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

#!/usr/bin/env perl

$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 3rd word
	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

	cmp	\$7,%r11d
	jb	.Lnocacheinfo

	mov	\$7,%eax
	xor	%ecx,%ecx
	cpuid
	mov	%ebx,8(%rdi)

.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	.Lnotintel
	or	\$0x00100000,%edx	# set reserved bit#20 to engage RC4_CHAR
.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
	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)	# cleax 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
___

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
___
}

print<<___;
.globl	OPENSSL_ia32_rdrand
.type	OPENSSL_ia32_rdrand,\@abi-omnipotent
.align	16
OPENSSL_ia32_rdrand:
	mov	\$8,%ecx
.Loop_rdrand:
	rdrand	%rax
	jc	.Lbreak_rdrand
	loop	.Loop_rdrand
.Lbreak_rdrand:
	cmp	\$0,%rax
	cmove	%rcx,%rax
	ret
.size	OPENSSL_ia32_rdrand,.-OPENSSL_ia32_rdrand

.globl	OPENSSL_ia32_rdseed
.type	OPENSSL_ia32_rdseed,\@abi-omnipotent
.align	16
OPENSSL_ia32_rdseed:
	mov	\$8,%ecx
.Loop_rdseed:
	rdseed	%rax
	jc	.Lbreak_rdseed
	loop	.Loop_rdseed
.Lbreak_rdseed:
	cmp	\$0,%rax
	cmove	%rcx,%rax
	ret
.size	OPENSSL_ia32_rdseed,.-OPENSSL_ia32_rdseed
___

close STDOUT;	# flush
Commit	Line	Data
14e21f86 AP	1	#!/usr/bin/env perl
14e21f86 AP	2
aa8f38e4 AP	3	$flavour = shift;
	4	$output = shift;
	5	if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
14e21f86	6
aa8f38e4	7	$win64=0; $win64=1 if ($flavour =~ /[nm]asm\|mingw64/ \|\| $output =~ /\.asm$/);
376729e1	8
aa8f38e4	9	$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
301799b8 AP	10	( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or
	11	( $xlate="${dir}perlasm/x86_64-xlate.pl" and -f $xlate) or
	12	die "can't locate x86_64-xlate.pl";
	13
46bf83f0 AP	14	open OUT,"\| \"$^X\" $xlate $flavour $output";
46bf83f0 AP	15	STDOUT=OUT;
376729e1	16
5fabb88a AP	17	($arg1,$arg2,$arg3,$arg4)=$win64?("%rcx","%rdx","%r8", "%r9") : # Win64 order
	18	("%rdi","%rsi","%rdx","%rcx"); # Unix order
	19
aa8f38e4 AP	20	print<<___;
aa8f38e4 AP	21	.extern OPENSSL_cpuid_setup
ddc20d4d	22	.hidden OPENSSL_cpuid_setup
aa8f38e4 AP	23	.section .init
aa8f38e4 AP	24	call OPENSSL_cpuid_setup
932cc129	25
ddc20d4d	26	.hidden OPENSSL_ia32cap_P
c5cd28bd	27	.comm OPENSSL_ia32cap_P,16,4
ddc20d4d	28
14e21f86	29	.text
376729e1 AP	30
376729e1 AP	31	.globl OPENSSL_atomic_add
aa8f38e4	32	.type OPENSSL_atomic_add,\@abi-omnipotent
376729e1 AP	33	.align 16
376729e1 AP	34	OPENSSL_atomic_add:
aa8f38e4 AP	35	movl ($arg1),%eax
	36	.Lspin: leaq ($arg2,%rax),%r8
	37	.byte 0xf0 # lock
	38	cmpxchgl %r8d,($arg1)
376729e1	39	jne .Lspin
e442c362	40	movl %r8d,%eax
aa8f38e4	41	.byte 0x48,0x98 # cltq/cdqe
376729e1 AP	42	ret
	43	.size OPENSSL_atomic_add,.-OPENSSL_atomic_add
	44
932cc129 AP	45	.globl OPENSSL_rdtsc
	46	.type OPENSSL_rdtsc,\@abi-omnipotent
	47	.align 16
	48	OPENSSL_rdtsc:
	49	rdtsc
	50	shl \$32,%rdx
	51	or %rdx,%rax
	52	ret
	53	.size OPENSSL_rdtsc,.-OPENSSL_rdtsc
	54
376729e1	55	.globl OPENSSL_ia32_cpuid
c5cd28bd	56	.type OPENSSL_ia32_cpuid,\@function,1
376729e1 AP	57	.align 16
376729e1 AP	58	OPENSSL_ia32_cpuid:
b9064221	59	mov %rbx,%r8 # save %rbx
9babf392 AP	60
9babf392 AP	61	xor %eax,%eax
c5cd28bd	62	mov %eax,8(%rdi) # clear 3rd word
9babf392	63	cpuid
761393bb AP	64	mov %eax,%r11d # max value for standard query level
761393bb AP	65
9babf392 AP	66	xor %eax,%eax
	67	cmp \$0x756e6547,%ebx # "Genu"
	68	setne %al
	69	mov %eax,%r9d
	70	cmp \$0x49656e69,%edx # "ineI"
	71	setne %al
	72	or %eax,%r9d
	73	cmp \$0x6c65746e,%ecx # "ntel"
	74	setne %al
5cd91b50	75	or %eax,%r9d # 0 indicates Intel CPU
5cd91b50 AP	76	jz .Lintel
	77
	78	cmp \$0x68747541,%ebx # "Auth"
	79	setne %al
	80	mov %eax,%r10d
	81	cmp \$0x69746E65,%edx # "enti"
	82	setne %al
	83	or %eax,%r10d
	84	cmp \$0x444D4163,%ecx # "cAMD"
	85	setne %al
	86	or %eax,%r10d # 0 indicates AMD CPU
	87	jnz .Lintel
	88
761393bb	89	# AMD specific
5cd91b50 AP	90	mov \$0x80000000,%eax
5cd91b50 AP	91	cpuid
b9064221 AP	92	cmp \$0x80000001,%eax
	93	jb .Lintel
	94	mov %eax,%r10d
	95	mov \$0x80000001,%eax
	96	cpuid
	97	or %ecx,%r9d
	98	and \$0x00000801,%r9d # isolate AMD XOP bit, 1<<11
	99
	100	cmp \$0x80000008,%r10d
5cd91b50 AP	101	jb .Lintel
	102
	103	mov \$0x80000008,%eax
	104	cpuid
	105	movzb %cl,%r10 # number of cores - 1
	106	inc %r10 # number of cores
9babf392	107
761393bb AP	108	mov \$1,%eax
	109	cpuid
	110	bt \$28,%edx # test hyper-threading bit
b9064221	111	jnc .Lgeneric
761393bb AP	112	shr \$16,%ebx # number of logical processors
761393bb AP	113	cmp %r10b,%bl
b9064221	114	ja .Lgeneric
761393bb	115	and \$0xefffffff,%edx # ~(1<<28)
b9064221	116	jmp .Lgeneric
761393bb	117
5cd91b50	118	.Lintel:
761393bb AP	119	cmp \$4,%r11d
	120	mov \$-1,%r10d
	121	jb .Lnocacheinfo
	122
	123	mov \$4,%eax
	124	mov \$0,%ecx # query L1D
	125	cpuid
	126	mov %eax,%r10d
	127	shr \$14,%r10d
	128	and \$0xfff,%r10d # number of cores -1 per L1D
	129
c5cd28bd AP	130	cmp \$7,%r11d
	131	jb .Lnocacheinfo
	132
	133	mov \$7,%eax
	134	xor %ecx,%ecx
	135	cpuid
	136	mov %ebx,8(%rdi)
	137
761393bb	138	.Lnocacheinfo:
932cc129	139	mov \$1,%eax
376729e1	140	cpuid
4bb90087	141	and \$0xbfefffff,%edx # force reserved bits to 0
932cc129	142	cmp \$0,%r9d
9babf392	143	jne .Lnotintel
4bb90087	144	or \$0x40000000,%edx # set reserved bit#30 on Intel CPUs
9babf392 AP	145	and \$15,%ah
9babf392 AP	146	cmp \$15,%ah # examine Family ID
4bb90087 AP	147	jne .Lnotintel
4bb90087 AP	148	or \$0x00100000,%edx # set reserved bit#20 to engage RC4_CHAR
9babf392	149	.Lnotintel:
3df2eff4	150	bt \$28,%edx # test hyper-threading bit
b9064221	151	jnc .Lgeneric
761393bb AP	152	and \$0xefffffff,%edx # ~(1<<28)
761393bb AP	153	cmp \$0,%r10d
b9064221	154	je .Lgeneric
761393bb AP	155
761393bb AP	156	or \$0x10000000,%edx # 1<<28
9babf392	157	shr \$16,%ebx
761393bb	158	cmp \$1,%bl # see if cache is shared
b9064221	159	ja .Lgeneric
932cc129	160	and \$0xefffffff,%edx # ~(1<<28)
b9064221 AP	161	.Lgeneric:
	162	and \$0x00000800,%r9d # isolate AMD XOP flag
	163	and \$0xfffff7ff,%ecx
2bc3ad28	164	or %ecx,%r9d # merge AMD XOP flag
b9064221	165
2bc3ad28 AP	166	mov %edx,%r10d # %r9d:%r10d is copy of %ecx:%edx
2bc3ad28 AP	167	bt \$27,%r9d # check OSXSAVE bit
b9064221 AP	168	jnc .Lclear_avx
	169	xor %ecx,%ecx # XCR0
	170	.byte 0x0f,0x01,0xd0 # xgetbv
	171	and \$6,%eax # isolate XMM and YMM state support
	172	cmp \$6,%eax
	173	je .Ldone
	174	.Lclear_avx:
	175	mov \$0xefffe7ff,%eax # ~(1<<28\|1<<12\|1<<11)
2bc3ad28	176	and %eax,%r9d # clear AVX, FMA and AMD XOP bits
c5cd28bd	177	andl \$0xffffffdf,8(%rdi) # cleax AVX2, ~(1<<5)
b9064221	178	.Ldone:
2bc3ad28 AP	179	shl \$32,%r9
2bc3ad28 AP	180	mov %r10d,%eax
b9064221	181	mov %r8,%rbx # restore %rbx
2bc3ad28	182	or %r9,%rax
376729e1 AP	183	ret
376729e1 AP	184	.size OPENSSL_ia32_cpuid,.-OPENSSL_ia32_cpuid
b2dba9bf AP	185
b2dba9bf AP	186	.globl OPENSSL_cleanse
aa8f38e4	187	.type OPENSSL_cleanse,\@abi-omnipotent
b2dba9bf AP	188	.align 16
	189	OPENSSL_cleanse:
	190	xor %rax,%rax
aa8f38e4	191	cmp \$15,$arg2
b2dba9bf	192	jae .Lot
7676eebf AP	193	cmp \$0,$arg2
7676eebf AP	194	je .Lret
b2dba9bf	195	.Little:
aa8f38e4 AP	196	mov %al,($arg1)
	197	sub \$1,$arg2
	198	lea 1($arg1),$arg1
b2dba9bf	199	jnz .Little
1fd79f66 AP	200	.Lret:
1fd79f66 AP	201	ret
b2dba9bf AP	202	.align 16
b2dba9bf AP	203	.Lot:
aa8f38e4	204	test \$7,$arg1
b2dba9bf	205	jz .Laligned
aa8f38e4 AP	206	mov %al,($arg1)
	207	lea -1($arg2),$arg2
	208	lea 1($arg1),$arg1
b2dba9bf AP	209	jmp .Lot
b2dba9bf AP	210	.Laligned:
aa8f38e4 AP	211	mov %rax,($arg1)
	212	lea -8($arg2),$arg2
	213	test \$-8,$arg2
	214	lea 8($arg1),$arg1
b2dba9bf	215	jnz .Laligned
aa8f38e4	216	cmp \$0,$arg2
b2dba9bf AP	217	jne .Little
	218	ret
	219	.size OPENSSL_cleanse,.-OPENSSL_cleanse
14e21f86	220	___
aa8f38e4 AP	221
	222	print<<___ if (!$win64);
	223	.globl OPENSSL_wipe_cpu
	224	.type OPENSSL_wipe_cpu,\@abi-omnipotent
	225	.align 16
	226	OPENSSL_wipe_cpu:
	227	pxor %xmm0,%xmm0
	228	pxor %xmm1,%xmm1
	229	pxor %xmm2,%xmm2
	230	pxor %xmm3,%xmm3
	231	pxor %xmm4,%xmm4
	232	pxor %xmm5,%xmm5
	233	pxor %xmm6,%xmm6
	234	pxor %xmm7,%xmm7
	235	pxor %xmm8,%xmm8
	236	pxor %xmm9,%xmm9
	237	pxor %xmm10,%xmm10
	238	pxor %xmm11,%xmm11
	239	pxor %xmm12,%xmm12
	240	pxor %xmm13,%xmm13
	241	pxor %xmm14,%xmm14
	242	pxor %xmm15,%xmm15
	243	xorq %rcx,%rcx
	244	xorq %rdx,%rdx
	245	xorq %rsi,%rsi
	246	xorq %rdi,%rdi
	247	xorq %r8,%r8
	248	xorq %r9,%r9
	249	xorq %r10,%r10
	250	xorq %r11,%r11
	251	leaq 8(%rsp),%rax
	252	ret
	253	.size OPENSSL_wipe_cpu,.-OPENSSL_wipe_cpu
	254	___
	255	print<<___ if ($win64);
	256	.globl OPENSSL_wipe_cpu
	257	.type OPENSSL_wipe_cpu,\@abi-omnipotent
	258	.align 16
	259	OPENSSL_wipe_cpu:
	260	pxor %xmm0,%xmm0
	261	pxor %xmm1,%xmm1
	262	pxor %xmm2,%xmm2
	263	pxor %xmm3,%xmm3
	264	pxor %xmm4,%xmm4
	265	pxor %xmm5,%xmm5
	266	xorq %rcx,%rcx
	267	xorq %rdx,%rdx
	268	xorq %r8,%r8
	269	xorq %r9,%r9
	270	xorq %r10,%r10
	271	xorq %r11,%r11
	272	leaq 8(%rsp),%rax
	273	ret
	274	.size OPENSSL_wipe_cpu,.-OPENSSL_wipe_cpu
	275	___
5fabb88a AP	276	{
	277	my $out="%r10";
	278	my $cnt="%rcx";
	279	my $max="%r11";
	280	my $lasttick="%r8d";
	281	my $lastdiff="%r9d";
	282	my $redzone=win64?8:-8;
	283
	284	print<<___;
	285	.globl OPENSSL_instrument_bus
	286	.type OPENSSL_instrument_bus,\@abi-omnipotent
	287	.align 16
	288	OPENSSL_instrument_bus:
	289	mov $arg1,$out # tribute to Win64
	290	mov $arg2,$cnt
	291	mov $arg2,$max
	292
	293	rdtsc # collect 1st tick
	294	mov %eax,$lasttick # lasttick = tick
	295	mov \$0,$lastdiff # lastdiff = 0
	296	clflush ($out)
b9064221	297	.byte 0xf0 # lock
5fabb88a AP	298	add $lastdiff,($out)
	299	jmp .Loop
	300	.align 16
	301	.Loop: rdtsc
	302	mov %eax,%edx
	303	sub $lasttick,%eax
	304	mov %edx,$lasttick
	305	mov %eax,$lastdiff
	306	clflush ($out)
b9064221	307	.byte 0xf0 # lock
5fabb88a AP	308	add %eax,($out)
	309	lea 4($out),$out
	310	sub \$1,$cnt
	311	jnz .Loop
	312
	313	mov $max,%rax
	314	ret
	315	.size OPENSSL_instrument_bus,.-OPENSSL_instrument_bus
	316
	317	.globl OPENSSL_instrument_bus2
	318	.type OPENSSL_instrument_bus2,\@abi-omnipotent
	319	.align 16
	320	OPENSSL_instrument_bus2:
	321	mov $arg1,$out # tribute to Win64
	322	mov $arg2,$cnt
	323	mov $arg3,$max
	324	mov $cnt,$redzone(%rsp)
	325
	326	rdtsc # collect 1st tick
	327	mov %eax,$lasttick # lasttick = tick
	328	mov \$0,$lastdiff # lastdiff = 0
	329
	330	clflush ($out)
b9064221	331	.byte 0xf0 # lock
5fabb88a AP	332	add $lastdiff,($out)
	333
	334	rdtsc # collect 1st diff
	335	mov %eax,%edx
	336	sub $lasttick,%eax # diff
	337	mov %edx,$lasttick # lasttick = tick
	338	mov %eax,$lastdiff # lastdiff = diff
	339	.Loop2:
	340	clflush ($out)
b9064221	341	.byte 0xf0 # lock
5fabb88a AP	342	add %eax,($out) # accumulate diff
	343
	344	sub \$1,$max
	345	jz .Ldone2
	346
	347	rdtsc
	348	mov %eax,%edx
	349	sub $lasttick,%eax # diff
	350	mov %edx,$lasttick # lasttick = tick
	351	cmp $lastdiff,%eax
	352	mov %eax,$lastdiff # lastdiff = diff
	353	mov \$0,%edx
	354	setne %dl
	355	sub %rdx,$cnt # conditional --$cnt
	356	lea ($out,%rdx,4),$out # conditional ++$out
	357	jnz .Loop2
	358
	359	.Ldone2:
	360	mov $redzone(%rsp),%rax
	361	sub $cnt,%rax
	362	ret
	363	.size OPENSSL_instrument_bus2,.-OPENSSL_instrument_bus2
	364	___
	365	}
aa8f38e4	366
301799b8 AP	367	print<<___;
	368	.globl OPENSSL_ia32_rdrand
	369	.type OPENSSL_ia32_rdrand,\@abi-omnipotent
	370	.align 16
	371	OPENSSL_ia32_rdrand:
	372	mov \$8,%ecx
	373	.Loop_rdrand:
	374	rdrand %rax
	375	jc .Lbreak_rdrand
	376	loop .Loop_rdrand
	377	.Lbreak_rdrand:
	378	cmp \$0,%rax
	379	cmove %rcx,%rax
	380	ret
4d01f276	381	.size OPENSSL_ia32_rdrand,.-OPENSSL_ia32_rdrand
f4d45640 AP	382
	383	.globl OPENSSL_ia32_rdseed
	384	.type OPENSSL_ia32_rdseed,\@abi-omnipotent
	385	.align 16
	386	OPENSSL_ia32_rdseed:
	387	mov \$8,%ecx
	388	.Loop_rdseed:
	389	rdseed %rax
	390	jc .Lbreak_rdseed
	391	loop .Loop_rdseed
	392	.Lbreak_rdseed:
	393	cmp \$0,%rax
	394	cmove %rcx,%rax
	395	ret
	396	.size OPENSSL_ia32_rdseed,.-OPENSSL_ia32_rdseed
301799b8 AP	397	___
301799b8 AP	398
5d863367	399	close STDOUT; # flush