From a43ce912fc025d11e1395506111f75fc194d7ba5 Mon Sep 17 00:00:00 2001
From: Jung-uk Kim <jkim@FreeBSD.org>
Date: Thu, 13 Sep 2018 19:18:07 +0000
Subject: Import OpenSSL 1.1.1.

---
 crypto/md5/asm/md5-x86_64.pl | 55 +++++++++++++++++++++++++++++++-------------
 1 file changed, 39 insertions(+), 16 deletions(-)

(limited to 'crypto/md5/asm/md5-x86_64.pl')

diff --git a/crypto/md5/asm/md5-x86_64.pl b/crypto/md5/asm/md5-x86_64.pl
index 381bf77e1c6b..386d8048ec02 100755
--- a/crypto/md5/asm/md5-x86_64.pl
+++ b/crypto/md5/asm/md5-x86_64.pl
@@ -1,11 +1,13 @@
-#!/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 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
+
+# MD5 optimized for AMD64.
 
 use strict;
 
@@ -25,8 +27,8 @@ sub round1_step
 	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' = $y */
@@ -43,13 +45,12 @@ EOF
 sub round2_step
 {
     my ($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) */
@@ -66,21 +67,31 @@ EOF
 #   %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:
@@ -91,16 +102,15 @@ EOF
 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 */
@@ -120,7 +130,7 @@ $0 =~ m/(.*[\/\\])[^\/\\]+$/; my $dir=$1; my $xlate;
 ( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or
 die "can't locate x86_64-xlate.pl";
 
-open OUT,"| \"$^X\" $xlate $flavour $output";
+open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\"";
 *STDOUT=*OUT;
 
 $code .= <<EOF;
@@ -130,11 +140,17 @@ $code .= <<EOF;
 .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)
@@ -179,7 +195,7 @@ round1_step( 0,'%ebx','%ecx','%edx','%eax','12','0x895cd7be','22');
 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');
@@ -196,7 +212,7 @@ round2_step( 0,'%ebx','%ecx','%edx','%eax','13','0x455a14ed','20');
 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');
@@ -251,13 +267,20 @@ $code .= <<EOF;
 	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
 
-- 
cgit v1.2.3