[thirdparty/openssl.git] / crypto / x86_64cpuid.pl

#! /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


$flavour = shift;
$output  = shift;
if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }

$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);

$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or
( $xlate="${dir}perlasm/x86_64-xlate.pl" and -f $xlate) or
die "can't locate x86_64-xlate.pl";

open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\"";
*STDOUT=*OUT;

($arg1,$arg2,$arg3,$arg4)=$win64?("%rcx","%rdx","%r8", "%r9") :	# Win64 order
				 ("%rdi","%rsi","%rdx","%rcx");	# Unix order

print<<___;
.extern		OPENSSL_cpuid_setup
.hidden		OPENSSL_cpuid_setup
.section	.init
	call	OPENSSL_cpuid_setup

.hidden	OPENSSL_ia32cap_P
.comm	OPENSSL_ia32cap_P,16,4

.text

.globl	OPENSSL_atomic_add
.type	OPENSSL_atomic_add,\@abi-omnipotent
.align	16
OPENSSL_atomic_add:
	movl	($arg1),%eax
.Lspin:	leaq	($arg2,%rax),%r8
	.byte	0xf0		# lock
	cmpxchgl	%r8d,($arg1)
	jne	.Lspin
	movl	%r8d,%eax
	.byte	0x48,0x98	# cltq/cdqe
	ret
.size	OPENSSL_atomic_add,.-OPENSSL_atomic_add

.globl	OPENSSL_rdtsc
.type	OPENSSL_rdtsc,\@abi-omnipotent
.align	16
OPENSSL_rdtsc:
	rdtsc
	shl	\$32,%rdx
	or	%rdx,%rax
	ret
.size	OPENSSL_rdtsc,.-OPENSSL_rdtsc

.globl	OPENSSL_ia32_cpuid
.type	OPENSSL_ia32_cpuid,\@function,1
.align	16
OPENSSL_ia32_cpuid:
	mov	%rbx,%r8		# save %rbx

	xor	%eax,%eax
	mov	%eax,8(%rdi)		# clear extended feature flags
	cpuid
	mov	%eax,%r11d		# max value for standard query level

	xor	%eax,%eax
	cmp	\$0x756e6547,%ebx	# "Genu"
	setne	%al
	mov	%eax,%r9d
	cmp	\$0x49656e69,%edx	# "ineI"
	setne	%al
	or	%eax,%r9d
	cmp	\$0x6c65746e,%ecx	# "ntel"
	setne	%al
	or	%eax,%r9d		# 0 indicates Intel CPU
	jz	.Lintel

	cmp	\$0x68747541,%ebx	# "Auth"
	setne	%al
	mov	%eax,%r10d
	cmp	\$0x69746E65,%edx	# "enti"
	setne	%al
	or	%eax,%r10d
	cmp	\$0x444D4163,%ecx	# "cAMD"
	setne	%al
	or	%eax,%r10d		# 0 indicates AMD CPU
	jnz	.Lintel

	# AMD specific
	mov	\$0x80000000,%eax
	cpuid
	cmp	\$0x80000001,%eax
	jb	.Lintel
	mov	%eax,%r10d
	mov	\$0x80000001,%eax
	cpuid
	or	%ecx,%r9d
	and	\$0x00000801,%r9d	# isolate AMD XOP bit, 1<<11

	cmp	\$0x80000008,%r10d
	jb	.Lintel

	mov	\$0x80000008,%eax
	cpuid
	movzb	%cl,%r10		# number of cores - 1
	inc	%r10			# number of cores

	mov	\$1,%eax
	cpuid
	bt	\$28,%edx		# test hyper-threading bit
	jnc	.Lgeneric
	shr	\$16,%ebx		# number of logical processors
	cmp	%r10b,%bl
	ja	.Lgeneric
	and	\$0xefffffff,%edx	# ~(1<<28)
	jmp	.Lgeneric

.Lintel:
	cmp	\$4,%r11d
	mov	\$-1,%r10d
	jb	.Lnocacheinfo

	mov	\$4,%eax
	mov	\$0,%ecx		# query L1D
	cpuid
	mov	%eax,%r10d
	shr	\$14,%r10d
	and	\$0xfff,%r10d		# number of cores -1 per L1D

.Lnocacheinfo:
	mov	\$1,%eax
	cpuid
	and	\$0xbfefffff,%edx	# force reserved bits to 0
	cmp	\$0,%r9d
	jne	.Lnotintel
	or	\$0x40000000,%edx	# set reserved bit#30 on Intel CPUs
	and	\$15,%ah
	cmp	\$15,%ah		# examine Family ID
	jne	.LnotP4
	or	\$0x00100000,%edx	# set reserved bit#20 to engage RC4_CHAR
.LnotP4:
	cmp	\$6,%ah
	jne	.Lnotintel
	and	\$0x0fff0ff0,%eax
	cmp	\$0x00050670,%eax	# Knights Landing
	je	.Lknights
	cmp	\$0x00080650,%eax	# Knights Mill (according to sde)
	jne	.Lnotintel
.Lknights:
	and	\$0xfbffffff,%ecx	# clear XSAVE flag to mimic Silvermont

.Lnotintel:
	bt	\$28,%edx		# test hyper-threading bit
	jnc	.Lgeneric
	and	\$0xefffffff,%edx	# ~(1<<28)
	cmp	\$0,%r10d
	je	.Lgeneric

	or	\$0x10000000,%edx	# 1<<28
	shr	\$16,%ebx
	cmp	\$1,%bl			# see if cache is shared
	ja	.Lgeneric
	and	\$0xefffffff,%edx	# ~(1<<28)
.Lgeneric:
	and	\$0x00000800,%r9d	# isolate AMD XOP flag
	and	\$0xfffff7ff,%ecx
	or	%ecx,%r9d		# merge AMD XOP flag

	mov	%edx,%r10d		# %r9d:%r10d is copy of %ecx:%edx

	cmp	\$7,%r11d
	jb	.Lno_extended_info
	mov	\$7,%eax
	xor	%ecx,%ecx
	cpuid
	bt	\$26,%r9d		# check XSAVE bit, cleared on Knights
	jc	.Lnotknights
	and	\$0xfff7ffff,%ebx	# clear ADCX/ADOX flag
.Lnotknights:
	mov	%ebx,8(%rdi)		# save extended feature flags
.Lno_extended_info:

	bt	\$27,%r9d		# check OSXSAVE bit
	jnc	.Lclear_avx
	xor	%ecx,%ecx		# XCR0
	.byte	0x0f,0x01,0xd0		# xgetbv
	and	\$6,%eax		# isolate XMM and YMM state support
	cmp	\$6,%eax
	je	.Ldone
.Lclear_avx:
	mov	\$0xefffe7ff,%eax	# ~(1<<28|1<<12|1<<11)
	and	%eax,%r9d		# clear AVX, FMA and AMD XOP bits
	andl	\$0xffffffdf,8(%rdi)	# clear AVX2, ~(1<<5)
.Ldone:
	shl	\$32,%r9
	mov	%r10d,%eax
	mov	%r8,%rbx		# restore %rbx
	or	%r9,%rax
	ret
.size	OPENSSL_ia32_cpuid,.-OPENSSL_ia32_cpuid

.globl  OPENSSL_cleanse
.type   OPENSSL_cleanse,\@abi-omnipotent
.align  16
OPENSSL_cleanse:
	xor	%rax,%rax
	cmp	\$15,$arg2
	jae	.Lot
	cmp	\$0,$arg2
	je	.Lret
.Little:
	mov	%al,($arg1)
	sub	\$1,$arg2
	lea	1($arg1),$arg1
	jnz	.Little
.Lret:
	ret
.align	16
.Lot:
	test	\$7,$arg1
	jz	.Laligned
	mov	%al,($arg1)
	lea	-1($arg2),$arg2
	lea	1($arg1),$arg1
	jmp	.Lot
.Laligned:
	mov	%rax,($arg1)
	lea	-8($arg2),$arg2
	test	\$-8,$arg2
	lea	8($arg1),$arg1
	jnz	.Laligned
	cmp	\$0,$arg2
	jne	.Little
	ret
.size	OPENSSL_cleanse,.-OPENSSL_cleanse

.globl  CRYPTO_memcmp
.type   CRYPTO_memcmp,\@abi-omnipotent
.align  16
CRYPTO_memcmp:
	xor	%rax,%rax
	xor	%r10,%r10
	cmp	\$0,$arg3
	je	.Lno_data
.Loop_cmp:
	mov	($arg1),%r10b
	lea	1($arg1),$arg1
	xor	($arg2),%r10b
	lea	1($arg2),$arg2
	or	%r10b,%al
	dec	$arg3
	jnz	.Loop_cmp
	neg	%rax
	shr	\$63,%rax
.Lno_data:
	ret
.size	CRYPTO_memcmp,.-CRYPTO_memcmp
___

print<<___ if (!$win64);
.globl	OPENSSL_wipe_cpu
.type	OPENSSL_wipe_cpu,\@abi-omnipotent
.align	16
OPENSSL_wipe_cpu:
	pxor	%xmm0,%xmm0
	pxor	%xmm1,%xmm1
	pxor	%xmm2,%xmm2
	pxor	%xmm3,%xmm3
	pxor	%xmm4,%xmm4
	pxor	%xmm5,%xmm5
	pxor	%xmm6,%xmm6
	pxor	%xmm7,%xmm7
	pxor	%xmm8,%xmm8
	pxor	%xmm9,%xmm9
	pxor	%xmm10,%xmm10
	pxor	%xmm11,%xmm11
	pxor	%xmm12,%xmm12
	pxor	%xmm13,%xmm13
	pxor	%xmm14,%xmm14
	pxor	%xmm15,%xmm15
	xorq	%rcx,%rcx
	xorq	%rdx,%rdx
	xorq	%rsi,%rsi
	xorq	%rdi,%rdi
	xorq	%r8,%r8
	xorq	%r9,%r9
	xorq	%r10,%r10
	xorq	%r11,%r11
	leaq	8(%rsp),%rax
	ret
.size	OPENSSL_wipe_cpu,.-OPENSSL_wipe_cpu
___
print<<___ if ($win64);
.globl	OPENSSL_wipe_cpu
.type	OPENSSL_wipe_cpu,\@abi-omnipotent
.align	16
OPENSSL_wipe_cpu:
	pxor	%xmm0,%xmm0
	pxor	%xmm1,%xmm1
	pxor	%xmm2,%xmm2
	pxor	%xmm3,%xmm3
	pxor	%xmm4,%xmm4
	pxor	%xmm5,%xmm5
	xorq	%rcx,%rcx
	xorq	%rdx,%rdx
	xorq	%r8,%r8
	xorq	%r9,%r9
	xorq	%r10,%r10
	xorq	%r11,%r11
	leaq	8(%rsp),%rax
	ret
.size	OPENSSL_wipe_cpu,.-OPENSSL_wipe_cpu
___
{
my $out="%r10";
my $cnt="%rcx";
my $max="%r11";
my $lasttick="%r8d";
my $lastdiff="%r9d";
my $redzone=win64?8:-8;

print<<___;
.globl	OPENSSL_instrument_bus
.type	OPENSSL_instrument_bus,\@abi-omnipotent
.align	16
OPENSSL_instrument_bus:
	mov	$arg1,$out	# tribute to Win64
	mov	$arg2,$cnt
	mov	$arg2,$max

	rdtsc			# collect 1st tick
	mov	%eax,$lasttick	# lasttick = tick
	mov	\$0,$lastdiff	# lastdiff = 0
	clflush	($out)
	.byte	0xf0		# lock
	add	$lastdiff,($out)
	jmp	.Loop
.align	16
.Loop:	rdtsc
	mov	%eax,%edx
	sub	$lasttick,%eax
	mov	%edx,$lasttick
	mov	%eax,$lastdiff
	clflush	($out)
	.byte	0xf0		# lock
	add	%eax,($out)
	lea	4($out),$out
	sub	\$1,$cnt
	jnz	.Loop

	mov	$max,%rax
	ret
.size	OPENSSL_instrument_bus,.-OPENSSL_instrument_bus

.globl	OPENSSL_instrument_bus2
.type	OPENSSL_instrument_bus2,\@abi-omnipotent
.align	16
OPENSSL_instrument_bus2:
	mov	$arg1,$out	# tribute to Win64
	mov	$arg2,$cnt
	mov	$arg3,$max
	mov	$cnt,$redzone(%rsp)

	rdtsc			# collect 1st tick
	mov	%eax,$lasttick	# lasttick = tick
	mov	\$0,$lastdiff	# lastdiff = 0

	clflush	($out)
	.byte	0xf0		# lock
	add	$lastdiff,($out)

	rdtsc			# collect 1st diff
	mov	%eax,%edx
	sub	$lasttick,%eax	# diff
	mov	%edx,$lasttick	# lasttick = tick
	mov	%eax,$lastdiff	# lastdiff = diff
.Loop2:
	clflush	($out)
	.byte	0xf0		# lock
	add	%eax,($out)	# accumulate diff

	sub	\$1,$max
	jz	.Ldone2

	rdtsc
	mov	%eax,%edx
	sub	$lasttick,%eax	# diff
	mov	%edx,$lasttick	# lasttick = tick
	cmp	$lastdiff,%eax
	mov	%eax,$lastdiff	# lastdiff = diff
	mov	\$0,%edx
	setne	%dl
	sub	%rdx,$cnt	# conditional --$cnt
	lea	($out,%rdx,4),$out	# conditional ++$out
	jnz	.Loop2

.Ldone2:
	mov	$redzone(%rsp),%rax
	sub	$cnt,%rax
	ret
.size	OPENSSL_instrument_bus2,.-OPENSSL_instrument_bus2
___
}

sub gen_random {
my $rdop = shift;
print<<___;
.globl	OPENSSL_ia32_${rdop}
.type	OPENSSL_ia32_${rdop},\@abi-omnipotent
.align	16
OPENSSL_ia32_${rdop}:
	mov	\$8,%ecx
.Loop_${rdop}:
	${rdop}	%rax
	jc	.Lbreak_${rdop}
	loop	.Loop_${rdop}
.Lbreak_${rdop}:
	cmp	\$0,%rax
	cmove	%rcx,%rax
	ret
.size	OPENSSL_ia32_${rdop},.-OPENSSL_ia32_${rdop}

.globl	OPENSSL_ia32_${rdop}_bytes
.type	OPENSSL_ia32_${rdop}_bytes,\@abi-omnipotent
.align	16
OPENSSL_ia32_${rdop}_bytes:
	xor	%rax, %rax	# return value
	cmp	\$0,$arg2
	je	.Ldone_${rdop}_bytes

	mov	\$8,%r11
.Loop_${rdop}_bytes:
	${rdop}	%r10
	jc	.Lbreak_${rdop}_bytes
	dec	%r11
	jnz	.Loop_${rdop}_bytes
	jmp	.Ldone_${rdop}_bytes

.align	16
.Lbreak_${rdop}_bytes:
	cmp	\$8,$arg2
	jb	.Ltail_${rdop}_bytes
	mov	%r10,($arg1)
	lea	8($arg1),$arg1
	add	\$8,%rax
	sub	\$8,$arg2
	jz	.Ldone_${rdop}_bytes
	mov	\$8,%r11
	jmp	.Loop_${rdop}_bytes

.align	16
.Ltail_${rdop}_bytes:
	mov	%r10b,($arg1)
	lea	1($arg1),$arg1
	inc	%rax
	shr	\$8,%r8
	dec	$arg2
	jnz	.Ltail_${rdop}_bytes

.Ldone_${rdop}_bytes:
	ret
.size	OPENSSL_ia32_${rdop}_bytes,.-OPENSSL_ia32_${rdop}_bytes
___
}
gen_random("rdrand");
gen_random("rdseed");

close STDOUT;	# flush
Commit	Line	Data
e0a65194 RS	1	#! /usr/bin/env perl
	2	# Copyright 2005-2016 The OpenSSL Project Authors. All Rights Reserved.
	3	#
	4	# Licensed under the OpenSSL license (the "License"). You may not use
	5	# this file except in compliance with the License. You can obtain a copy
	6	# in the file LICENSE in the source distribution or at
	7	# https://www.openssl.org/source/license.html
	8
14e21f86	9
aa8f38e4 AP	10	$flavour = shift;
	11	$output = shift;
	12	if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
14e21f86	13
aa8f38e4	14	$win64=0; $win64=1 if ($flavour =~ /[nm]asm\|mingw64/ \|\| $output =~ /\.asm$/);
376729e1	15
aa8f38e4	16	$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
301799b8 AP	17	( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or
	18	( $xlate="${dir}perlasm/x86_64-xlate.pl" and -f $xlate) or
	19	die "can't locate x86_64-xlate.pl";
	20
cfe1d992	21	open OUT,"\| \"$^X\" \"$xlate\" $flavour \"$output\"";
46bf83f0	22	STDOUT=OUT;
376729e1	23
5fabb88a AP	24	($arg1,$arg2,$arg3,$arg4)=$win64?("%rcx","%rdx","%r8", "%r9") : # Win64 order
	25	("%rdi","%rsi","%rdx","%rcx"); # Unix order
	26
aa8f38e4 AP	27	print<<___;
aa8f38e4 AP	28	.extern OPENSSL_cpuid_setup
ddc20d4d	29	.hidden OPENSSL_cpuid_setup
aa8f38e4 AP	30	.section .init
aa8f38e4 AP	31	call OPENSSL_cpuid_setup
932cc129	32
ddc20d4d	33	.hidden OPENSSL_ia32cap_P
c5cd28bd	34	.comm OPENSSL_ia32cap_P,16,4
ddc20d4d	35
14e21f86	36	.text
376729e1 AP	37
376729e1 AP	38	.globl OPENSSL_atomic_add
aa8f38e4	39	.type OPENSSL_atomic_add,\@abi-omnipotent
376729e1 AP	40	.align 16
376729e1 AP	41	OPENSSL_atomic_add:
aa8f38e4 AP	42	movl ($arg1),%eax
	43	.Lspin: leaq ($arg2,%rax),%r8
	44	.byte 0xf0 # lock
	45	cmpxchgl %r8d,($arg1)
376729e1	46	jne .Lspin
e442c362	47	movl %r8d,%eax
aa8f38e4	48	.byte 0x48,0x98 # cltq/cdqe
376729e1 AP	49	ret
	50	.size OPENSSL_atomic_add,.-OPENSSL_atomic_add
	51
932cc129 AP	52	.globl OPENSSL_rdtsc
	53	.type OPENSSL_rdtsc,\@abi-omnipotent
	54	.align 16
	55	OPENSSL_rdtsc:
	56	rdtsc
	57	shl \$32,%rdx
	58	or %rdx,%rax
	59	ret
	60	.size OPENSSL_rdtsc,.-OPENSSL_rdtsc
	61
376729e1	62	.globl OPENSSL_ia32_cpuid
c5cd28bd	63	.type OPENSSL_ia32_cpuid,\@function,1
376729e1 AP	64	.align 16
376729e1 AP	65	OPENSSL_ia32_cpuid:
b9064221	66	mov %rbx,%r8 # save %rbx
9babf392 AP	67
9babf392 AP	68	xor %eax,%eax
8e4d861c	69	mov %eax,8(%rdi) # clear extended feature flags
9babf392	70	cpuid
761393bb AP	71	mov %eax,%r11d # max value for standard query level
761393bb AP	72
9babf392 AP	73	xor %eax,%eax
	74	cmp \$0x756e6547,%ebx # "Genu"
	75	setne %al
	76	mov %eax,%r9d
	77	cmp \$0x49656e69,%edx # "ineI"
	78	setne %al
	79	or %eax,%r9d
	80	cmp \$0x6c65746e,%ecx # "ntel"
	81	setne %al
5cd91b50	82	or %eax,%r9d # 0 indicates Intel CPU
5cd91b50 AP	83	jz .Lintel
	84
	85	cmp \$0x68747541,%ebx # "Auth"
	86	setne %al
	87	mov %eax,%r10d
	88	cmp \$0x69746E65,%edx # "enti"
	89	setne %al
	90	or %eax,%r10d
	91	cmp \$0x444D4163,%ecx # "cAMD"
	92	setne %al
	93	or %eax,%r10d # 0 indicates AMD CPU
	94	jnz .Lintel
	95
761393bb	96	# AMD specific
5cd91b50 AP	97	mov \$0x80000000,%eax
5cd91b50 AP	98	cpuid
b9064221 AP	99	cmp \$0x80000001,%eax
	100	jb .Lintel
	101	mov %eax,%r10d
	102	mov \$0x80000001,%eax
	103	cpuid
	104	or %ecx,%r9d
	105	and \$0x00000801,%r9d # isolate AMD XOP bit, 1<<11
	106
	107	cmp \$0x80000008,%r10d
5cd91b50 AP	108	jb .Lintel
	109
	110	mov \$0x80000008,%eax
	111	cpuid
	112	movzb %cl,%r10 # number of cores - 1
	113	inc %r10 # number of cores
9babf392	114
761393bb AP	115	mov \$1,%eax
	116	cpuid
	117	bt \$28,%edx # test hyper-threading bit
b9064221	118	jnc .Lgeneric
761393bb AP	119	shr \$16,%ebx # number of logical processors
761393bb AP	120	cmp %r10b,%bl
b9064221	121	ja .Lgeneric
761393bb	122	and \$0xefffffff,%edx # ~(1<<28)
b9064221	123	jmp .Lgeneric
761393bb	124
5cd91b50	125	.Lintel:
761393bb AP	126	cmp \$4,%r11d
	127	mov \$-1,%r10d
	128	jb .Lnocacheinfo
	129
	130	mov \$4,%eax
	131	mov \$0,%ecx # query L1D
	132	cpuid
	133	mov %eax,%r10d
	134	shr \$14,%r10d
	135	and \$0xfff,%r10d # number of cores -1 per L1D
	136
	137	.Lnocacheinfo:
932cc129	138	mov \$1,%eax
376729e1	139	cpuid
4bb90087	140	and \$0xbfefffff,%edx # force reserved bits to 0
932cc129	141	cmp \$0,%r9d
9babf392	142	jne .Lnotintel
4bb90087	143	or \$0x40000000,%edx # set reserved bit#30 on Intel CPUs
9babf392 AP	144	and \$15,%ah
9babf392 AP	145	cmp \$15,%ah # examine Family ID
738a9dd5	146	jne .LnotP4
4bb90087	147	or \$0x00100000,%edx # set reserved bit#20 to engage RC4_CHAR
738a9dd5 AP	148	.LnotP4:
	149	cmp \$6,%ah
	150	jne .Lnotintel
	151	and \$0x0fff0ff0,%eax
	152	cmp \$0x00050670,%eax # Knights Landing
	153	je .Lknights
	154	cmp \$0x00080650,%eax # Knights Mill (according to sde)
	155	jne .Lnotintel
	156	.Lknights:
	157	and \$0xfbffffff,%ecx # clear XSAVE flag to mimic Silvermont
	158
9babf392	159	.Lnotintel:
3df2eff4	160	bt \$28,%edx # test hyper-threading bit
b9064221	161	jnc .Lgeneric
761393bb AP	162	and \$0xefffffff,%edx # ~(1<<28)
761393bb AP	163	cmp \$0,%r10d
b9064221	164	je .Lgeneric
761393bb AP	165
761393bb AP	166	or \$0x10000000,%edx # 1<<28
9babf392	167	shr \$16,%ebx
761393bb	168	cmp \$1,%bl # see if cache is shared
b9064221	169	ja .Lgeneric
932cc129	170	and \$0xefffffff,%edx # ~(1<<28)
b9064221 AP	171	.Lgeneric:
	172	and \$0x00000800,%r9d # isolate AMD XOP flag
	173	and \$0xfffff7ff,%ecx
2bc3ad28	174	or %ecx,%r9d # merge AMD XOP flag
b9064221	175
2bc3ad28	176	mov %edx,%r10d # %r9d:%r10d is copy of %ecx:%edx
8e4d861c AP	177
	178	cmp \$7,%r11d
	179	jb .Lno_extended_info
	180	mov \$7,%eax
	181	xor %ecx,%ecx
	182	cpuid
738a9dd5 AP	183	bt \$26,%r9d # check XSAVE bit, cleared on Knights
	184	jc .Lnotknights
	185	and \$0xfff7ffff,%ebx # clear ADCX/ADOX flag
	186	.Lnotknights:
8e4d861c AP	187	mov %ebx,8(%rdi) # save extended feature flags
	188	.Lno_extended_info:
	189
2bc3ad28	190	bt \$27,%r9d # check OSXSAVE bit
b9064221 AP	191	jnc .Lclear_avx
	192	xor %ecx,%ecx # XCR0
	193	.byte 0x0f,0x01,0xd0 # xgetbv
	194	and \$6,%eax # isolate XMM and YMM state support
	195	cmp \$6,%eax
	196	je .Ldone
	197	.Lclear_avx:
	198	mov \$0xefffe7ff,%eax # ~(1<<28\|1<<12\|1<<11)
2bc3ad28	199	and %eax,%r9d # clear AVX, FMA and AMD XOP bits
574496c0	200	andl \$0xffffffdf,8(%rdi) # clear AVX2, ~(1<<5)
b9064221	201	.Ldone:
2bc3ad28 AP	202	shl \$32,%r9
2bc3ad28 AP	203	mov %r10d,%eax
b9064221	204	mov %r8,%rbx # restore %rbx
2bc3ad28	205	or %r9,%rax
376729e1 AP	206	ret
376729e1 AP	207	.size OPENSSL_ia32_cpuid,.-OPENSSL_ia32_cpuid
b2dba9bf AP	208
b2dba9bf AP	209	.globl OPENSSL_cleanse
aa8f38e4	210	.type OPENSSL_cleanse,\@abi-omnipotent
b2dba9bf AP	211	.align 16
	212	OPENSSL_cleanse:
	213	xor %rax,%rax
aa8f38e4	214	cmp \$15,$arg2
b2dba9bf	215	jae .Lot
7676eebf AP	216	cmp \$0,$arg2
7676eebf AP	217	je .Lret
b2dba9bf	218	.Little:
aa8f38e4 AP	219	mov %al,($arg1)
	220	sub \$1,$arg2
	221	lea 1($arg1),$arg1
b2dba9bf	222	jnz .Little
1fd79f66 AP	223	.Lret:
1fd79f66 AP	224	ret
b2dba9bf AP	225	.align 16
b2dba9bf AP	226	.Lot:
aa8f38e4	227	test \$7,$arg1
b2dba9bf	228	jz .Laligned
aa8f38e4 AP	229	mov %al,($arg1)
	230	lea -1($arg2),$arg2
	231	lea 1($arg1),$arg1
b2dba9bf AP	232	jmp .Lot
b2dba9bf AP	233	.Laligned:
aa8f38e4 AP	234	mov %rax,($arg1)
	235	lea -8($arg2),$arg2
	236	test \$-8,$arg2
	237	lea 8($arg1),$arg1
b2dba9bf	238	jnz .Laligned
aa8f38e4	239	cmp \$0,$arg2
b2dba9bf AP	240	jne .Little
	241	ret
	242	.size OPENSSL_cleanse,.-OPENSSL_cleanse
e33826f0 AP	243
	244	.globl CRYPTO_memcmp
	245	.type CRYPTO_memcmp,\@abi-omnipotent
	246	.align 16
	247	CRYPTO_memcmp:
	248	xor %rax,%rax
	249	xor %r10,%r10
	250	cmp \$0,$arg3
	251	je .Lno_data
	252	.Loop_cmp:
	253	mov ($arg1),%r10b
	254	lea 1($arg1),$arg1
	255	xor ($arg2),%r10b
	256	lea 1($arg2),$arg2
	257	or %r10b,%al
	258	dec $arg3
	259	jnz .Loop_cmp
	260	neg %rax
	261	shr \$63,%rax
	262	.Lno_data:
	263	ret
	264	.size CRYPTO_memcmp,.-CRYPTO_memcmp
14e21f86	265	___
aa8f38e4 AP	266
	267	print<<___ if (!$win64);
	268	.globl OPENSSL_wipe_cpu
	269	.type OPENSSL_wipe_cpu,\@abi-omnipotent
	270	.align 16
	271	OPENSSL_wipe_cpu:
	272	pxor %xmm0,%xmm0
	273	pxor %xmm1,%xmm1
	274	pxor %xmm2,%xmm2
	275	pxor %xmm3,%xmm3
	276	pxor %xmm4,%xmm4
	277	pxor %xmm5,%xmm5
	278	pxor %xmm6,%xmm6
	279	pxor %xmm7,%xmm7
	280	pxor %xmm8,%xmm8
	281	pxor %xmm9,%xmm9
	282	pxor %xmm10,%xmm10
	283	pxor %xmm11,%xmm11
	284	pxor %xmm12,%xmm12
	285	pxor %xmm13,%xmm13
	286	pxor %xmm14,%xmm14
	287	pxor %xmm15,%xmm15
	288	xorq %rcx,%rcx
	289	xorq %rdx,%rdx
	290	xorq %rsi,%rsi
	291	xorq %rdi,%rdi
	292	xorq %r8,%r8
	293	xorq %r9,%r9
	294	xorq %r10,%r10
	295	xorq %r11,%r11
	296	leaq 8(%rsp),%rax
	297	ret
	298	.size OPENSSL_wipe_cpu,.-OPENSSL_wipe_cpu
	299	___
	300	print<<___ if ($win64);
	301	.globl OPENSSL_wipe_cpu
	302	.type OPENSSL_wipe_cpu,\@abi-omnipotent
	303	.align 16
	304	OPENSSL_wipe_cpu:
	305	pxor %xmm0,%xmm0
	306	pxor %xmm1,%xmm1
	307	pxor %xmm2,%xmm2
	308	pxor %xmm3,%xmm3
	309	pxor %xmm4,%xmm4
	310	pxor %xmm5,%xmm5
	311	xorq %rcx,%rcx
	312	xorq %rdx,%rdx
	313	xorq %r8,%r8
	314	xorq %r9,%r9
	315	xorq %r10,%r10
	316	xorq %r11,%r11
	317	leaq 8(%rsp),%rax
	318	ret
	319	.size OPENSSL_wipe_cpu,.-OPENSSL_wipe_cpu
	320	___
5fabb88a AP	321	{
	322	my $out="%r10";
	323	my $cnt="%rcx";
	324	my $max="%r11";
	325	my $lasttick="%r8d";
	326	my $lastdiff="%r9d";
	327	my $redzone=win64?8:-8;
	328
	329	print<<___;
	330	.globl OPENSSL_instrument_bus
	331	.type OPENSSL_instrument_bus,\@abi-omnipotent
	332	.align 16
	333	OPENSSL_instrument_bus:
	334	mov $arg1,$out # tribute to Win64
	335	mov $arg2,$cnt
	336	mov $arg2,$max
	337
	338	rdtsc # collect 1st tick
	339	mov %eax,$lasttick # lasttick = tick
	340	mov \$0,$lastdiff # lastdiff = 0
	341	clflush ($out)
b9064221	342	.byte 0xf0 # lock
5fabb88a AP	343	add $lastdiff,($out)
	344	jmp .Loop
	345	.align 16
	346	.Loop: rdtsc
	347	mov %eax,%edx
	348	sub $lasttick,%eax
	349	mov %edx,$lasttick
	350	mov %eax,$lastdiff
	351	clflush ($out)
b9064221	352	.byte 0xf0 # lock
5fabb88a AP	353	add %eax,($out)
	354	lea 4($out),$out
	355	sub \$1,$cnt
	356	jnz .Loop
	357
	358	mov $max,%rax
	359	ret
	360	.size OPENSSL_instrument_bus,.-OPENSSL_instrument_bus
	361
	362	.globl OPENSSL_instrument_bus2
	363	.type OPENSSL_instrument_bus2,\@abi-omnipotent
	364	.align 16
	365	OPENSSL_instrument_bus2:
	366	mov $arg1,$out # tribute to Win64
	367	mov $arg2,$cnt
	368	mov $arg3,$max
	369	mov $cnt,$redzone(%rsp)
	370
	371	rdtsc # collect 1st tick
	372	mov %eax,$lasttick # lasttick = tick
	373	mov \$0,$lastdiff # lastdiff = 0
	374
	375	clflush ($out)
b9064221	376	.byte 0xf0 # lock
5fabb88a AP	377	add $lastdiff,($out)
	378
	379	rdtsc # collect 1st diff
	380	mov %eax,%edx
	381	sub $lasttick,%eax # diff
	382	mov %edx,$lasttick # lasttick = tick
	383	mov %eax,$lastdiff # lastdiff = diff
	384	.Loop2:
	385	clflush ($out)
b9064221	386	.byte 0xf0 # lock
5fabb88a AP	387	add %eax,($out) # accumulate diff
	388
	389	sub \$1,$max
	390	jz .Ldone2
	391
	392	rdtsc
	393	mov %eax,%edx
	394	sub $lasttick,%eax # diff
	395	mov %edx,$lasttick # lasttick = tick
	396	cmp $lastdiff,%eax
	397	mov %eax,$lastdiff # lastdiff = diff
	398	mov \$0,%edx
	399	setne %dl
	400	sub %rdx,$cnt # conditional --$cnt
	401	lea ($out,%rdx,4),$out # conditional ++$out
	402	jnz .Loop2
	403
	404	.Ldone2:
	405	mov $redzone(%rsp),%rax
	406	sub $cnt,%rax
	407	ret
	408	.size OPENSSL_instrument_bus2,.-OPENSSL_instrument_bus2
	409	___
	410	}
aa8f38e4	411
9c940446 AP	412	sub gen_random {
9c940446 AP	413	my $rdop = shift;
301799b8	414	print<<___;
9c940446 AP	415	.globl OPENSSL_ia32_${rdop}
9c940446 AP	416	.type OPENSSL_ia32_${rdop},\@abi-omnipotent
301799b8	417	.align 16
9c940446	418	OPENSSL_ia32_${rdop}:
301799b8	419	mov \$8,%ecx
9c940446 AP	420	.Loop_${rdop}:
	421	${rdop} %rax
	422	jc .Lbreak_${rdop}
	423	loop .Loop_${rdop}
	424	.Lbreak_${rdop}:
301799b8 AP	425	cmp \$0,%rax
	426	cmove %rcx,%rax
	427	ret
9c940446	428	.size OPENSSL_ia32_${rdop},.-OPENSSL_ia32_${rdop}
f4d45640	429
9c940446 AP	430	.globl OPENSSL_ia32_${rdop}_bytes
9c940446 AP	431	.type OPENSSL_ia32_${rdop}_bytes,\@abi-omnipotent
f4d45640	432	.align 16
9c940446 AP	433	OPENSSL_ia32_${rdop}_bytes:
	434	xor %rax, %rax # return value
	435	cmp \$0,$arg2
	436	je .Ldone_${rdop}_bytes
	437
	438	mov \$8,%r11
	439	.Loop_${rdop}_bytes:
	440	${rdop} %r10
	441	jc .Lbreak_${rdop}_bytes
	442	dec %r11
	443	jnz .Loop_${rdop}_bytes
	444	jmp .Ldone_${rdop}_bytes
	445
	446	.align 16
	447	.Lbreak_${rdop}_bytes:
	448	cmp \$8,$arg2
	449	jb .Ltail_${rdop}_bytes
	450	mov %r10,($arg1)
	451	lea 8($arg1),$arg1
	452	add \$8,%rax
	453	sub \$8,$arg2
	454	jz .Ldone_${rdop}_bytes
	455	mov \$8,%r11
	456	jmp .Loop_${rdop}_bytes
	457
	458	.align 16
	459	.Ltail_${rdop}_bytes:
	460	mov %r10b,($arg1)
	461	lea 1($arg1),$arg1
	462	inc %rax
	463	shr \$8,%r8
	464	dec $arg2
	465	jnz .Ltail_${rdop}_bytes
	466
	467	.Ldone_${rdop}_bytes:
f4d45640	468	ret
9c940446	469	.size OPENSSL_ia32_${rdop}_bytes,.-OPENSSL_ia32_${rdop}_bytes
301799b8	470	___
9c940446 AP	471	}
	472	gen_random("rdrand");
	473	gen_random("rdseed");
301799b8	474
5d863367	475	close STDOUT; # flush