Merge remote-tracking branch 'origin/master'

# Conflicts:
#	README.md

crypto/rc4/asm/rc4-586.pl

@@ -1,4 +1,11 @@
-#!/usr/bin/env perl
+#! /usr/bin/env perl
+# Copyright 1998-2016 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
+
 
 # ====================================================================
 # [Re]written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
@@ -43,6 +50,9 @@
 # Westmere	5.1/+94%(**)
 # Sandy Bridge	5.0/+8%
 # Atom		12.6/+6%
+# VIA Nano	6.4/+9%
+# Ivy Bridge	4.9/±0%
+# Bulldozer	4.9/+15%
 #
 # (*)	PIII can actually deliver 6.6 cycles per byte with MMX code,
 #	but this specific code performs poorly on Core2. And vice
@@ -60,6 +70,9 @@ $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
 push(@INC,"${dir}","${dir}../../perlasm");
 require "x86asm.pl";
 
+$output=pop;
+open STDOUT,">$output";
+
 &asm_init($ARGV[0],"rc4-586.pl",$x86only = $ARGV[$#ARGV] eq "386");
 
 $xx="eax";
@@ -144,7 +157,7 @@ if ($alt=0) {
 	&movd	($i>0?"mm1":"mm2",&DWP(0,$dat,$ty,4));
 
 	# (*)	This is the key to Core2 and Westmere performance.
-	#	Whithout movz out-of-order execution logic confuses
+	#	Without movz out-of-order execution logic confuses
 	#	itself and fails to reorder loads and stores. Problem
 	#	appears to be fixed in Sandy Bridge...
   }
@@ -304,7 +317,7 @@ $ido="ecx";
 $idx="edx";
 
 # void RC4_set_key(RC4_KEY *key,int len,const unsigned char *data);
-&function_begin("private_RC4_set_key");
+&function_begin("RC4_set_key");
 	&mov	($out,&wparam(0));		# load key
 	&mov	($idi,&wparam(1));		# load len
 	&mov	($inp,&wparam(2));		# load data
@@ -382,7 +395,7 @@ $idx="edx";
 	&xor	("eax","eax");
 	&mov	(&DWP(-8,$out),"eax");		# key->x=0;
 	&mov	(&DWP(-4,$out),"eax");		# key->y=0;
-&function_end("private_RC4_set_key");
+&function_end("RC4_set_key");
 
 # const char *RC4_options(void);
 &function_begin_B("RC4_options");
@@ -412,3 +425,4 @@ $idx="edx";
 
 &asm_finish();
 
+close STDOUT;

crypto/rc4/asm/rc4-c64xplus.pl

@@ -0,0 +1,190 @@
+#! /usr/bin/env perl
+# Copyright 2014-2016 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@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/.
+# ====================================================================
+#
+# RC4 for C64x+.
+#
+# April 2014
+#
+# RC4 subroutine processes one byte in 7.0 cycles, which is 3x faster
+# than TI CGT-generated code. Loop is scheduled in such way that
+# there is only one reference to memory in each cycle. This is done
+# to avoid L1D memory banking conflicts, see SPRU871 TI publication
+# for further details. Otherwise it should be possible to schedule
+# the loop for iteration interval of 6...
+
+($KEY,$LEN,$INP,$OUT)=("A4","B4","A6","B6");
+
+($KEYA,$XX,$TY,$xx,$ONE,$ret)=map("A$_",(5,7,8,9,1,2));
+($KEYB,$YY,$TX,$tx,$SUM,$dat)=map("B$_",(5,7,8,9,1,2));
+
+$code.=<<___;
+	.text
+
+	.if	.ASSEMBLER_VERSION<7000000
+	.asg	0,__TI_EABI__
+	.endif
+	.if	__TI_EABI__
+	.nocmp
+	.asg	RC4,_RC4
+	.asg	RC4_set_key,_RC4_set_key
+	.asg	RC4_options,_RC4_options
+	.endif
+
+	.global	_RC4
+	.align	16
+_RC4:
+	.asmfunc
+	MV	$LEN,B0
+  [!B0]	BNOP	B3			; if (len==0) return;
+||[B0]	ADD	$KEY,2,$KEYA
+||[B0]	ADD	$KEY,2,$KEYB
+  [B0]	MVK	1,$ONE
+||[B0]	LDBU	*${KEYA}[-2],$XX	; key->x
+  [B0]	LDBU	*${KEYB}[-1],$YY	; key->y
+||	NOP	4
+
+	ADD4	$ONE,$XX,$XX
+	LDBU	*${KEYA}[$XX],$TX
+||	MVC	$LEN,ILC
+	NOP	4
+;;==================================================
+	SPLOOP	7
+||	ADD4	$TX,$YY,$YY
+
+	LDBU	*${KEYB}[$YY],$TY
+||	MVD	$XX,$xx
+||	ADD4	$ONE,$XX,$XX
+	LDBU	*${KEYA}[$XX],$tx
+	CMPEQ	$YY,$XX,B0
+||	NOP	3
+	STB	$TX,*${KEYB}[$YY]
+||[B0]	ADD4	$TX,$YY,$YY
+	STB	$TY,*${KEYA}[$xx]
+||[!B0]	ADD4	$tx,$YY,$YY
+||[!B0]	MVD	$tx,$TX
+	ADD4	$TY,$TX,$SUM		; [0,0] $TX is not replaced by $tx yet!
+||	NOP	2
+	LDBU	*$INP++,$dat
+||	NOP	2
+	LDBU	*${KEYB}[$SUM],$ret
+||	NOP	5
+	XOR.L	$dat,$ret,$ret
+	SPKERNEL
+||	STB	$ret,*$OUT++
+;;==================================================
+	SUB4	$XX,$ONE,$XX
+||	NOP	5
+	STB	$XX,*${KEYA}[-2]	; key->x
+||	SUB4	$YY,$TX,$YY
+||	BNOP	B3	
+	STB	$YY,*${KEYB}[-1]	; key->y
+||	NOP	5
+	.endasmfunc
+
+	.global	_RC4_set_key
+	.align	16
+_RC4_set_key:
+	.asmfunc
+	.if	.BIG_ENDIAN
+	MVK	0x00000404,$ONE
+||	MVK	0x00000203,B0
+	MVKH	0x04040000,$ONE
+||	MVKH	0x00010000,B0
+	.else
+	MVK	0x00000404,$ONE
+||	MVK	0x00000100,B0
+	MVKH	0x04040000,$ONE
+||	MVKH	0x03020000,B0
+	.endif
+	ADD	$KEY,2,$KEYA
+||	ADD	$KEY,2,$KEYB
+||	ADD	$INP,$LEN,$ret		; end of input
+	LDBU	*${INP}++,$dat
+||	MVK	0,$TX
+	STH	$TX,*${KEY}++		; key->x=key->y=0
+||	MV	B0,A0
+||	MVK	64-4,B0
+
+;;==================================================
+	SPLOOPD	1
+||	MVC	B0,ILC
+
+	STNW	A0,*${KEY}++
+||	ADD4	$ONE,A0,A0
+	SPKERNEL
+;;==================================================
+
+	MVK	0,$YY
+||	MVK	0,$XX
+	MVK	1,$ONE
+||	MVK	256-1,B0
+
+;;==================================================
+	SPLOOPD	8
+||	MVC	B0,ILC
+
+	ADD4	$dat,$YY,$YY
+||	CMPEQ	$INP,$ret,A0		; end of input?
+	LDBU	*${KEYB}[$YY],$TY
+||	MVD	$XX,$xx
+||	ADD4	$ONE,$XX,$XX
+	LDBU	*${KEYA}[$XX],$tx
+||[A0]	SUB	$INP,$LEN,$INP		; rewind
+	LDBU	*${INP}++,$dat
+||	CMPEQ	$YY,$XX,B0
+||	NOP	3
+	STB	$TX,*${KEYB}[$YY]
+||[B0]	ADD4	$TX,$YY,$YY
+	STB	$TY,*${KEYA}[$xx]
+||[!B0]	ADD4	$tx,$YY,$YY
+||[!B0]	MV	$tx,$TX
+	SPKERNEL
+;;==================================================
+
+	BNOP	B3,5
+	.endasmfunc
+
+	.global	_RC4_options
+	.align	16
+_RC4_options:
+_rc4_options:
+	.asmfunc
+	BNOP	B3,1
+	ADDKPC	_rc4_options,B4
+	.if	__TI_EABI__
+	MVKL	\$PCR_OFFSET(rc4_options,_rc4_options),A4
+	MVKH	\$PCR_OFFSET(rc4_options,_rc4_options),A4
+	.else
+	MVKL	(rc4_options-_rc4_options),A4
+	MVKH	(rc4_options-_rc4_options),A4
+	.endif
+	ADD	B4,A4,A4
+	.endasmfunc
+
+	.if	__TI_EABI__
+	.sect	".text:rc4_options.const"
+	.else
+	.sect	".const:rc4_options"
+	.endif
+	.align	4
+rc4_options:
+	.cstring "rc4(sploop,char)"
+	.cstring "RC4 for C64+, CRYPTOGAMS by <appro\@openssl.org>"
+	.align	4
+___
+
+print $code;
+close STDOUT;

crypto/rc4/asm/rc4-ia64.pl

@@ -1,4 +1,11 @@
-#!/usr/bin/env perl
+#! /usr/bin/env perl
+# 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
+
 #
 # ====================================================================
 # Written by David Mosberger <David.Mosberger@acm.org> based on the
@@ -164,6 +171,9 @@
 #     random input data).
 #
 
+$output = pop;
+open STDOUT,">$output";
+
 $phases = 4;		# number of stages/phases in the pipelined-loop
 $unroll_count = 6;	# number of times we unrolled it
 $pComI = (1 << 0);
@@ -753,3 +763,5 @@ $code.=<<___;
 ___
 
 print $code;
+
+close STDOUT;

crypto/rc4/asm/rc4-md5-x86_64.pl

@@ -1,4 +1,11 @@
-#!/usr/bin/env perl
+#! /usr/bin/env perl
+# Copyright 2011-2016 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@openssl.org> for the OpenSSL
@@ -29,10 +36,16 @@
 # Core2		6.5	5.8	12.3	7.7	+60%
 # Westmere	4.3	5.2	9.5	7.0	+36%
 # Sandy Bridge	4.2	5.5	9.7	6.8	+43%
+# Ivy Bridge	4.1	5.2	9.3	6.0	+54%
+# Haswell	4.0	5.0	9.0	5.7	+60%
+# Skylake	6.3(**)	5.0	11.3	5.3	+110%
 # Atom		9.3	6.5	15.8	11.1	+42%
+# VIA Nano	6.3	5.4	11.7	8.6	+37%
+# Bulldozer	4.5	5.4	9.9	7.7	+29%
 #
 # (*)	rc4-x86_64.pl delivers 5.3 on Opteron, so real improvement
 #	is +53%...
+# (**)	unidentified anomaly;
 
 my ($rc4,$md5)=(1,1);	# what to generate?
 my $D="#" if (!$md5);	# if set to "#", MD5 is stitched into RC4(),
@@ -51,7 +64,7 @@ $0 =~ m/(.*[\/\\])[^\/\\]+$/; 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;
 
 my ($dat,$in0,$out,$ctx,$inp,$len, $func,$nargs);

crypto/rc4/asm/rc4-parisc.pl

@@ -1,4 +1,11 @@
-#!/usr/bin/env perl
+#! /usr/bin/env perl
+# Copyright 2009-2016 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
@@ -243,9 +250,9 @@ ___
 
 $code.=<<___;
 
-	.EXPORT	private_RC4_set_key,ENTRY,ARGW0=GR,ARGW1=GR,ARGW2=GR
+	.EXPORT	RC4_set_key,ENTRY,ARGW0=GR,ARGW1=GR,ARGW2=GR
 	.ALIGN	8
-private_RC4_set_key
+RC4_set_key
 	.PROC
 	.CALLINFO	NO_CALLS
 	.ENTRY

crypto/rc4/asm/rc4-s390x.pl

@@ -1,4 +1,11 @@
-#!/usr/bin/env perl
+#! /usr/bin/env perl
+# Copyright 2009-2016 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
@@ -33,7 +40,7 @@ if ($flavour =~ /3[12]/) {
 	$g="g";
 }
 
-while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {}
+while (($output=shift) && ($output!~/\w[\w\-]*\.\w+$/)) {}
 open STDOUT,">$output";
 
 $rp="%r14";
@@ -171,10 +178,10 @@ $ikey="%r7";
 $iinp="%r8";
 
 $code.=<<___;
-.globl	private_RC4_set_key
-.type	private_RC4_set_key,\@function
+.globl	RC4_set_key
+.type	RC4_set_key,\@function
 .align	64
-private_RC4_set_key:
+RC4_set_key:
 	stm${g}	%r6,%r8,6*$SIZE_T($sp)
 	lhi	$cnt,256
 	la	$idx,0(%r0)
@@ -210,7 +217,7 @@ private_RC4_set_key:
 .Ldone:
 	lm${g}	%r6,%r8,6*$SIZE_T($sp)
 	br	$rp
-.size	private_RC4_set_key,.-private_RC4_set_key
+.size	RC4_set_key,.-RC4_set_key
 
 ___
 }

crypto/rc4/asm/rc4-x86_64.pl

@@ -1,4 +1,11 @@
-#!/usr/bin/env perl
+#! /usr/bin/env perl
+# 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
+
 #
 # ====================================================================
 # Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
@@ -50,13 +57,13 @@
 # As was shown by Zou Nanhai loop unrolling can improve Intel EM64T
 # performance by >30% [unlike P4 32-bit case that is]. But this is
 # provided that loads are reordered even more aggressively! Both code
-# pathes, AMD64 and EM64T, reorder loads in essentially same manner
+# paths, AMD64 and EM64T, reorder loads in essentially same manner
 # as my IA-64 implementation. On Opteron this resulted in modest 5%
 # improvement [I had to test it], while final Intel P4 performance
 # achieves respectful 432MBps on 2.8GHz processor now. For reference.
 # If executed on Xeon, current RC4_CHAR code-path is 2.7x faster than
 # RC4_INT code-path. While if executed on Opteron, it's only 25%
-# slower than the RC4_INT one [meaning that if CPU µ-arch detection
+# slower than the RC4_INT one [meaning that if CPU µ-arch detection
 # is not implemented, then this final RC4_CHAR code-path should be
 # preferred, as it provides better *all-round* performance].
 
@@ -92,6 +99,9 @@
 # Westmere	4.2/+60%
 # Sandy Bridge	4.2/+120%
 # Atom		9.3/+80%
+# VIA Nano	6.4/+4%
+# Ivy Bridge	4.1/+30%
+# Bulldozer	4.5/+30%(*)
 #
 # (*)	But corresponding loop has less instructions, which should have
 #	positive effect on upcoming Bulldozer, which has one less ALU.
@@ -112,7 +122,7 @@ $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
 ( $xlate="${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;
 
 $dat="%rdi";	    # arg1
@@ -430,10 +440,10 @@ $idx="%r8";
 $ido="%r9";
 
 $code.=<<___;
-.globl	private_RC4_set_key
-.type	private_RC4_set_key,\@function,3
+.globl	RC4_set_key
+.type	RC4_set_key,\@function,3
 .align	16
-private_RC4_set_key:
+RC4_set_key:
 	lea	8($dat),$dat
 	lea	($inp,$len),$inp
 	neg	$len
@@ -500,7 +510,7 @@ private_RC4_set_key:
 	mov	%eax,-8($dat)
 	mov	%eax,-4($dat)
 	ret
-.size	private_RC4_set_key,.-private_RC4_set_key
+.size	RC4_set_key,.-RC4_set_key
 
 .globl	RC4_options
 .type	RC4_options,\@abi-omnipotent
@@ -645,16 +655,16 @@ key_se_handler:
 	.rva	.LSEH_end_RC4
 	.rva	.LSEH_info_RC4
 
-	.rva	.LSEH_begin_private_RC4_set_key
-	.rva	.LSEH_end_private_RC4_set_key
-	.rva	.LSEH_info_private_RC4_set_key
+	.rva	.LSEH_begin_RC4_set_key
+	.rva	.LSEH_end_RC4_set_key
+	.rva	.LSEH_info_RC4_set_key
 
 .section	.xdata
 .align	8
 .LSEH_info_RC4:
 	.byte	9,0,0,0
 	.rva	stream_se_handler
-.LSEH_info_private_RC4_set_key:
+.LSEH_info_RC4_set_key:
 	.byte	9,0,0,0
 	.rva	key_se_handler
 ___