-#!/usr/bin/perl -w
-#
-# MD5 optimized for AMD64.
-#
+#! /usr/bin/env perl
# Author: Marc Bevand <bevand_m (at) epita.fr>
-# Licence: I hereby disclaim the copyright on this code and place it
-# in the public domain.
+# Copyright 2005-2016 The OpenSSL Project Authors. All Rights Reserved.
#
+# Licensed under the Apache License 2.0 (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
+
+# MD5 optimized for AMD64.
use strict;
xor $y, %r11d /* y ^ ... */
lea $T_i($dst,%r10d),$dst /* Const + dst + ... */
and $x, %r11d /* x & ... */
- xor $z, %r11d /* z ^ ... */
mov $k_next*4(%rsi),%r10d /* (NEXT STEP) X[$k_next] */
+ xor $z, %r11d /* z ^ ... */
add %r11d, $dst /* dst += ... */
rol \$$s, $dst /* dst <<< s */
mov $y, %r11d /* (NEXT STEP) z' = $y */
sub round2_step
{
my ($pos, $dst, $x, $y, $z, $k_next, $T_i, $s) = @_;
- $code .= " mov 1*4(%rsi), %r10d /* (NEXT STEP) X[1] */\n" if ($pos == -1);
$code .= " mov %edx, %r11d /* (NEXT STEP) z' = %edx */\n" if ($pos == -1);
$code .= " mov %edx, %r12d /* (NEXT STEP) z' = %edx */\n" if ($pos == -1);
$code .= <<EOF;
not %r11d /* not z */
- lea $T_i($dst,%r10d),$dst /* Const + dst + ... */
and $x, %r12d /* x & z */
+ lea $T_i($dst,%r10d),$dst /* Const + dst + ... */
and $y, %r11d /* y & (not z) */
mov $k_next*4(%rsi),%r10d /* (NEXT STEP) X[$k_next] */
or %r11d, %r12d /* (y & (not z)) | (x & z) */
# %r10d = X[k_next]
# %r11d = y' (copy of y for the next step)
# Each round3_step() takes about 4.2 clocks (8 instructions, 1.9 IPC)
+{ my $round3_alter=0;
sub round3_step
{
my ($pos, $dst, $x, $y, $z, $k_next, $T_i, $s) = @_;
- $code .= " mov 5*4(%rsi), %r10d /* (NEXT STEP) X[5] */\n" if ($pos == -1);
$code .= " mov %ecx, %r11d /* (NEXT STEP) y' = %ecx */\n" if ($pos == -1);
$code .= <<EOF;
lea $T_i($dst,%r10d),$dst /* Const + dst + ... */
- mov $k_next*4(%rsi),%r10d /* (NEXT STEP) X[$k_next] */
xor $z, %r11d /* z ^ ... */
+ mov $k_next*4(%rsi),%r10d /* (NEXT STEP) X[$k_next] */
xor $x, %r11d /* x ^ ... */
add %r11d, $dst /* dst += ... */
+EOF
+ $code .= <<EOF if ($round3_alter);
rol \$$s, $dst /* dst <<< s */
mov $x, %r11d /* (NEXT STEP) y' = $x */
+EOF
+ $code .= <<EOF if (!$round3_alter);
+ mov $x, %r11d /* (NEXT STEP) y' = $x */
+ rol \$$s, $dst /* dst <<< s */
+EOF
+ $code .= <<EOF;
add $x, $dst /* dst += x */
EOF
+ $round3_alter^=1;
+}
}
# round4_step() does:
sub round4_step
{
my ($pos, $dst, $x, $y, $z, $k_next, $T_i, $s) = @_;
- $code .= " mov 0*4(%rsi), %r10d /* (NEXT STEP) X[0] */\n" if ($pos == -1);
$code .= " mov \$0xffffffff, %r11d\n" if ($pos == -1);
$code .= " xor %edx, %r11d /* (NEXT STEP) not z' = not %edx*/\n"
if ($pos == -1);
$code .= <<EOF;
lea $T_i($dst,%r10d),$dst /* Const + dst + ... */
or $x, %r11d /* x | ... */
+ mov $k_next*4(%rsi),%r10d /* (NEXT STEP) X[$k_next] */
xor $y, %r11d /* y ^ ... */
add %r11d, $dst /* dst += ... */
- mov $k_next*4(%rsi),%r10d /* (NEXT STEP) X[$k_next] */
mov \$0xffffffff, %r11d
rol \$$s, $dst /* dst <<< s */
xor $y, %r11d /* (NEXT STEP) not z' = not $y */
EOF
}
-my $flavour = shift;
-my $output = shift;
-if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
+no warnings qw(uninitialized);
+# $output is the last argument if it looks like a file (it has an extension)
+# $flavour is the first argument if it doesn't look like a file
+my $output = $#ARGV >= 0 && $ARGV[$#ARGV] =~ m|\.\w+$| ? pop : undef;
+my $flavour = $#ARGV >= 0 && $ARGV[0] !~ m|\.| ? shift : undef;
my $win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or
die "can't locate x86_64-xlate.pl";
-no warnings qw(uninitialized);
-open STDOUT,"| $^X $xlate $flavour $output";
+open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\""
+ or die "can't call $xlate: $!";
+*STDOUT=*OUT;
$code .= <<EOF;
.text
.globl md5_block_asm_data_order
.type md5_block_asm_data_order,\@function,3
md5_block_asm_data_order:
+.cfi_startproc
push %rbp
+.cfi_push %rbp
push %rbx
+.cfi_push %rbx
push %r12
+.cfi_push %r12
push %r14
+.cfi_push %r14
push %r15
+.cfi_push %r15
.Lprologue:
# rdi = arg #1 (ctx, MD5_CTX pointer)
round1_step( 0,'%eax','%ebx','%ecx','%edx','13','0x6b901122', '7');
round1_step( 0,'%edx','%eax','%ebx','%ecx','14','0xfd987193','12');
round1_step( 0,'%ecx','%edx','%eax','%ebx','15','0xa679438e','17');
-round1_step( 1,'%ebx','%ecx','%edx','%eax', '0','0x49b40821','22');
+round1_step( 1,'%ebx','%ecx','%edx','%eax', '1','0x49b40821','22');
round2_step(-1,'%eax','%ebx','%ecx','%edx', '6','0xf61e2562', '5');
round2_step( 0,'%edx','%eax','%ebx','%ecx','11','0xc040b340', '9');
round2_step( 0,'%eax','%ebx','%ecx','%edx', '2','0xa9e3e905', '5');
round2_step( 0,'%edx','%eax','%ebx','%ecx', '7','0xfcefa3f8', '9');
round2_step( 0,'%ecx','%edx','%eax','%ebx','12','0x676f02d9','14');
-round2_step( 1,'%ebx','%ecx','%edx','%eax', '0','0x8d2a4c8a','20');
+round2_step( 1,'%ebx','%ecx','%edx','%eax', '5','0x8d2a4c8a','20');
round3_step(-1,'%eax','%ebx','%ecx','%edx', '8','0xfffa3942', '4');
round3_step( 0,'%edx','%eax','%ebx','%ecx','11','0x8771f681','11');
mov %edx, 3*4(%rbp) # ctx->D = D
mov (%rsp),%r15
+.cfi_restore %r15
mov 8(%rsp),%r14
+.cfi_restore %r14
mov 16(%rsp),%r12
+.cfi_restore %r12
mov 24(%rsp),%rbx
+.cfi_restore %rbx
mov 32(%rsp),%rbp
+.cfi_restore %rbp
add \$40,%rsp
+.cfi_adjust_cfa_offset -40
.Lepilogue:
ret
+.cfi_endproc
.size md5_block_asm_data_order,.-md5_block_asm_data_order
EOF
print $code;
-close STDOUT;
+close STDOUT or die "error closing STDOUT";