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

#!/usr/bin/env perl

$output=shift;
$masm=1 if ($output =~ /\.asm/);
open STDOUT,">$output" || die "can't open $output: $!";

print<<___ if(defined($masm));
_TEXT	SEGMENT
PUBLIC	OPENSSL_rdtsc

PUBLIC	OPENSSL_atomic_add
ALIGN	16
OPENSSL_atomic_add	PROC
	mov	eax,DWORD PTR[rcx]
\$Lspin:	lea	r8,DWORD PTR[rdx+rax]
lock	cmpxchg	DWORD PTR[rcx],r8d
	jne	\$Lspin
	mov	eax,r8d
	cdqe    
	ret
OPENSSL_atomic_add	ENDP

PUBLIC	OPENSSL_wipe_cpu
ALIGN	16
OPENSSL_wipe_cpu	PROC
	pxor	xmm0,xmm0
	pxor	xmm1,xmm1
	pxor	xmm2,xmm2
	pxor	xmm3,xmm3
	pxor	xmm4,xmm4
	pxor	xmm5,xmm5
	xor	rcx,rcx
	xor	rdx,rdx
	xor	r8,r8
	xor	r9,r9
	xor	r10,r10
	xor	r11,r11
	lea	rax,QWORD PTR[rsp+8]
	ret
OPENSSL_wipe_cpu	ENDP
_TEXT	ENDS

CRT\$XIU	SEGMENT
EXTRN	OPENSSL_cpuid_setup:PROC
DQ	OPENSSL_cpuid_setup
CRT\$XIU	ENDS

___
print<<___ if(!defined($masm));
.text

.globl	OPENSSL_atomic_add
.type	OPENSSL_atomic_add,\@function
.align	16
OPENSSL_atomic_add:
	movl	(%rdi),%eax
.Lspin:	leaq	(%rsi,%rax),%r8
lock;	cmpxchgl	%r8d,(%rdi)
	jne	.Lspin
	movl	%r8d,%eax
	.byte	0x48,0x98
	ret
.size	OPENSSL_atomic_add,.-OPENSSL_atomic_add

.globl	OPENSSL_wipe_cpu
.type	OPENSSL_wipe_cpu,\@function
.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

.section	.init
	call	OPENSSL_cpuid_setup

___

open STDOUT,"| $^X perlasm/x86_64-xlate.pl $output";
print<<___;
.text

.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,\@abi-omnipotent
.align	16
OPENSSL_ia32_cpuid:
	mov	%rbx,%r8

	xor	%eax,%eax
	cpuid
	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

	mov	\$1,%eax
	cpuid
	cmp	\$0,%r9d
	jne	.Lnotintel
	or	\$0x00100000,%edx	# use reserved 20th bit to engage RC4_CHAR
	and	\$15,%ah
	cmp	\$15,%ah		# examine Family ID
	je	.Lnotintel
	or	\$0x40000000,%edx	# use reserved bit to skip unrolled loop
.Lnotintel:
	bt	\$28,%edx		# test hyper-threading bit
	jnc	.Ldone
	shr	\$16,%ebx
	cmp	\$1,%bl			# see if cache is shared
	ja	.Ldone
	and	\$0xefffffff,%edx	# ~(1<<28)
.Ldone:
	shl	\$32,%rcx
	mov	%edx,%eax
	mov	%r8,%rbx
	or	%rcx,%rax
	ret
.size	OPENSSL_ia32_cpuid,.-OPENSSL_ia32_cpuid
___
close STDOUT;	# flush
Commit	Line	Data
14e21f86 AP	1	#!/usr/bin/env perl
	2
	3	$output=shift;
0794f3a7	4	$masm=1 if ($output =~ /\.asm/);
14e21f86 AP	5	open STDOUT,">$output" \|\| die "can't open $output: $!";
14e21f86 AP	6
0794f3a7	7	print<<___ if(defined($masm));
376729e1	8	_TEXT SEGMENT
14e21f86	9	PUBLIC OPENSSL_rdtsc
376729e1 AP	10
	11	PUBLIC OPENSSL_atomic_add
	12	ALIGN 16
	13	OPENSSL_atomic_add PROC
	14	mov eax,DWORD PTR[rcx]
	15	\$Lspin: lea r8,DWORD PTR[rdx+rax]
	16	lock cmpxchg DWORD PTR[rcx],r8d
	17	jne \$Lspin
	18	mov eax,r8d
	19	cdqe
	20	ret
	21	OPENSSL_atomic_add ENDP
	22
	23	PUBLIC OPENSSL_wipe_cpu
	24	ALIGN 16
	25	OPENSSL_wipe_cpu PROC
	26	pxor xmm0,xmm0
	27	pxor xmm1,xmm1
	28	pxor xmm2,xmm2
	29	pxor xmm3,xmm3
	30	pxor xmm4,xmm4
	31	pxor xmm5,xmm5
	32	xor rcx,rcx
	33	xor rdx,rdx
	34	xor r8,r8
	35	xor r9,r9
	36	xor r10,r10
	37	xor r11,r11
	38	lea rax,QWORD PTR[rsp+8]
	39	ret
	40	OPENSSL_wipe_cpu ENDP
376729e1 AP	41	_TEXT ENDS
	42
	43	CRT\$XIU SEGMENT
	44	EXTRN OPENSSL_cpuid_setup:PROC
	45	DQ OPENSSL_cpuid_setup
	46	CRT\$XIU ENDS
0794f3a7	47
14e21f86	48	___
0794f3a7	49	print<<___ if(!defined($masm));
14e21f86	50	.text
376729e1 AP	51
	52	.globl OPENSSL_atomic_add
	53	.type OPENSSL_atomic_add,\@function
	54	.align 16
	55	OPENSSL_atomic_add:
	56	movl (%rdi),%eax
8d56c8b5 AP	57	.Lspin: leaq (%rsi,%rax),%r8
8d56c8b5 AP	58	lock; cmpxchgl %r8d,(%rdi)
376729e1	59	jne .Lspin
8d56c8b5 AP	60	movl %r8d,%eax
8d56c8b5 AP	61	.byte 0x48,0x98
376729e1 AP	62	ret
	63	.size OPENSSL_atomic_add,.-OPENSSL_atomic_add
	64
	65	.globl OPENSSL_wipe_cpu
	66	.type OPENSSL_wipe_cpu,\@function
	67	.align 16
	68	OPENSSL_wipe_cpu:
	69	pxor %xmm0,%xmm0
	70	pxor %xmm1,%xmm1
	71	pxor %xmm2,%xmm2
	72	pxor %xmm3,%xmm3
	73	pxor %xmm4,%xmm4
	74	pxor %xmm5,%xmm5
	75	pxor %xmm6,%xmm6
	76	pxor %xmm7,%xmm7
	77	pxor %xmm8,%xmm8
	78	pxor %xmm9,%xmm9
	79	pxor %xmm10,%xmm10
	80	pxor %xmm11,%xmm11
	81	pxor %xmm12,%xmm12
	82	pxor %xmm13,%xmm13
	83	pxor %xmm14,%xmm14
	84	pxor %xmm15,%xmm15
8d56c8b5 AP	85	xorq %rcx,%rcx
	86	xorq %rdx,%rdx
	87	xorq %rsi,%rsi
	88	xorq %rdi,%rdi
	89	xorq %r8,%r8
	90	xorq %r9,%r9
	91	xorq %r10,%r10
	92	xorq %r11,%r11
	93	leaq 8(%rsp),%rax
376729e1 AP	94	ret
	95	.size OPENSSL_wipe_cpu,.-OPENSSL_wipe_cpu
	96
0794f3a7 AP	97	.section .init
	98	call OPENSSL_cpuid_setup
	99
	100	___
	101
	102	open STDOUT,"\| $^X perlasm/x86_64-xlate.pl $output";
	103	print<<___;
	104	.text
	105
	106	.globl OPENSSL_rdtsc
	107	.type OPENSSL_rdtsc,\@abi-omnipotent
	108	.align 16
	109	OPENSSL_rdtsc:
	110	rdtsc
	111	shl \$32,%rdx
	112	or %rdx,%rax
	113	ret
	114	.size OPENSSL_rdtsc,.-OPENSSL_rdtsc
	115
376729e1	116	.globl OPENSSL_ia32_cpuid
0794f3a7	117	.type OPENSSL_ia32_cpuid,\@abi-omnipotent
376729e1 AP	118	.align 16
376729e1 AP	119	OPENSSL_ia32_cpuid:
0794f3a7 AP	120	mov %rbx,%r8
	121
	122	xor %eax,%eax
	123	cpuid
	124	xor %eax,%eax
	125	cmp \$0x756e6547,%ebx # "Genu"
	126	setne %al
	127	mov %eax,%r9d
	128	cmp \$0x49656e69,%edx # "ineI"
	129	setne %al
	130	or %eax,%r9d
	131	cmp \$0x6c65746e,%ecx # "ntel"
	132	setne %al
	133	or %eax,%r9d
	134
	135	mov \$1,%eax
376729e1	136	cpuid
0794f3a7 AP	137	cmp \$0,%r9d
	138	jne .Lnotintel
	139	or \$0x00100000,%edx # use reserved 20th bit to engage RC4_CHAR
	140	and \$15,%ah
	141	cmp \$15,%ah # examine Family ID
	142	je .Lnotintel
	143	or \$0x40000000,%edx # use reserved bit to skip unrolled loop
	144	.Lnotintel:
	145	bt \$28,%edx # test hyper-threading bit
	146	jnc .Ldone
	147	shr \$16,%ebx
	148	cmp \$1,%bl # see if cache is shared
	149	ja .Ldone
	150	and \$0xefffffff,%edx # ~(1<<28)
	151	.Ldone:
	152	shl \$32,%rcx
	153	mov %edx,%eax
	154	mov %r8,%rbx
	155	or %rcx,%rax
376729e1 AP	156	ret
376729e1 AP	157	.size OPENSSL_ia32_cpuid,.-OPENSSL_ia32_cpuid
14e21f86	158	___
0794f3a7	159	close STDOUT; # flush