X-Git-Url: https://git.librecmc.org/?a=blobdiff_plain;ds=sidebyside;f=crypto%2Fsha%2Fasm%2Fsha512-x86_64.pl;h=f2ebdfdb68b64e4341da3e192186cd2f7bdd8f27;hb=0d7903f83f84bba1d29225efd999c633a0c5ba01;hp=aec7b4568c0deb6b05cc81f2635941e616d1ce11;hpb=340daf6a06aca2bf9a694eee091df42eddfae5cf;p=oweals%2Fopenssl.git

diff --git a/crypto/sha/asm/sha512-x86_64.pl b/crypto/sha/asm/sha512-x86_64.pl
index aec7b4568c..f2ebdfdb68 100755
--- a/crypto/sha/asm/sha512-x86_64.pl
+++ b/crypto/sha/asm/sha512-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@openssl.org> for the OpenSSL
@@ -34,7 +41,7 @@
 # level parallelism, on a given CPU implementation in this case.
 #
 # Special note on Intel EM64T. While Opteron CPU exhibits perfect
-# perfromance ratio of 1.5 between 64- and 32-bit flavors [see above],
+# performance ratio of 1.5 between 64- and 32-bit flavors [see above],
 # [currently available] EM64T CPUs apparently are far from it. On the
 # contrary, 64-bit version, sha512_block, is ~30% *slower* than 32-bit
 # sha256_block:-( This is presumably because 64-bit shifts/rotates
@@ -86,11 +93,16 @@
 # Sandy Bridge	17.4	14.2(+23%)  11.6(+50%(**))  11.2    8.10(+38%(**))
 # Ivy Bridge	12.6	10.5(+20%)  10.3(+22%)	    8.17    7.22(+13%)
 # Haswell	12.2	9.28(+31%)  7.80(+56%)	    7.66    5.40(+42%)
+# Skylake	11.4	9.03(+26%)  7.70(+48%)      7.25    5.20(+40%)
 # Bulldozer	21.1	13.6(+54%)  13.6(+54%(***)) 13.5    8.58(+57%)
+# Ryzen		11.0	9.02(+22%)  2.05(+440%)     7.05    5.67(+20%)
 # VIA Nano	23.0	16.5(+39%)  -		    14.7    -
 # Atom		23.0	18.9(+22%)  -		    14.7    -
+# Silvermont	27.4	20.6(+33%)  -               17.5    -
+# Knights L	27.4	21.0(+30%)  19.6(+40%)	    17.5    12.8(+37%)
+# Goldmont	18.9	14.3(+32%)  4.16(+350%)     12.0    -
 #
-# (*)	whichever best applicable;
+# (*)	whichever best applicable, including SHAEXT;
 # (**)	switch from ror to shrd stands for fair share of improvement;
 # (***)	execution time is fully determined by remaining integer-only
 #	part, body_00_15; reducing the amount of SIMD instructions
@@ -123,14 +135,14 @@ if (!$avx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
 	$avx = ($1>=10) + ($1>=11);
 }
 
-if (!$avx && `$ENV{CC} -v 2>&1` =~ /(^clang version|based on LLVM) ([3-9]\.[0-9]+)/) {
+if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|.*based on LLVM) ([3-9]\.[0-9]+)/) {
 	$avx = ($2>=3.0) + ($2>3.0);
 }
 
 $shaext=1;	### set to zero if compiling for 1.0.1
 $avx=1		if (!$shaext && $avx);
 
-open OUT,"| \"$^X\" $xlate $flavour $output";
+open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\"";
 *STDOUT=*OUT;
 
 if ($output =~ /512/) {
@@ -166,7 +178,7 @@ $Tbl="%rbp";
 $_ctx="16*$SZ+0*8(%rsp)";
 $_inp="16*$SZ+1*8(%rsp)";
 $_end="16*$SZ+2*8(%rsp)";
-$_rsp="16*$SZ+3*8(%rsp)";
+$_rsp="`16*$SZ+3*8`(%rsp)";
 $framesz="16*$SZ+4*8";
 
 
@@ -259,6 +271,7 @@ $code=<<___;
 .type	$func,\@function,3
 .align	16
 $func:
+.cfi_startproc
 ___
 $code.=<<___ if ($SZ==4 || $avx);
 	lea	OPENSSL_ia32cap_P(%rip),%r11
@@ -291,13 +304,20 @@ $code.=<<___ if ($SZ==4);
 	jnz	.Lssse3_shortcut
 ___
 $code.=<<___;
+	mov	%rsp,%rax		# copy %rsp
+.cfi_def_cfa_register	%rax
 	push	%rbx
+.cfi_push	%rbx
 	push	%rbp
+.cfi_push	%rbp
 	push	%r12
+.cfi_push	%r12
 	push	%r13
+.cfi_push	%r13
 	push	%r14
+.cfi_push	%r14
 	push	%r15
-	mov	%rsp,%r11		# copy %rsp
+.cfi_push	%r15
 	shl	\$4,%rdx		# num*16
 	sub	\$$framesz,%rsp
 	lea	($inp,%rdx,$SZ),%rdx	# inp+num*16*$SZ
@@ -305,7 +325,8 @@ $code.=<<___;
 	mov	$ctx,$_ctx		# save ctx, 1st arg
 	mov	$inp,$_inp		# save inp, 2nd arh
 	mov	%rdx,$_end		# save end pointer, "3rd" arg
-	mov	%r11,$_rsp		# save copy of %rsp
+	mov	%rax,$_rsp		# save copy of %rsp
+.cfi_cfa_expression	$_rsp,deref,+8
 .Lprologue:
 
 	mov	$SZ*0($ctx),$A
@@ -372,15 +393,24 @@ $code.=<<___;
 	jb	.Lloop
 
 	mov	$_rsp,%rsi
-	mov	(%rsi),%r15
-	mov	8(%rsi),%r14
-	mov	16(%rsi),%r13
-	mov	24(%rsi),%r12
-	mov	32(%rsi),%rbp
-	mov	40(%rsi),%rbx
-	lea	48(%rsi),%rsp
+.cfi_def_cfa	%rsi,8
+	mov	-48(%rsi),%r15
+.cfi_restore	%r15
+	mov	-40(%rsi),%r14
+.cfi_restore	%r14
+	mov	-32(%rsi),%r13
+.cfi_restore	%r13
+	mov	-24(%rsi),%r12
+.cfi_restore	%r12
+	mov	-16(%rsi),%rbp
+.cfi_restore	%rbp
+	mov	-8(%rsi),%rbx
+.cfi_restore	%rbx
+	lea	(%rsi),%rsp
+.cfi_def_cfa_register	%rsp
 .Lepilogue:
 	ret
+.cfi_endproc
 .size	$func,.-$func
 ___
 
@@ -750,14 +780,22 @@ $code.=<<___;
 .type	${func}_ssse3,\@function,3
 .align	64
 ${func}_ssse3:
+.cfi_startproc
 .Lssse3_shortcut:
+	mov	%rsp,%rax		# copy %rsp
+.cfi_def_cfa_register	%rax
 	push	%rbx
+.cfi_push	%rbx
 	push	%rbp
+.cfi_push	%rbp
 	push	%r12
+.cfi_push	%r12
 	push	%r13
+.cfi_push	%r13
 	push	%r14
+.cfi_push	%r14
 	push	%r15
-	mov	%rsp,%r11		# copy %rsp
+.cfi_push	%r15
 	shl	\$4,%rdx		# num*16
 	sub	\$`$framesz+$win64*16*4`,%rsp
 	lea	($inp,%rdx,$SZ),%rdx	# inp+num*16*$SZ
@@ -765,7 +803,8 @@ ${func}_ssse3:
 	mov	$ctx,$_ctx		# save ctx, 1st arg
 	mov	$inp,$_inp		# save inp, 2nd arh
 	mov	%rdx,$_end		# save end pointer, "3rd" arg
-	mov	%r11,$_rsp		# save copy of %rsp
+	mov	%rax,$_rsp		# save copy of %rsp
+.cfi_cfa_expression	$_rsp,deref,+8
 ___
 $code.=<<___ if ($win64);
 	movaps	%xmm6,16*$SZ+32(%rsp)
@@ -1064,6 +1103,7 @@ $code.=<<___;
 	jb	.Lloop_ssse3
 
 	mov	$_rsp,%rsi
+.cfi_def_cfa	%rsi,8
 ___
 $code.=<<___ if ($win64);
 	movaps	16*$SZ+32(%rsp),%xmm6
@@ -1072,15 +1112,23 @@ $code.=<<___ if ($win64);
 	movaps	16*$SZ+80(%rsp),%xmm9
 ___
 $code.=<<___;
-	mov	(%rsi),%r15
-	mov	8(%rsi),%r14
-	mov	16(%rsi),%r13
-	mov	24(%rsi),%r12
-	mov	32(%rsi),%rbp
-	mov	40(%rsi),%rbx
-	lea	48(%rsi),%rsp
+	mov	-48(%rsi),%r15
+.cfi_restore	%r15
+	mov	-40(%rsi),%r14
+.cfi_restore	%r14
+	mov	-32(%rsi),%r13
+.cfi_restore	%r13
+	mov	-24(%rsi),%r12
+.cfi_restore	%r12
+	mov	-16(%rsi),%rbp
+.cfi_restore	%rbp
+	mov	-8(%rsi),%rbx
+.cfi_restore	%rbx
+	lea	(%rsi),%rsp
+.cfi_def_cfa_register	%rsp
 .Lepilogue_ssse3:
 	ret
+.cfi_endproc
 .size	${func}_ssse3,.-${func}_ssse3
 ___
 }
@@ -1094,14 +1142,22 @@ $code.=<<___;
 .type	${func}_xop,\@function,3
 .align	64
 ${func}_xop:
+.cfi_startproc
 .Lxop_shortcut:
+	mov	%rsp,%rax		# copy %rsp
+.cfi_def_cfa_register	%rax
 	push	%rbx
+.cfi_push	%rbx
 	push	%rbp
+.cfi_push	%rbp
 	push	%r12
+.cfi_push	%r12
 	push	%r13
+.cfi_push	%r13
 	push	%r14
+.cfi_push	%r14
 	push	%r15
-	mov	%rsp,%r11		# copy %rsp
+.cfi_push	%r15
 	shl	\$4,%rdx		# num*16
 	sub	\$`$framesz+$win64*16*($SZ==4?4:6)`,%rsp
 	lea	($inp,%rdx,$SZ),%rdx	# inp+num*16*$SZ
@@ -1109,7 +1165,8 @@ ${func}_xop:
 	mov	$ctx,$_ctx		# save ctx, 1st arg
 	mov	$inp,$_inp		# save inp, 2nd arh
 	mov	%rdx,$_end		# save end pointer, "3rd" arg
-	mov	%r11,$_rsp		# save copy of %rsp
+	mov	%rax,$_rsp		# save copy of %rsp
+.cfi_cfa_expression	$_rsp,deref,+8
 ___
 $code.=<<___ if ($win64);
 	movaps	%xmm6,16*$SZ+32(%rsp)
@@ -1436,6 +1493,7 @@ $code.=<<___;
 	jb	.Lloop_xop
 
 	mov	$_rsp,%rsi
+.cfi_def_cfa	%rsi,8
 	vzeroupper
 ___
 $code.=<<___ if ($win64);
@@ -1449,15 +1507,23 @@ $code.=<<___ if ($win64 && $SZ>4);
 	movaps	16*$SZ+112(%rsp),%xmm11
 ___
 $code.=<<___;
-	mov	(%rsi),%r15
-	mov	8(%rsi),%r14
-	mov	16(%rsi),%r13
-	mov	24(%rsi),%r12
-	mov	32(%rsi),%rbp
-	mov	40(%rsi),%rbx
-	lea	48(%rsi),%rsp
+	mov	-48(%rsi),%r15
+.cfi_restore	%r15
+	mov	-40(%rsi),%r14
+.cfi_restore	%r14
+	mov	-32(%rsi),%r13
+.cfi_restore	%r13
+	mov	-24(%rsi),%r12
+.cfi_restore	%r12
+	mov	-16(%rsi),%rbp
+.cfi_restore	%rbp
+	mov	-8(%rsi),%rbx
+.cfi_restore	%rbx
+	lea	(%rsi),%rsp
+.cfi_def_cfa_register	%rsp
 .Lepilogue_xop:
 	ret
+.cfi_endproc
 .size	${func}_xop,.-${func}_xop
 ___
 }
@@ -1470,14 +1536,22 @@ $code.=<<___;
 .type	${func}_avx,\@function,3
 .align	64
 ${func}_avx:
+.cfi_startproc
 .Lavx_shortcut:
+	mov	%rsp,%rax		# copy %rsp
+.cfi_def_cfa_register	%rax
 	push	%rbx
+.cfi_push	%rbx
 	push	%rbp
+.cfi_push	%rbp
 	push	%r12
+.cfi_push	%r12
 	push	%r13
+.cfi_push	%r13
 	push	%r14
+.cfi_push	%r14
 	push	%r15
-	mov	%rsp,%r11		# copy %rsp
+.cfi_push	%r15
 	shl	\$4,%rdx		# num*16
 	sub	\$`$framesz+$win64*16*($SZ==4?4:6)`,%rsp
 	lea	($inp,%rdx,$SZ),%rdx	# inp+num*16*$SZ
@@ -1485,7 +1559,8 @@ ${func}_avx:
 	mov	$ctx,$_ctx		# save ctx, 1st arg
 	mov	$inp,$_inp		# save inp, 2nd arh
 	mov	%rdx,$_end		# save end pointer, "3rd" arg
-	mov	%r11,$_rsp		# save copy of %rsp
+	mov	%rax,$_rsp		# save copy of %rsp
+.cfi_cfa_expression	$_rsp,deref,+8
 ___
 $code.=<<___ if ($win64);
 	movaps	%xmm6,16*$SZ+32(%rsp)
@@ -1744,6 +1819,7 @@ $code.=<<___;
 	jb	.Lloop_avx
 
 	mov	$_rsp,%rsi
+.cfi_def_cfa	%rsi,8
 	vzeroupper
 ___
 $code.=<<___ if ($win64);
@@ -1757,15 +1833,23 @@ $code.=<<___ if ($win64 && $SZ>4);
 	movaps	16*$SZ+112(%rsp),%xmm11
 ___
 $code.=<<___;
-	mov	(%rsi),%r15
-	mov	8(%rsi),%r14
-	mov	16(%rsi),%r13
-	mov	24(%rsi),%r12
-	mov	32(%rsi),%rbp
-	mov	40(%rsi),%rbx
-	lea	48(%rsi),%rsp
+	mov	-48(%rsi),%r15
+.cfi_restore	%r15
+	mov	-40(%rsi),%r14
+.cfi_restore	%r14
+	mov	-32(%rsi),%r13
+.cfi_restore	%r13
+	mov	-24(%rsi),%r12
+.cfi_restore	%r12
+	mov	-16(%rsi),%rbp
+.cfi_restore	%rbp
+	mov	-8(%rsi),%rbx
+.cfi_restore	%rbx
+	lea	(%rsi),%rsp
+.cfi_def_cfa_register	%rsp
 .Lepilogue_avx:
 	ret
+.cfi_endproc
 .size	${func}_avx,.-${func}_avx
 ___
 
@@ -1773,7 +1857,7 @@ if ($avx>1) {{
 ######################################################################
 # AVX2+BMI code path
 #
-my $a5=$SZ==4?"%esi":"%rsi";	# zap $inp 
+my $a5=$SZ==4?"%esi":"%rsi";	# zap $inp
 my $PUSH8=8*2*$SZ;
 use integer;
 
@@ -1821,14 +1905,22 @@ $code.=<<___;
 .type	${func}_avx2,\@function,3
 .align	64
 ${func}_avx2:
+.cfi_startproc
 .Lavx2_shortcut:
+	mov	%rsp,%rax		# copy %rsp
+.cfi_def_cfa_register	%rax
 	push	%rbx
+.cfi_push	%rbx
 	push	%rbp
+.cfi_push	%rbp
 	push	%r12
+.cfi_push	%r12
 	push	%r13
+.cfi_push	%r13
 	push	%r14
+.cfi_push	%r14
 	push	%r15
-	mov	%rsp,%r11		# copy %rsp
+.cfi_push	%r15
 	sub	\$`2*$SZ*$rounds+4*8+$win64*16*($SZ==4?4:6)`,%rsp
 	shl	\$4,%rdx		# num*16
 	and	\$-256*$SZ,%rsp		# align stack frame
@@ -1837,7 +1929,8 @@ ${func}_avx2:
 	mov	$ctx,$_ctx		# save ctx, 1st arg
 	mov	$inp,$_inp		# save inp, 2nd arh
 	mov	%rdx,$_end		# save end pointer, "3rd" arg
-	mov	%r11,$_rsp		# save copy of %rsp
+	mov	%rax,$_rsp		# save copy of %rsp
+.cfi_cfa_expression	$_rsp,deref,+8
 ___
 $code.=<<___ if ($win64);
 	movaps	%xmm6,16*$SZ+32(%rsp)
@@ -2118,6 +2211,7 @@ $code.=<<___;
 .Ldone_avx2:
 	lea	($Tbl),%rsp
 	mov	$_rsp,%rsi
+.cfi_def_cfa	%rsi,8
 	vzeroupper
 ___
 $code.=<<___ if ($win64);
@@ -2131,15 +2225,23 @@ $code.=<<___ if ($win64 && $SZ>4);
 	movaps	16*$SZ+112(%rsp),%xmm11
 ___
 $code.=<<___;
-	mov	(%rsi),%r15
-	mov	8(%rsi),%r14
-	mov	16(%rsi),%r13
-	mov	24(%rsi),%r12
-	mov	32(%rsi),%rbp
-	mov	40(%rsi),%rbx
-	lea	48(%rsi),%rsp
+	mov	-48(%rsi),%r15
+.cfi_restore	%r15
+	mov	-40(%rsi),%r14
+.cfi_restore	%r14
+	mov	-32(%rsi),%r13
+.cfi_restore	%r13
+	mov	-24(%rsi),%r12
+.cfi_restore	%r12
+	mov	-16(%rsi),%rbp
+.cfi_restore	%rbp
+	mov	-8(%rsi),%rbx
+.cfi_restore	%rbx
+	lea	(%rsi),%rsp
+.cfi_def_cfa_register	%rsp
 .Lepilogue_avx2:
 	ret
+.cfi_endproc
 .size	${func}_avx2,.-${func}_avx2
 ___
 }}
@@ -2199,7 +2301,6 @@ ___
 $code.=<<___;
 	mov	%rax,%rsi		# put aside Rsp
 	mov	16*$SZ+3*8(%rax),%rax	# pull $_rsp
-	lea	48(%rax),%rax
 
 	mov	-8(%rax),%rbx
 	mov	-16(%rax),%rbp
@@ -2262,7 +2363,9 @@ $code.=<<___;
 	pop	%rsi
 	ret
 .size	se_handler,.-se_handler
+___
 
+$code.=<<___ if ($SZ==4 && $shaext);
 .type	shaext_handler,\@abi-omnipotent
 .align	16
 shaext_handler:
@@ -2295,14 +2398,16 @@ shaext_handler:
 
 	jmp	.Lin_prologue
 .size	shaext_handler,.-shaext_handler
+___
 
+$code.=<<___;
 .section	.pdata
 .align	4
 	.rva	.LSEH_begin_$func
 	.rva	.LSEH_end_$func
 	.rva	.LSEH_info_$func
 ___
-$code.=<<___ if ($SZ==4 && $shext);
+$code.=<<___ if ($SZ==4 && $shaext);
 	.rva	.LSEH_begin_${func}_shaext
 	.rva	.LSEH_end_${func}_shaext
 	.rva	.LSEH_info_${func}_shaext
@@ -2335,10 +2440,12 @@ $code.=<<___;
 	.rva	se_handler
 	.rva	.Lprologue,.Lepilogue			# HandlerData[]
 ___
-$code.=<<___ if ($SZ==4);
+$code.=<<___ if ($SZ==4 && $shaext);
 .LSEH_info_${func}_shaext:
 	.byte	9,0,0,0
 	.rva	shaext_handler
+___
+$code.=<<___ if ($SZ==4);
 .LSEH_info_${func}_ssse3:
 	.byte	9,0,0,0
 	.rva	se_handler