align 2"; }
|
||||
else
|
||||
{ ""; }
|
||||
};
|
||||
my $quad = sub {
|
||||
shift;
|
||||
my @ret;
|
||||
my ($hi,$lo);
|
||||
for (@_) {
|
||||
if (/^0x([0-9a-f]*?)([0-9a-f]{1,8})$/io)
|
||||
{ $hi=$1?"0x$1":"0"; $lo="0x$2"; }
|
||||
elsif (/^([0-9]+)$/o)
|
||||
{ $hi=$1>>32; $lo=$1&0xffffffff; } # error-prone with 32-bit perl
|
||||
else
|
||||
{ $hi=undef; $lo=$_; }
|
||||
|
||||
if (defined($hi))
|
||||
{ push(@ret,$flavour=~/le$/o?".long\t$lo,$hi":".long\t$hi,$lo"); }
|
||||
else
|
||||
{ push(@ret,".quad $lo"); }
|
||||
}
|
||||
join("\n",@ret);
|
||||
};
|
||||
|
||||
################################################################
|
||||
# simplified mnemonics not handled by at least one assembler
|
||||
################################################################
|
||||
my $cmplw = sub {
|
||||
my $f = shift;
|
||||
my $cr = 0; $cr = shift if ($#_>1);
|
||||
# Some out-of-date 32-bit GNU assembler just can't handle cmplw...
|
||||
($flavour =~ /linux.*32/) ?
|
||||
" .long ".sprintf "0x%x",31<<26|$cr<<23|$_[0]<<16|$_[1]<<11|64 :
|
||||
" cmplw ".join(',',$cr,@_);
|
||||
};
|
||||
my $bdnz = sub {
|
||||
my $f = shift;
|
||||
my $bo = $f=~/[\+\-]/ ? 16+9 : 16; # optional "to be taken" hint
|
||||
" bc $bo,0,".shift;
|
||||
} if ($flavour!~/linux/);
|
||||
my $bltlr = sub {
|
||||
my $f = shift;
|
||||
my $bo = $f=~/\-/ ? 12+2 : 12; # optional "not to be taken" hint
|
||||
($flavour =~ /linux/) ? # GNU as doesn't allow most recent hints
|
||||
" .long ".sprintf "0x%x",19<<26|$bo<<21|16<<1 :
|
||||
" bclr $bo,0";
|
||||
};
|
||||
my $bnelr = sub {
|
||||
my $f = shift;
|
||||
my $bo = $f=~/\-/ ? 4+2 : 4; # optional "not to be taken" hint
|
||||
($flavour =~ /linux/) ? # GNU as doesn't allow most recent hints
|
||||
" .long ".sprintf "0x%x",19<<26|$bo<<21|2<<16|16<<1 :
|
||||
" bclr $bo,2";
|
||||
};
|
||||
my $beqlr = sub {
|
||||
my $f = shift;
|
||||
my $bo = $f=~/-/ ? 12+2 : 12; # optional "not to be taken" hint
|
||||
($flavour =~ /linux/) ? # GNU as doesn't allow most recent hints
|
||||
" .long ".sprintf "0x%X",19<<26|$bo<<21|2<<16|16<<1 :
|
||||
" bclr $bo,2";
|
||||
};
|
||||
# GNU assembler can't handle extrdi rA,rS,16,48, or when sum of last two
|
||||
# arguments is 64, with "operand out of range" error.
|
||||
my $extrdi = sub {
|
||||
my ($f,$ra,$rs,$n,$b) = @_;
|
||||
$b = ($b+$n)&63; $n = 64-$n;
|
||||
" rldicl $ra,$rs,$b,$n";
|
||||
};
|
||||
my $vmr = sub {
|
||||
my ($f,$vx,$vy) = @_;
|
||||
" vor $vx,$vy,$vy";
|
||||
};
|
||||
|
||||
# PowerISA 2.06 stuff
|
||||
sub vsxmem_op {
|
||||
my ($f, $vrt, $ra, $rb, $op) = @_;
|
||||
" .long ".sprintf "0x%X",(31<<26)|($vrt<<21)|($ra<<16)|($rb<<11)|($op*2+1);
|
||||
}
|
||||
# made-up unaligned memory reference AltiVec/VMX instructions
|
||||
my $lvx_u = sub { vsxmem_op(@_, 844); }; # lxvd2x
|
||||
my $stvx_u = sub { vsxmem_op(@_, 972); }; # stxvd2x
|
||||
my $lvdx_u = sub { vsxmem_op(@_, 588); }; # lxsdx
|
||||
my $stvdx_u = sub { vsxmem_op(@_, 716); }; # stxsdx
|
||||
my $lvx_4w = sub { vsxmem_op(@_, 780); }; # lxvw4x
|
||||
my $stvx_4w = sub { vsxmem_op(@_, 908); }; # stxvw4x
|
||||
|
||||
# PowerISA 2.07 stuff
|
||||
sub vcrypto_op {
|
||||
my ($f, $vrt, $vra, $vrb, $op) = @_;
|
||||
" .long ".sprintf "0x%X",(4<<26)|($vrt<<21)|($vra<<16)|($vrb<<11)|$op;
|
||||
}
|
||||
my $vcipher = sub { vcrypto_op(@_, 1288); };
|
||||
my $vcipherlast = sub { vcrypto_op(@_, 1289); };
|
||||
my $vncipher = sub { vcrypto_op(@_, 1352); };
|
||||
my $vncipherlast= sub { vcrypto_op(@_, 1353); };
|
||||
my $vsbox = sub { vcrypto_op(@_, 0, 1480); };
|
||||
my $vshasigmad = sub { my ($st,$six)=splice(@_,-2); vcrypto_op(@_, $st<<4|$six, 1730); };
|
||||
my $vshasigmaw = sub { my ($st,$six)=splice(@_,-2); vcrypto_op(@_, $st<<4|$six, 1666); };
|
||||
my $vpmsumb = sub { vcrypto_op(@_, 1032); };
|
||||
my $vpmsumd = sub { vcrypto_op(@_, 1224); };
|
||||
my $vpmsubh = sub { vcrypto_op(@_, 1096); };
|
||||
my $vpmsumw = sub { vcrypto_op(@_, 1160); };
|
||||
my $vaddudm = sub { vcrypto_op(@_, 192); };
|
||||
|
||||
my $mtsle = sub {
|
||||
my ($f, $arg) = @_;
|
||||
" .long ".sprintf "0x%X",(31<<26)|($arg<<21)|(147*2);
|
||||
};
|
||||
|
||||
while($line=<>) {
|
||||
|
||||
$line =~ s|[#!;].*$||; # get rid of asm-style comments...
|
||||
$line =~ s|/\*.*\*/||; # ... and C-style comments...
|
||||
$line =~ s|^\s+||; # ... and skip white spaces in beginning...
|
||||
$line =~ s|\s+$||; # ... and at the end
|
||||
|
||||
{
|
||||
$line =~ s|\b\.L(\w+)|L$1|g; # common denominator for Locallabel
|
||||
$line =~ s|\bL(\w+)|\.L$1|g if ($dotinlocallabels);
|
||||
}
|
||||
|
||||
{
|
||||
$line =~ s|(^[\.\w]+)\:\s*||;
|
||||
my $label = $1;
|
||||
if ($label) {
|
||||
printf "%s:",($GLOBALS{$label} or $label);
|
||||
printf "\n.localentry\t$GLOBALS{$label},0" if ($GLOBALS{$label} && $flavour =~ /linux.*64le/);
|
||||
}
|
||||
}
|
||||
|
||||
{
|
||||
$line =~ s|^\s*(\.?)(\w+)([\.\+\-]?)\s*||;
|
||||
my $c = $1; $c = "\t" if ($c eq "");
|
||||
my $mnemonic = $2;
|
||||
my $f = $3;
|
||||
my $opcode = eval("\$$mnemonic");
|
||||
$line =~ s/\b(c?[rf]|v|vs)([0-9]+)\b/$2/g if ($c ne "." and $flavour !~ /osx/);
|
||||
if (ref($opcode) eq 'CODE') { $line = &$opcode($f,split(',',$line)); }
|
||||
elsif ($mnemonic) { $line = $c.$mnemonic.$f."\t".$line; }
|
||||
}
|
||||
|
||||
print $line if ($line);
|
||||
print "\n";
|
||||
}
|
||||
|
||||
close STDOUT;
|
||||
Loading…
Reference in New Issue