[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:
.cfi_startproc
	mov	%rbx,%r8		# save %rbx
.cfi_register	%rbx,%r8

	xor	%eax,%eax
	mov	%rax,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
	movd	%eax,%xmm0		# put aside processor id
	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:
	movd	%xmm0,%eax		# restore processor id
	and	\$0x0fff0ff0,%eax
	cmp	\$0x00050650,%eax	# Skylake-X
	jne	.Lnotskylakex
	and	\$0xfffeffff,%ebx	# ~(1<<16)
					# suppress AVX512F flag on Skylake-X
.Lnotskylakex:
	mov	%ebx,8(%rdi)		# save extended feature flags
	mov	%ecx,12(%rdi)
.Lno_extended_info:

	bt	\$27,%r9d		# check OSXSAVE bit
	jnc	.Lclear_avx
	xor	%ecx,%ecx		# XCR0
	.byte	0x0f,0x01,0xd0		# xgetbv
	and	\$0xe6,%eax		# isolate XMM, YMM and ZMM state support
	cmp	\$0xe6,%eax
	je	.Ldone
	andl	\$0x3fdeffff,8(%rdi)	# ~(1<<31|1<<30|1<<21|1<<16)
					# clear AVX512F+BW+VL+FIMA, all of
					# them are EVEX-encoded, which requires
					# ZMM state support even if one uses
					# only XMM and YMM :-(
	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
	mov	\$0x3fdeffdf,%eax	# ~(1<<31|1<<30|1<<21|1<<16|1<<5)
	and	%eax,8(%rdi)		# clear AVX2 and AVX512* bits
.Ldone:
	shl	\$32,%r9
	mov	%r10d,%eax
	mov	%r8,%rbx		# restore %rbx
.cfi_restore	%rbx
	or	%r9,%rax
	ret
.cfi_endproc
.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:
5e32cfb2	66	.cfi_startproc
b9064221	67	mov %rbx,%r8 # save %rbx
5e32cfb2	68	.cfi_register %rbx,%r8
9babf392 AP	69
9babf392 AP	70	xor %eax,%eax
d6ee8f3d	71	mov %rax,8(%rdi) # clear extended feature flags
9babf392	72	cpuid
761393bb AP	73	mov %eax,%r11d # max value for standard query level
761393bb AP	74
9babf392 AP	75	xor %eax,%eax
	76	cmp \$0x756e6547,%ebx # "Genu"
	77	setne %al
	78	mov %eax,%r9d
	79	cmp \$0x49656e69,%edx # "ineI"
	80	setne %al
	81	or %eax,%r9d
	82	cmp \$0x6c65746e,%ecx # "ntel"
	83	setne %al
5cd91b50	84	or %eax,%r9d # 0 indicates Intel CPU
5cd91b50 AP	85	jz .Lintel
	86
	87	cmp \$0x68747541,%ebx # "Auth"
	88	setne %al
	89	mov %eax,%r10d
	90	cmp \$0x69746E65,%edx # "enti"
	91	setne %al
	92	or %eax,%r10d
	93	cmp \$0x444D4163,%ecx # "cAMD"
	94	setne %al
	95	or %eax,%r10d # 0 indicates AMD CPU
	96	jnz .Lintel
	97
761393bb	98	# AMD specific
5cd91b50 AP	99	mov \$0x80000000,%eax
5cd91b50 AP	100	cpuid
b9064221 AP	101	cmp \$0x80000001,%eax
	102	jb .Lintel
	103	mov %eax,%r10d
	104	mov \$0x80000001,%eax
	105	cpuid
	106	or %ecx,%r9d
	107	and \$0x00000801,%r9d # isolate AMD XOP bit, 1<<11
	108
	109	cmp \$0x80000008,%r10d
5cd91b50 AP	110	jb .Lintel
	111
	112	mov \$0x80000008,%eax
	113	cpuid
	114	movzb %cl,%r10 # number of cores - 1
	115	inc %r10 # number of cores
9babf392	116
761393bb AP	117	mov \$1,%eax
	118	cpuid
	119	bt \$28,%edx # test hyper-threading bit
b9064221	120	jnc .Lgeneric
761393bb AP	121	shr \$16,%ebx # number of logical processors
761393bb AP	122	cmp %r10b,%bl
b9064221	123	ja .Lgeneric
761393bb	124	and \$0xefffffff,%edx # ~(1<<28)
b9064221	125	jmp .Lgeneric
761393bb	126
5cd91b50	127	.Lintel:
761393bb AP	128	cmp \$4,%r11d
	129	mov \$-1,%r10d
	130	jb .Lnocacheinfo
	131
	132	mov \$4,%eax
	133	mov \$0,%ecx # query L1D
	134	cpuid
	135	mov %eax,%r10d
	136	shr \$14,%r10d
	137	and \$0xfff,%r10d # number of cores -1 per L1D
	138
	139	.Lnocacheinfo:
932cc129	140	mov \$1,%eax
376729e1	141	cpuid
79337628	142	movd %eax,%xmm0 # put aside processor id
4bb90087	143	and \$0xbfefffff,%edx # force reserved bits to 0
932cc129	144	cmp \$0,%r9d
9babf392	145	jne .Lnotintel
4bb90087	146	or \$0x40000000,%edx # set reserved bit#30 on Intel CPUs
9babf392 AP	147	and \$15,%ah
9babf392 AP	148	cmp \$15,%ah # examine Family ID
64d92d74	149	jne .LnotP4
4bb90087	150	or \$0x00100000,%edx # set reserved bit#20 to engage RC4_CHAR
64d92d74 AP	151	.LnotP4:
	152	cmp \$6,%ah
	153	jne .Lnotintel
d84df594	154	and \$0x0fff0ff0,%eax
64d92d74 AP	155	cmp \$0x00050670,%eax # Knights Landing
	156	je .Lknights
	157	cmp \$0x00080650,%eax # Knights Mill (according to sde)
	158	jne .Lnotintel
	159	.Lknights:
	160	and \$0xfbffffff,%ecx # clear XSAVE flag to mimic Silvermont
	161
9babf392	162	.Lnotintel:
3df2eff4	163	bt \$28,%edx # test hyper-threading bit
b9064221	164	jnc .Lgeneric
761393bb AP	165	and \$0xefffffff,%edx # ~(1<<28)
761393bb AP	166	cmp \$0,%r10d
b9064221	167	je .Lgeneric
761393bb AP	168
761393bb AP	169	or \$0x10000000,%edx # 1<<28
9babf392	170	shr \$16,%ebx
761393bb	171	cmp \$1,%bl # see if cache is shared
b9064221	172	ja .Lgeneric
932cc129	173	and \$0xefffffff,%edx # ~(1<<28)
b9064221 AP	174	.Lgeneric:
	175	and \$0x00000800,%r9d # isolate AMD XOP flag
	176	and \$0xfffff7ff,%ecx
2bc3ad28	177	or %ecx,%r9d # merge AMD XOP flag
b9064221	178
2bc3ad28	179	mov %edx,%r10d # %r9d:%r10d is copy of %ecx:%edx
1aed5e1a AP	180
	181	cmp \$7,%r11d
	182	jb .Lno_extended_info
	183	mov \$7,%eax
	184	xor %ecx,%ecx
	185	cpuid
64d92d74 AP	186	bt \$26,%r9d # check XSAVE bit, cleared on Knights
	187	jc .Lnotknights
	188	and \$0xfff7ffff,%ebx # clear ADCX/ADOX flag
	189	.Lnotknights:
79337628 AP	190	movd %xmm0,%eax # restore processor id
	191	and \$0x0fff0ff0,%eax
	192	cmp \$0x00050650,%eax # Skylake-X
	193	jne .Lnotskylakex
	194	and \$0xfffeffff,%ebx # ~(1<<16)
	195	# suppress AVX512F flag on Skylake-X
	196	.Lnotskylakex:
1aed5e1a	197	mov %ebx,8(%rdi) # save extended feature flags
d6ee8f3d	198	mov %ecx,12(%rdi)
1aed5e1a AP	199	.Lno_extended_info:
1aed5e1a AP	200
2bc3ad28	201	bt \$27,%r9d # check OSXSAVE bit
b9064221 AP	202	jnc .Lclear_avx
	203	xor %ecx,%ecx # XCR0
	204	.byte 0x0f,0x01,0xd0 # xgetbv
66bee01c AP	205	and \$0xe6,%eax # isolate XMM, YMM and ZMM state support
	206	cmp \$0xe6,%eax
	207	je .Ldone
88ac224c AP	208	andl \$0x3fdeffff,8(%rdi) # ~(1<<31\|1<<30\|1<<21\|1<<16)
	209	# clear AVX512F+BW+VL+FIMA, all of
	210	# them are EVEX-encoded, which requires
	211	# ZMM state support even if one uses
	212	# only XMM and YMM :-(
b9064221 AP	213	and \$6,%eax # isolate XMM and YMM state support
	214	cmp \$6,%eax
	215	je .Ldone
	216	.Lclear_avx:
	217	mov \$0xefffe7ff,%eax # ~(1<<28\|1<<12\|1<<11)
2bc3ad28	218	and %eax,%r9d # clear AVX, FMA and AMD XOP bits
66bee01c	219	mov \$0x3fdeffdf,%eax # ~(1<<31\|1<<30\|1<<21\|1<<16\|1<<5)
d67e7554	220	and %eax,8(%rdi) # clear AVX2 and AVX512* bits
b9064221	221	.Ldone:
2bc3ad28 AP	222	shl \$32,%r9
2bc3ad28 AP	223	mov %r10d,%eax
b9064221	224	mov %r8,%rbx # restore %rbx
5e32cfb2	225	.cfi_restore %rbx
2bc3ad28	226	or %r9,%rax
376729e1	227	ret
5e32cfb2	228	.cfi_endproc
376729e1	229	.size OPENSSL_ia32_cpuid,.-OPENSSL_ia32_cpuid
b2dba9bf AP	230
b2dba9bf AP	231	.globl OPENSSL_cleanse
aa8f38e4	232	.type OPENSSL_cleanse,\@abi-omnipotent
b2dba9bf AP	233	.align 16
	234	OPENSSL_cleanse:
	235	xor %rax,%rax
aa8f38e4	236	cmp \$15,$arg2
b2dba9bf	237	jae .Lot
7676eebf AP	238	cmp \$0,$arg2
7676eebf AP	239	je .Lret
b2dba9bf	240	.Little:
aa8f38e4 AP	241	mov %al,($arg1)
	242	sub \$1,$arg2
	243	lea 1($arg1),$arg1
b2dba9bf	244	jnz .Little
1fd79f66 AP	245	.Lret:
1fd79f66 AP	246	ret
b2dba9bf AP	247	.align 16
b2dba9bf AP	248	.Lot:
aa8f38e4	249	test \$7,$arg1
b2dba9bf	250	jz .Laligned
aa8f38e4 AP	251	mov %al,($arg1)
	252	lea -1($arg2),$arg2
	253	lea 1($arg1),$arg1
b2dba9bf AP	254	jmp .Lot
b2dba9bf AP	255	.Laligned:
aa8f38e4 AP	256	mov %rax,($arg1)
	257	lea -8($arg2),$arg2
	258	test \$-8,$arg2
	259	lea 8($arg1),$arg1
b2dba9bf	260	jnz .Laligned
aa8f38e4	261	cmp \$0,$arg2
b2dba9bf AP	262	jne .Little
	263	ret
	264	.size OPENSSL_cleanse,.-OPENSSL_cleanse
e33826f0 AP	265
	266	.globl CRYPTO_memcmp
	267	.type CRYPTO_memcmp,\@abi-omnipotent
	268	.align 16
	269	CRYPTO_memcmp:
	270	xor %rax,%rax
	271	xor %r10,%r10
	272	cmp \$0,$arg3
	273	je .Lno_data
	274	.Loop_cmp:
	275	mov ($arg1),%r10b
	276	lea 1($arg1),$arg1
	277	xor ($arg2),%r10b
	278	lea 1($arg2),$arg2
	279	or %r10b,%al
	280	dec $arg3
	281	jnz .Loop_cmp
	282	neg %rax
	283	shr \$63,%rax
	284	.Lno_data:
	285	ret
	286	.size CRYPTO_memcmp,.-CRYPTO_memcmp
14e21f86	287	___
aa8f38e4 AP	288
	289	print<<___ if (!$win64);
	290	.globl OPENSSL_wipe_cpu
	291	.type OPENSSL_wipe_cpu,\@abi-omnipotent
	292	.align 16
	293	OPENSSL_wipe_cpu:
	294	pxor %xmm0,%xmm0
	295	pxor %xmm1,%xmm1
	296	pxor %xmm2,%xmm2
	297	pxor %xmm3,%xmm3
	298	pxor %xmm4,%xmm4
	299	pxor %xmm5,%xmm5
	300	pxor %xmm6,%xmm6
	301	pxor %xmm7,%xmm7
	302	pxor %xmm8,%xmm8
	303	pxor %xmm9,%xmm9
	304	pxor %xmm10,%xmm10
	305	pxor %xmm11,%xmm11
	306	pxor %xmm12,%xmm12
	307	pxor %xmm13,%xmm13
	308	pxor %xmm14,%xmm14
	309	pxor %xmm15,%xmm15
	310	xorq %rcx,%rcx
	311	xorq %rdx,%rdx
	312	xorq %rsi,%rsi
	313	xorq %rdi,%rdi
	314	xorq %r8,%r8
	315	xorq %r9,%r9
	316	xorq %r10,%r10
	317	xorq %r11,%r11
	318	leaq 8(%rsp),%rax
	319	ret
	320	.size OPENSSL_wipe_cpu,.-OPENSSL_wipe_cpu
	321	___
	322	print<<___ if ($win64);
	323	.globl OPENSSL_wipe_cpu
	324	.type OPENSSL_wipe_cpu,\@abi-omnipotent
	325	.align 16
	326	OPENSSL_wipe_cpu:
	327	pxor %xmm0,%xmm0
	328	pxor %xmm1,%xmm1
	329	pxor %xmm2,%xmm2
	330	pxor %xmm3,%xmm3
	331	pxor %xmm4,%xmm4
	332	pxor %xmm5,%xmm5
	333	xorq %rcx,%rcx
	334	xorq %rdx,%rdx
	335	xorq %r8,%r8
	336	xorq %r9,%r9
	337	xorq %r10,%r10
	338	xorq %r11,%r11
	339	leaq 8(%rsp),%rax
	340	ret
	341	.size OPENSSL_wipe_cpu,.-OPENSSL_wipe_cpu
	342	___
5fabb88a AP	343	{
	344	my $out="%r10";
	345	my $cnt="%rcx";
	346	my $max="%r11";
	347	my $lasttick="%r8d";
	348	my $lastdiff="%r9d";
	349	my $redzone=win64?8:-8;
	350
	351	print<<___;
	352	.globl OPENSSL_instrument_bus
	353	.type OPENSSL_instrument_bus,\@abi-omnipotent
	354	.align 16
	355	OPENSSL_instrument_bus:
	356	mov $arg1,$out # tribute to Win64
	357	mov $arg2,$cnt
	358	mov $arg2,$max
	359
	360	rdtsc # collect 1st tick
	361	mov %eax,$lasttick # lasttick = tick
	362	mov \$0,$lastdiff # lastdiff = 0
	363	clflush ($out)
b9064221	364	.byte 0xf0 # lock
5fabb88a AP	365	add $lastdiff,($out)
	366	jmp .Loop
	367	.align 16
	368	.Loop: rdtsc
	369	mov %eax,%edx
	370	sub $lasttick,%eax
	371	mov %edx,$lasttick
	372	mov %eax,$lastdiff
	373	clflush ($out)
b9064221	374	.byte 0xf0 # lock
5fabb88a AP	375	add %eax,($out)
	376	lea 4($out),$out
	377	sub \$1,$cnt
	378	jnz .Loop
	379
	380	mov $max,%rax
	381	ret
	382	.size OPENSSL_instrument_bus,.-OPENSSL_instrument_bus
	383
	384	.globl OPENSSL_instrument_bus2
	385	.type OPENSSL_instrument_bus2,\@abi-omnipotent
	386	.align 16
	387	OPENSSL_instrument_bus2:
	388	mov $arg1,$out # tribute to Win64
	389	mov $arg2,$cnt
	390	mov $arg3,$max
	391	mov $cnt,$redzone(%rsp)
	392
	393	rdtsc # collect 1st tick
	394	mov %eax,$lasttick # lasttick = tick
	395	mov \$0,$lastdiff # lastdiff = 0
	396
	397	clflush ($out)
b9064221	398	.byte 0xf0 # lock
5fabb88a AP	399	add $lastdiff,($out)
	400
	401	rdtsc # collect 1st diff
	402	mov %eax,%edx
	403	sub $lasttick,%eax # diff
	404	mov %edx,$lasttick # lasttick = tick
	405	mov %eax,$lastdiff # lastdiff = diff
	406	.Loop2:
	407	clflush ($out)
b9064221	408	.byte 0xf0 # lock
5fabb88a AP	409	add %eax,($out) # accumulate diff
	410
	411	sub \$1,$max
	412	jz .Ldone2
	413
	414	rdtsc
	415	mov %eax,%edx
	416	sub $lasttick,%eax # diff
	417	mov %edx,$lasttick # lasttick = tick
	418	cmp $lastdiff,%eax
	419	mov %eax,$lastdiff # lastdiff = diff
	420	mov \$0,%edx
	421	setne %dl
	422	sub %rdx,$cnt # conditional --$cnt
	423	lea ($out,%rdx,4),$out # conditional ++$out
	424	jnz .Loop2
	425
	426	.Ldone2:
	427	mov $redzone(%rsp),%rax
	428	sub $cnt,%rax
	429	ret
	430	.size OPENSSL_instrument_bus2,.-OPENSSL_instrument_bus2
	431	___
	432	}
aa8f38e4	433
9c940446 AP	434	sub gen_random {
9c940446 AP	435	my $rdop = shift;
301799b8	436	print<<___;
9c940446 AP	437	.globl OPENSSL_ia32_${rdop}
9c940446 AP	438	.type OPENSSL_ia32_${rdop},\@abi-omnipotent
301799b8	439	.align 16
9c940446	440	OPENSSL_ia32_${rdop}:
301799b8	441	mov \$8,%ecx
9c940446 AP	442	.Loop_${rdop}:
	443	${rdop} %rax
	444	jc .Lbreak_${rdop}
	445	loop .Loop_${rdop}
	446	.Lbreak_${rdop}:
301799b8 AP	447	cmp \$0,%rax
	448	cmove %rcx,%rax
	449	ret
9c940446	450	.size OPENSSL_ia32_${rdop},.-OPENSSL_ia32_${rdop}
f4d45640	451
9c940446 AP	452	.globl OPENSSL_ia32_${rdop}_bytes
9c940446 AP	453	.type OPENSSL_ia32_${rdop}_bytes,\@abi-omnipotent
f4d45640	454	.align 16
9c940446 AP	455	OPENSSL_ia32_${rdop}_bytes:
	456	xor %rax, %rax # return value
	457	cmp \$0,$arg2
	458	je .Ldone_${rdop}_bytes
	459
	460	mov \$8,%r11
	461	.Loop_${rdop}_bytes:
	462	${rdop} %r10
	463	jc .Lbreak_${rdop}_bytes
	464	dec %r11
	465	jnz .Loop_${rdop}_bytes
	466	jmp .Ldone_${rdop}_bytes
	467
	468	.align 16
	469	.Lbreak_${rdop}_bytes:
	470	cmp \$8,$arg2
	471	jb .Ltail_${rdop}_bytes
	472	mov %r10,($arg1)
	473	lea 8($arg1),$arg1
	474	add \$8,%rax
	475	sub \$8,$arg2
	476	jz .Ldone_${rdop}_bytes
	477	mov \$8,%r11
	478	jmp .Loop_${rdop}_bytes
	479
	480	.align 16
	481	.Ltail_${rdop}_bytes:
	482	mov %r10b,($arg1)
	483	lea 1($arg1),$arg1
	484	inc %rax
	485	shr \$8,%r8
	486	dec $arg2
	487	jnz .Ltail_${rdop}_bytes
	488
	489	.Ldone_${rdop}_bytes:
f4d45640	490	ret
9c940446	491	.size OPENSSL_ia32_${rdop}_bytes,.-OPENSSL_ia32_${rdop}_bytes
301799b8	492	___
9c940446 AP	493	}
	494	gen_random("rdrand");
	495	gen_random("rdseed");
301799b8	496
5d863367	497	close STDOUT; # flush